JP2003209572A - Device for allocating band - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a band allocating device capable of observing the allocation of an insured band to traffic that changes like a burst, and suppressing buffer overflow in an OUT, an increase in a round trip time due to allocation delay, the allocation of an unreasonably small amount of band to a customer who sporadically uses the band, the occurrence of an invalid band, or the like. <P>SOLUTION: This band allocating device for dynamically allocating at least one portion of a shared band to each line is provided with: a traffic observing means 10 for observing the use situation of the band allocated to each line; a band request receiving means 20 for receiving the declared value of an allocation request band periodically inputted to each line; a band predicting means 30 for predicting the allocation request band of each line on the basis of at least either the declared value of the used band of each line detected by the traffic observing means 10 or the declared value of the allocation request band received by the band request receiving means 20; and a band allocating means 40 for allocating a band corresponding to a result predicted by the band predicting means 30 to each line. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a bandwidth allocating apparatus and a bandwidth allocating method, for example, DBA (Dynamic Bandwidth Allocation: Dyna).
B-PON (Broadb) compatible with mic Bandwidth Assignment
and Passive Optical Network) and is used for a device that dynamically allocates a shared band to each of a plurality of lines.

[0002]

2. Description of the Related Art In a B-PON communication system to which DBA is applied, the upstream information of a customer is received in the customer's house,
ONT (Opti that transmits the received customer's upstream information to the network side
cal network terminal) is installed, and OLT (optical subscriber line terminal equipment:
Optical Line Terminal) is supposed to be installed. The B-PON connects the ONT and the OLT with an optical fiber and an optical branching device.

The OLT has an O having an upstream band to be transmitted.
For NT, line from ONT for each customer or T-CO
Bands are dynamically allocated based on the allocation request band declaration for each NT (Transmission Container) or the used band observed in the OLT.

[0004]

When the ONT repeatedly declares the allocation request bandwidth to the OLT, there is a limit to the bandwidth that can be used for the declaration, so the declaration is usually repeated at a predetermined interval. . For this reason, the allocation request bandwidth declared by the ONT is actually information at a certain point in the past, not current information.

Accordingly, for example, TCP (Transmission Con)
In the case where the traffic whose transmission band changes in a burst manner appears on the line like the communication using the control protocol), there is a large difference between the allocation request band declared by ONT and the band currently required. Is likely to occur.

Further, even when the OLT tries to grasp the band by observing the band actually used by each ONT, the result obtained by continuing the observation for a certain period of time is detected as the used band. Therefore, it is possible to understand only the bandwidth used a little before the present time. Therefore, when the true allocation request bandwidth of the customer is rapidly increased, a delay occurs until the change in the customer traffic arriving at the ONT is actually detected by the ONT, so that the bandwidth allocation is delayed.

In some cases, a customer and a telecommunications carrier may make a contract to guarantee bandwidth allocation to each customer with the required bandwidth as an upper limit. However, if the delay in the actual bandwidth allocation is significant, the problem of failing to comply with the guaranteed bandwidth contract occurs. Further, if the allocation of the guaranteed bandwidth is significantly delayed, the buffer overflows in the ONT even if the input is less than the guaranteed bandwidth. Further, if the round trip time increases due to the bandwidth allocation delay, there is a problem that the TCP traffic cannot fully utilize the wide band.

Further, when there are customers who use a band sporadically and customers who use a wide band continuously, the bandwidth allocation to the former customer is smaller than that of the latter customer, which causes unfairness. There is also. That is, assuming that the bandwidth is allocated only for the queue length that is the value declared by the ONT, it is premised that cells received from the network under the control of the ONT are always accumulated in the buffer.

[0009] This not only causes an increase in storage delay time, but may cause an unfairness among customers, such as a disadvantage for TCP traffic customers as compared to UDP (User Datagram Protocol) traffic customers. Means

As an example of unfairness, for a plurality of customers who constantly send out UDP traffic for the maximum bandwidth,
There may be one TCP traffic customer competing. U
In the case of DP, there is no feedback from the traffic receiving terminal to the transmitting terminal, so that the customer can send a packet regardless of the availability of the shared band.

For this reason, in UDP, the cells more than the guaranteed bandwidth are always accumulated in the ONT queue, and the guaranteed bandwidth is always secured. On the other hand, TC
P takes the process of increasing the TCP window size by slow start. That is, the transmission band is suppressed by feeding back from the receiving terminal to the transmitting terminal for the purpose of avoiding congestion on the way route and reducing packet loss.

As a result, when the UDP traffic occupies the shared band, the TCP traffic customer who newly starts sending the traffic cannot increase the sending traffic and can secure the guaranteed band. Disappear. That is, it seems that the contracted bandwidth is not protected by the TCP traffic customer. Further, in order to comply with the guaranteed bandwidth, there is a method of fixedly allocating the guaranteed bandwidth irrespective of the amount of allocation request bandwidth. However, when the guaranteed bandwidth is fixedly allocated, the bandwidth is continuously wasted while the customer does not use the allocated bandwidth. Further, this method has a problem in that fairness among customers regarding allocation of a band exceeding the guaranteed band cannot be secured.

The present invention adheres to guaranteed bandwidth allocation for traffic that changes in bursts, suppresses buffer overflow at the ONT, suppresses an increase in round trip time due to allocation delay, and improperly applies to customers who use the bandwidth sporadically. An object of the present invention is to provide a band allocation device and a band allocation method capable of suppressing a small amount of band allocation and suppressing the generation of an invalid band.

[0014]

According to a first aspect of the present invention, in a band allocating device for dynamically allocating at least a part of a shared band shared among a plurality of lines to each line, use of a band allocated to each line is used. Traffic observing means for observing the situation, band request receiving means for receiving the declared value of the allocation request bandwidth input for each line, use bandwidth of each line detected by the traffic observing means, and reception of the band request receiving means A band predicting unit for predicting the request band for allocation of each line based on at least one of the declared values of the requested band for allocation, and a band allocating unit for allocating a band corresponding to the prediction result of the band predicting unit for each line are provided. It is characterized by that.

By observing the use status of the band allocated to the line by using the traffic observing means, it is possible to grasp the change or tendency of the actual use band. Also,
By monitoring the reported value of the allocation request bandwidth received by the bandwidth request receiving means, it is possible to grasp the change and tendency of the declared value. The band predicting unit predicts the current allocation request band of each line based on at least one of the used band of each line detected by the traffic observing unit and the declared value of the allocation request band received by the band request receiving unit. .

The allocation request bandwidth received as the declared value is
Considering the delay caused by the reception of it, it is the declared value slightly before the present time. The observed value detected by observing the traffic during a certain period is also an averaged observed value at a time point slightly before the present. Therefore, the declared value and the observed value do not always match the band currently required.

However, by monitoring the observed value of the traffic observing means and the declared value received by the request receiving means, the change or tendency of the actual bandwidth used (traffic characteristics)
Therefore, it is possible to predict the allocation request bandwidth at the present time from the observed values or declared values at a plurality of past times using, for example, linear prediction.

By allocating the band in accordance with the predicted allocation request band, it becomes possible to reduce the delay time of the band allocation control and reduce the allocation of the invalid band which is not actually used. According to a second aspect of the present invention, in the band allocating device according to the first aspect, the band predicting means is configured to perform linear prediction.
And second predicting means for performing non-linear prediction,
Used band identification means for obtaining the ratio of all of the allocated band used based on the history of past observation values for each line and comparing the input band with the declared value of the input band and use of the allocated band Based on at least one of the history of the situation, excess identification means for obtaining the ratio of the declared value exceeding the predicted value of the first prediction means for each line and comparing it with a second threshold, and identification of the used band identification means Selection means for selecting one of the prediction value of the first prediction means and the prediction value of the second prediction means based on the result and the identification result of the excess identification means. And

By performing the linear prediction, it is possible to relatively accurately predict a stationary allocation request band with a gradual change. Also, by performing non-linear prediction using a step function, it is possible to obtain a prediction value corresponding to a burst-like change in the allocation request bandwidth. The following effects can be expected by using these two types of prediction in combination.

(1) Observing guaranteed bandwidth allocation for bursty traffic (2) Suppressing buffer overflow at ONT (3) Suppressing increase in round trip time due to allocation delay (4) Sporadic use of bandwidth Suppressing unreasonably small bandwidth allocation to the customer (5) Suppressing invalid bandwidth The selecting means, based on the identification result of the used bandwidth identifying means and the identification result of the excess identifying means, the predicted value of the first predicting means. It is possible to appropriately select either one of the prediction value of the second prediction means and the prediction value of the second prediction means.

According to a third aspect of the present invention, in the bandwidth allocating apparatus according to the second aspect, the selecting means has the used bandwidth identifying means and the excess identifying means from a time when a prediction is started until a predetermined time period or a predetermined number of times elapses. The prediction value of the second prediction means is selected regardless of the identification result of. In order to make a linear prediction, past data of at least two time points are required. However, immediately after the declared value of the bandwidth allocation request and / or the observed value of the used bandwidth has changed from 0 to a finite value other than 0, there is not enough past data for linear prediction, so the linear predicted value is unreasonably low. The possibility cannot be denied. Therefore, immediately after the declared value of the band allocation request and / or the observed value of the used band becomes a finite value from 0 to a non-zero value, the predicted value of the second prediction means is used for predicting on the safe side.

According to a fourth aspect of the present invention, in the bandwidth allocating device according to the second or third aspect, the second predicting means includes a guaranteed bandwidth of a line, a current used bandwidth, a current allocated bandwidth and the first predicting means. The feature is that the maximum value among the predicted values is selected and output as the predicted value. By adopting the maximum value of the guaranteed bandwidth of the line, the currently used bandwidth, the current allocated bandwidth, and the linearly predicted value as the predicted value, it is possible to comply with the guaranteed bandwidth allocation even for traffic that changes in bursts. Become.

According to a fifth aspect of the present invention, in the bandwidth allocating device according to the second or third aspect, there is further provided guaranteed band identifying means for comparing the previously predicted advance predicted value with a threshold of the guaranteed band of the line, and If the preceding prediction value is less than or equal to the guaranteed bandwidth of the line, the prediction means of No. 2 is the maximum of the guaranteed bandwidth of the line, the currently used bandwidth, the current allocated bandwidth, and the predicted value of the first prediction means. If a value is selected as the predicted value and the preceding predicted value exceeds the guaranteed bandwidth of the line, the available shared band is divided by the total number of lines or the guaranteed bandwidth ratio of all lines, and the current used band, the current band The maximum value among the allocated bandwidth and the predicted value of the first prediction means is selected as the predicted value.

If the preceding predicted value (for example, the previously calculated predicted value) is less than or equal to the guaranteed bandwidth of the line, the guaranteed bandwidth of the line,
By selecting the maximum value among the currently used bandwidth, the currently allocated bandwidth, and the linear prediction value as the prediction value, it becomes possible to comply with the guaranteed bandwidth allocation even for the traffic that changes in a burst.

If the preceding predicted value exceeds the guaranteed bandwidth of the line, the available shared bandwidth is divided by the total number of lines or the guaranteed bandwidth ratio of all the lines, the current used band, the current allocated band and By selecting the maximum value of the linear prediction values as the prediction value, it becomes possible to appropriately allocate the band even to the traffic that changes in a burst and to evenly allocate the shared band to each line.

According to a sixth aspect of the present invention, in the bandwidth allocating device according to the second or third aspect, the first predicting means is a predetermined value for the declared value of the used bandwidth or the allocation request bandwidth observed at at least two points in the past. The present invention is characterized in that a value obtained by multiplying a coefficient is predicted as the current allocation request bandwidth. Based on the data of the past two time points, linear prediction can be applied to predict the current allocation request bandwidth.

According to a seventh aspect of the present invention, in the bandwidth allocating apparatus according to the second or third aspect, the first predicting means is linear to a used bandwidth observed at at least two past times or a declared allocation request bandwidth. Characteristic is that the result of adding the proportional value and the value obtained by multiplying the previously declared allocation request bandwidth by a coefficient or the value obtained by multiplying the increase bandwidth of the previous declared bandwidth with respect to the pre-previous round declared bandwidth by a coefficient is used as the current predicted value. And

In the present invention, the observation of the used band, the reception of the declared value, the prediction of the band and the band allocation are repeated in a constant time period (T), and the used band has a constant rate in proportion to the elapsed time. It is expected to increase or decrease in. By using a value that is linearly proportional to the observed value or the declared value of the band at the past two points in time, the currently required band can be appropriately determined in response to the gradual change in traffic. Also, by using a multiple of the declared value or a multiple of the bandwidth increase of the declared value per one cycle (T),
It is possible to cope with a burst increase in the allocation request bandwidth without delay.

According to an eighth aspect, in the bandwidth allocating device according to the seventh aspect, after the declared value is transmitted by ONT, the declared value becomes O.
It is characterized in that the number of control cycles corresponding to the time required to be reflected in the band allocation in LT is applied as the coefficient. For example, supposing that the time required for the declared value to be reflected from the ONT until it is reflected in the bandwidth allocation in the OLT is (2T), the declared value used for prediction actually allocates the bandwidth. Since the value is (2T) hours before the time point, it is necessary to reflect the change in the declared value corresponding to the (2T) time in the predicted value. That is, when the required time is (2T), a value that is twice the amount of change in the declared value per cycle may be determined as the predicted value.

Further, the bandwidth allocation methods of claims 9 to 16 have the same characteristics as those of claims 1 to 8, respectively. A seventeenth aspect is the band allocation device according to the first aspect,
The band predicting unit identifies whether or not there is band allocation of the band allocating unit for the observation period of the line observed by the traffic observing unit, and if there is no band allocation, use of the band of the corresponding line The feature is that the prediction is performed assuming that the situation is other than zero.

When the bandwidth allocation is repeatedly performed at regular intervals, the bandwidth allocation to a certain line is temporarily 0.
It also occurs when. When the band allocation becomes 0, the traffic observation value also becomes 0, and the band prediction value may become 0. That is, the predicted value fluctuates unnecessarily, and band allocation is adversely affected. According to the seventeenth aspect, when there is no band allocation, the prediction is performed assuming that the band usage status is other than 0, so that unnecessary fluctuation of the prediction value can be prevented.

A thirteenth aspect of the present invention is the band allocation method according to the ninth aspect, wherein the band prediction means determines whether or not there is band allocation in the band allocation procedure for an observation period of a line observed in the traffic observation procedure. Is identified, and when there is no band allocation, the band usage status of the corresponding line is considered to be other than 0 and prediction is performed.

According to the eighteenth aspect, similarly to the seventeenth aspect, when the band is not allocated, the prediction is performed assuming that the band usage status is other than 0, so that unnecessary fluctuation of the prediction value can be prevented. . According to a nineteenth aspect of the present invention, in the band allocating device according to the first aspect, while the traffic observing means has a zero band allocation of the band allocating means for the observed line, the traffic observing means is based on the usage status of the band previously observed. And outputs an observation value other than 0.

In claim 19, while the bandwidth allocation is 0, the traffic observing means outputs an observation value other than 0 based on the usage status of the band observed before that, so that the prediction value is unnecessary. It can prevent fluctuation. According to a twentieth aspect of the present invention, in the bandwidth allocating method according to the ninth aspect, in the traffic observing procedure, while the bandwidth allocating of the bandwidth allocating procedure for the observed line is 0, based on the usage status of the bandwidth previously observed. And outputs an observation value other than 0.

In the twentieth aspect, as in the nineteenth aspect, while the band allocation is 0, an observation value other than 0 is output in the traffic observation procedure based on the usage status of the band observed before that. , It is possible to prevent unnecessary fluctuation of the predicted value.
According to a twenty-first aspect of the present invention, in the bandwidth allocating device according to the first aspect, the band allocating means allocates a band of one cell or more in addition to the band corresponding to the prediction result in a period for detecting an increase in the allocation request band. Alternatively, the smaller one of the result of adding one or more cells to the band corresponding to the prediction result and the guaranteed band of the line, whichever is smaller, is selectively allocated.

According to a twenty-second aspect of the present invention, in the band allocating apparatus according to the first aspect, the band predicting means is the band allocation means when the actual allocation band is less than the prediction band predicted as the band to be allocated to the line. Considering that the amount of information corresponding to the band obtained by subtracting the actual allocated band from the predicted band is retained in the buffer in the allocated device as a queue,
It is characterized in that a band corresponding to a difference between the actual allocated band and the predicted band is predicted to decrease from a band used by the line.

A thirty-third aspect of the present invention is the band allocating device according to the first aspect, wherein the band predicting means outputs a second prediction band obtained by multiplying the first prediction band of the prediction result by a predetermined band prediction coefficient. At the same time, when the second prediction band does not have a band increase of 1 cell or more per first period for detecting an increase in the allocation request band as compared with the first prediction band, the first prediction It is characterized in that a result obtained by adding one cell to the band per the first period is output as a prediction result.

A thirty-fourth aspect of the present invention is the bandwidth allocation device according to the first aspect, wherein the band prediction means multiplies at least one of a previously observed used band and a reported value of the allocated band in order to predict a change in the band. While holding the coefficient of, the utilization rate of the bandwidth observed by the traffic observing means is a predetermined value or more, whether the bandwidth observed by the traffic observing means is an increasing tendency or a decreasing tendency, The plurality of coefficients are selected and predicted according to the result of identifying at least one condition of the increasing tendency and whether or not the current allocated bandwidth exceeds a predetermined value and whether or not the declared value is 0. It is characterized by performing.

In a twenty-fifth aspect of the present invention, in the bandwidth allocation device of the first aspect, the bandwidth request receiving means normalizes the declared value for each line according to the reporting cycle, and the traffic observing means allocates the observed value to the observation cycle or the bandwidth allocation. Normalize according to the cycle,
The bandwidth predicting means predicts the allocation request bandwidth of each line based on the normalized declared value and the observed value.

A twenty-sixth aspect of the present invention is the bandwidth allocating device according to the second aspect, wherein the second predicting means is a line having a positive predicted value of a bandwidth allocation request for the entire usable shared band or the entire usable shared band. Bandwidth divided by a number, or the total available shared bandwidth is the guaranteed bandwidth ratio of the line for which the predicted value of the bandwidth request is positive, or the entire available shared bandwidth is the guaranteed bandwidth ratio of all the lines, and the current allocated bandwidth, The present invention is characterized in that the maximum value is selected and output from the currently used band and the predicted value of the first prediction means.

A twenty-seventh aspect of the present invention is the bandwidth allocating apparatus according to the fifth aspect, wherein the second predicting means is the same as the case where the preceding prediction value is 0 when the preceding prediction value is 0. It is characterized by selecting a predicted value. According to a twenty-eighth aspect of the present invention, in the band allocating apparatus according to the first aspect, when the predicted value of the band prediction means or the increased predicted value exceeds the maximum band allowed for the line, the predicted value is replaced with the maximum band. It is characterized by

A thirty-ninth aspect of the present invention is the band allocation device according to the first aspect, wherein the band predicting means averages the used bands when the present used band is within a predetermined deviation from the average of the past used bands. The feature is that the band is predicted so that the value becomes a value. According to a thirtieth aspect of the present invention, in the bandwidth allocating device according to the second aspect, the prediction means is configured such that the actual allocation bandwidth allocated by the allocation means is less than the allocation bandwidth predicted to be allocated by the prediction means. A value obtained by adding a value corresponding to the difference between the actual allocated band and the predicted allocated band to the declared value received by the request band receiving unit, and the difference between the actual allocated band and the predicted allocated band. The predicted value using at least one of the values obtained by subtracting the value corresponding to the above from the usage status of the line observed by the traffic observing means, and the predicted value by the first predicting means, and comparing the predicted value by the first predicting means. And a predicted value by the second predicting means are selected.

According to a thirty-first aspect of the present invention, in the band allocating apparatus according to the second aspect, the first predicting means uses a linear predictive function used by the linear predicting means from a value obtained by multiplying a previous used bandwidth by a coefficient, and a second used bandwidth. It is characterized in that a value obtained by subtracting a value obtained by multiplying by is multiplied by a band increase coefficient is used.

A thirty-second aspect of the present invention is the band allocating device according to the first aspect, wherein the band predicting means converts a calculated value obtained by multiplying the allocation request band by a predetermined coefficient into the number of cells per cycle of the allocation request declaration, When the length of the buffer of the allocated device is exceeded, the calculated value is replaced with the buffer length of the allocated device per cycle of the allocation request declaration.

A thirty-third aspect of the present invention is the band allocating method according to the ninth aspect, wherein in the band allocating procedure, at least one cell is added in addition to the band corresponding to the prediction result during a period for detecting an increase in the allocation request band. It is characterized by allocating a band or selectively allocating a smaller band of a result obtained by adding a band of one cell or more to a band corresponding to the prediction result and a guaranteed band of the line.

A thirty-fourth aspect of the present invention is the band allocating method according to the ninth aspect, wherein in the band predicting procedure, when the actual band allocated by the band allocating procedure is less than the predicted band predicted as the band to be allocated to the line, It is considered that the amount of information corresponding to a band obtained by subtracting the actual allocation band from the prediction band is accumulated in the buffer in the allocated device as a queue, and the difference between the actual allocation band and the prediction band from the band used by the line. It is characterized by predicting that the band corresponding to is reduced.

A thirty-fifth aspect of the present invention is the band allocation method according to the ninth aspect, wherein in the band prediction procedure, a second prediction band obtained by multiplying the first prediction band of the prediction result by a predetermined band prediction coefficient is output. And the second prediction band is the first
When there is no band increase of 1 cell or more per first period for detecting the increase of the allocation request band as compared with the predicted band of 1), 1 cell per 1st period is added to the 1st predicted band. It is characterized in that the added result is output as a prediction result.

A thirty-sixth aspect, in the bandwidth allocating method according to the ninth aspect, holds a plurality of coefficients for multiplying at least one of the used bandwidth and the declared value of the allotted bandwidth observed in the past in order to predict a change in the bandwidth. In the bandwidth prediction procedure, the usage rate of the bandwidth observed in the traffic observation procedure is a predetermined value or more, whether the bandwidth observed in the traffic observation procedure is increasing or decreasing tendency, the increase of the used bandwidth The plurality of coefficients are selected and predicted according to the tendency and the result of identifying at least one condition of whether or not the current allocated bandwidth exceeds a predetermined value and whether or not the declared value is 0. It is characterized by

A thirty-seventh aspect of the present invention is the bandwidth allocating method according to the ninth aspect, wherein in the bandwidth request receiving procedure, the declared value for each line is normalized in accordance with the reporting cycle, and in the traffic observation procedure, the observed value is allocated to the observation cycle or the bandwidth allocation. It is characterized in that it is normalized according to the cycle, and the bandwidth requesting procedure predicts the allocation request bandwidth of each line based on the normalized declared value and observed value.

According to a thirty-eighth aspect of the present invention, in the bandwidth allocating method according to the tenth aspect, in the second predicting procedure, the entire usable shared band or the entire usable shared band is a line for which the predicted value of the bandwidth allocation request is positive. Bandwidth divided by a number, or the total available shared bandwidth is the guaranteed bandwidth ratio of the line for which the predicted value of the bandwidth request is positive, or the entire available shared bandwidth is the guaranteed bandwidth ratio of all the lines, and the current allocated bandwidth, The present invention is characterized in that the maximum value is selected and output from the currently used band and the predicted value of the first prediction means.

A thirty-ninth aspect of the present invention is the band allocation method according to the thirteenth aspect, wherein in the second prediction procedure, when the preceding prediction value is 0, it is the same as when the preceding prediction value exceeds the guaranteed bandwidth of the line. It is characterized by selecting a predicted value. According to a 40th aspect of the present invention, in the bandwidth allocating method according to the 9th aspect, when the predicted value in the bandwidth prediction procedure or the increased predicted value exceeds a maximum bandwidth allowed for a line, the predicted value is set to the maximum bandwidth. It is characterized by replacing.

According to a forty-first aspect of the present invention, in the band allocation method according to the ninth aspect, in the band prediction procedure, when the currently used band is within a predetermined deviation from the average of the past used band, the average of the used band is used. The feature is that the band is predicted so that the value becomes a value.

A 42nd aspect of the present invention is the bandwidth allocating method according to the 10th aspect, wherein in the predicting procedure, the actual allocated bandwidth allocated by the allocating procedure is less than the allocated bandwidth predicted to be allocated in the predicting procedure. In this case, a value obtained by adding a value corresponding to the difference between the actual allocated band and the predicted allocated band to the declared value received in the request band reception procedure, and the actual allocated band and the predicted allocated band. And comparing the predicted value using at least one of the values obtained by subtracting the value corresponding to the difference from the usage status of the line observed in the traffic observation procedure with the predicted value obtained in the first prediction procedure. It is characterized in that the prediction value according to the prediction procedure and the prediction value according to the second prediction procedure are selected.

A thirty-third aspect of the present invention is a band allocating method according to the tenth aspect, wherein, as a linear prediction function used in the linear prediction procedure, a value obtained by multiplying a previously used band by a coefficient by a coefficient is used. It is characterized in that a value obtained by multiplying the subtracted value by a band increase coefficient is used. According to a 44th aspect of the present invention, in the bandwidth allocating method according to the 9th aspect, in the bandwidth predicting procedure, a calculated value obtained by multiplying the allocation request bandwidth by a predetermined coefficient is converted into a number of cells per cycle of the allocation request declaration, and the allocated apparatus In the case where the length of the buffer owned by the device exceeds the buffer length, the calculated value is replaced by the buffer length of the allocated device per cycle of the allocation request declaration.

A forty-fifth aspect is characterized in that, in the bandwidth allocating device according to the second aspect, the first predicting means compares the predicted value with the declared value of the allocation request band and outputs the larger one as a linear predicted value. And According to a 46th aspect of the present invention, in the bandwidth allocation method according to the 10th aspect, in the first prediction procedure, the predicted value is compared with the declared value of the allocation request bandwidth, and the larger one is output as a linear predicted value.

[0056]

BEST MODE FOR CARRYING OUT THE INVENTION (First Embodiment) One embodiment of a band allocation apparatus and a band allocation method of the present invention will be described with reference to FIGS. This form corresponds to claims 1 and 9. FIG. 1 is a block diagram showing the configuration of the band allocation device of this embodiment. FIG. 2 is a flowchart showing an operation example (1) of the band prediction unit in this embodiment. FIG. 3 is a flowchart showing an operation example (2) of the band prediction unit in this embodiment. FIG. 4 is a flowchart showing an operation example (3) of the band prediction unit in this embodiment.

In this embodiment, the traffic observing means, the band request receiving means, the band predicting means and the band allocating means of claim 1 are the traffic observing section 10, the band request receiving section 20, the band predicting section 30 and the band allocating section 40, respectively. Equivalent to.
For example, when a telecommunications carrier uses a communication system such as B-PON to provide communication lines to a large number of customers, the bandwidth available on each customer's line is dynamically allocated to each customer (each line). Is desirable. In that case, a common band that can be allocated to any line is provided, and a part of the common band is dynamically allocated to each line. The band allocation device shown in FIG. 1 performs such dynamic band allocation.

As shown in FIG. 1, the band allocating device comprises a traffic observing section 10, a band request receiving section 20, a band predicting section 30, and a band allocating section 40. Band allocation unit 40
Allocates a band for each line according to the size of the band predicted by the band predicting unit 30 (the band currently required). The traffic observing unit 10 includes a band allocating unit 4 for each line.
The bandwidth allocated by 0 and the amount of traffic actually flowing through the line, that is, the usage status of the bandwidth are observed.

The bandwidth request receiving unit 20 receives the declared value of the allocation request bandwidth for each line, which is input by the customer or the like. The band prediction unit 30 predicts the current allocation request band based on at least one of the used band information obtained by the observation of the traffic observation unit 10 and the declared value of the allocation request band received by the band request receiving unit 20. .

Since there is a control time delay in the declared value of the used bandwidth obtained by the observation of the traffic observing section 10 and the allocation request bandwidth received by the bandwidth request receiving section 20, a time point slightly before (past) the present time It reflects the size of the bandwidth needed in. In order to reduce the delay in dynamic band allocation due to such a delay, the band prediction unit 30 performs prediction based on past information and obtains the size of the band required at the present time. The band predicted by the band prediction unit 30 is allocated to each line by the band allocation unit 40.

As a concrete operation of the band predicting section 30, an operation as shown in FIG. 2, FIG. 3 or FIG. 4 can be considered. In the operation example of FIG. 2, it is assumed that the prediction is performed using only the observation value of the traffic observation unit 10. Further, in the operation example of FIG. 3, it is assumed that the prediction is performed by using both the observed value and the declared value, and in the operation example of FIG. 4, the case is estimated that only the declared value is used for the prediction.

First, a description will be given with reference to FIG. The band is represented by the number of cells per unit time. Also, this processing is performed for each line. First step S1
At 1, 0 is set to the predicted value as the initial value. In the next step S12, the observation value (used band) of the traffic observation unit 10 is referred to, and it is discriminated whether or not the band of one cell or more is used. When the band of one cell or more is used, the process proceeds to the next step S13.

In step S13, the traffic observation unit 1
Calculate the predicted value of the band currently required based on the observed value of 0 (used band just before). In step S14, whether or not an unused band exists in the allocated band is identified for each line with reference to the observation value of the traffic observation unit 10. If all the allocated bandwidth is used,
The process proceeds from step S14 to S16, and correction is performed so as to increase the predicted value.

If there is an unused band in the allocated band, the process proceeds from step S14 to S15. In step S15, the observation value of the traffic observation unit 10 is referred to,
Whether or not the actual bandwidth used is 0 is identified for each line. If the used bandwidth is not 0, the process proceeds from step S15 to S17,
Correct to reduce the predicted value. If the band is not used at all, the process returns from step S15 to S11.

In the calculation of the predicted value in step S13 and the correction processing in steps S16 and S17, as shown in FIG.
It is conceivable to apply various conditions as shown in (1) to (8) to obtain the predicted value. Next, description will be made with reference to FIG. The band is represented by the number of cells as a unit. Also, this processing is performed for each line. In the first step S21, 0 is set to the predicted value as the initial value.

In the next step S22, reference is made to the declared value of the band received by the band request receiving unit 20 to identify whether or not there is a band allocation request for one or more cells. If there is a bandwidth allocation request for one cell or more, the process proceeds to the next step S23. In step S23, the predicted value of the band currently required is calculated based on the observed value of the traffic observing unit 10 (used band just before) and the declared value of the band received by the band request receiving unit 20.

In step S24, the traffic observation unit 1
By referring to the observed value of 0, whether or not an unused band exists in the allocated band is identified for each line. If all the allocated bands are used, steps S24 to S2
Proceed to step 6 and make corrections so that the predicted value increases. If there is an unused band in the allocated band, step S
It progresses from 24 to S25. In step S25, the observation value of the traffic observing unit 10 is referred to, and whether or not the actual bandwidth used is 0 is identified for each line.

If the used bandwidth is not 0, step S25
From S to S27, the predicted value is corrected so as to decrease.
If the band is not used at all, step S
It returns from 25 to S21. In the calculation of the predicted value in step S23 and the correction processing in steps S26 and S27, it is conceivable to apply various conditions as shown in (1) to (8) in FIG. 2 to obtain the predicted value.

Next, description will be made with reference to FIG. The band is represented by the number of cells as a unit. Also, this processing is performed for each line. In the first step S31, 0 is set to the predicted value as the initial value. In the next step S32, the declared value of the band received by the band request receiving unit 20 is referred to, and it is discriminated whether or not there is a band allocation request for one or more cells. If there is a bandwidth allocation request for one cell or more, the process proceeds to the next step S33.

In step S33, the bandwidth request receiving unit 20
Calculate the expected value of the band required at present based on the reported value of the band received by. In step S34, the tendency of the change in the declared value is checked from the history of the declared value received by the bandwidth request receiving unit 20, and it is discriminated for each line whether or not there is a decreasing tendency.
When the change in the declared value tends to increase, the process proceeds from step S34 to S35, and the predicted value is corrected so as to increase.

If the change in the declared value tends to decrease, the process proceeds from step S34 to S36, and the predicted value is corrected so as to decrease. In step S37, the history of the declared values received by the bandwidth request receiving unit 20 is checked to identify whether the declared values of 0 appear consecutively. If the declared value of 0 appears consecutively, the process returns to step S31, and if not, the process returns to step S33.

In the calculation of the predicted value in step S33 and the correction processing in steps S35 and S36, FIG.
It is conceivable to apply various conditions as shown in (1) to (8) to obtain the predicted value. As described above, the bandwidth prediction unit 30 predicts the current allocation request bandwidth based on the past (slightly before) observed value or the declared value, so that it is possible to allocate an appropriate bandwidth according to the current situation for each line. it can.

(1.1st Embodiment) An embodiment of the bandwidth allocating apparatus and the bandwidth allocating method of the present invention will be described. This form corresponds to claims 21 and 33. This form is a modification of the first embodiment. First
Parts different from the embodiment will be described below. In this form, as the initial value of step S21 shown in FIG. 2 and FIG. 3, in addition to the predicted band, one cell or more is allocated in each cycle in which an increase in the allocation request band is detected.
Alternatively, steps S13, S16, S17 of FIG. 2 and FIG.
The prediction band obtained by increasing the band obtained by adding one cell per period for detecting an increase in the allocation request band is used as the band predicted in at least one of steps S23, S26, and S27.

When the initial value is set to 1 cell or more in every cycle in which an increase in the allocation request bandwidth is detected, it is possible to detect the presence or absence of the allocation request bandwidth in a desired cycle. In addition, when the prediction band is increased for the prediction, it is possible to detect the increase in the allocation request band more quickly than when the prediction is performed based on the observed value. (1.2nd Embodiment) One embodiment of a band allocation device and a band allocation method of the present invention will be described. This form corresponds to claims 21 and 33. This form is a modification of the first embodiment. The parts different from the first embodiment will be described below.

In this embodiment, as the initial value of step S21 shown in FIG. 2 and FIG. 3, the bandwidth allocated for the purpose of detecting the increase in the allocation request bandwidth is set every cycle in which the increase in the allocation request bandwidth is detected. It is characterized in that a small band is allocated from the band when one cell is allocated and the guaranteed band.

Therefore, unlike the first embodiment, a band that does not need to be fixedly guaranteed and allocated is not allocated for the purpose of detecting an increase in the allocation request band, and a band equal to or more than the contract is allocated. There is an effect that the allocation can be reduced and the accommodation efficiency can be increased as compared with the first embodiment. (Embodiment 1.3) An embodiment of the band allocating apparatus and the band allocating method of the present invention will be described. This form corresponds to claims 23 and 35. This form is a modification of the first embodiment. The parts different from the first embodiment will be described below.

In this embodiment, steps S13 and S of FIG.
16, S17 and steps S23, S26, S2 of FIG.
A prediction band obtained by multiplying the band to be predicted by at least one of 7 by a predetermined band increase coefficient is used. Further, compared to the predicted band before the increased predicted band is multiplied by the band increase coefficient by the band allocating device according to claim 1, the band for one cell or more is not increased per cycle in which it is desired to detect an increase in the allocation request band. In this case, a band obtained by adding one cell per cycle for detecting an increase in the allocation request band to the prediction band before the premium is a premium band.

In this embodiment, when the predicted bandwidth increased by the bandwidth increase coefficient is larger than the bandwidth obtained by adding one cell per cycle in which an increase in the allocation request bandwidth is detected, it is the same as in the first embodiment. By comparison, it is possible to more quickly predict an increase in the allocation request bandwidth. Also, if the predicted bandwidth increased by the bandwidth increase coefficient is less than the bandwidth obtained by adding one cell per cycle in which an increase in the allocation request bandwidth is detected, it is 1 in each cycle in which an increase in the allocation request bandwidth is detected. Since the bandwidth includes cells, there is an effect that the speed of detecting an increase in the allocation request bandwidth is not likely to decrease as compared with the first embodiment.

(1.4th Embodiment: without staircase prediction)
Adaptive Band Increasing Factor) One embodiment of the band allocating device and the band allocating method of the present invention will be described. This form corresponds to claims 24 and 36. This form is a modification of the first embodiment. The parts different from the first embodiment will be described below. The characteristic of this form lies in the band prediction unit 30. The band prediction unit 30 in this form is
When the band usage by the traffic observation unit 10 increases and the currently allocated band is less than or equal to a predetermined band,
Alternatively, the declared value received by the bandwidth request receiving unit 20 is positive,
In addition, if at least one of the currently allocated bandwidth is equal to or less than a predetermined bandwidth is true, it is predicted that the allocation request bandwidth will increase with a large bandwidth ratio, and the bandwidth usage by the traffic observing unit 10 will decrease. , Or at least one of the declared values received by the band request receiving unit 20 is true, the allocation request band is predicted to decrease with a small band ratio.

For this reason, in this embodiment, when the bandwidth is equal to or less than a predetermined bandwidth, for example, the allocation guarantee is guaranteed in the case of the GFR service, the bandwidth to be used can be quickly increased when the bandwidth to be used is increased. Is. Further, when the bandwidth to be used decreases, or when the declared value is 0 and the bandwidth is excessively allocated, the predicted bandwidth is predicted slowly, so that the problem that the allocated bandwidth decreases more than necessary can be solved. Therefore, it is possible to avoid the problem that the prediction band oscillates and oscillates with an unnecessarily large width with respect to the band that is actually required, particularly when changing the allocated band with a large band ratio.

(1.5th Embodiment) An embodiment of the band allocating apparatus and the band allocating method of the present invention will be described. This form corresponds to claims 28 and 40. This form is a modification of the first embodiment. First
Parts different from the embodiment will be described below. This mode is characterized in that, when the band value estimated by the band prediction unit 30 or the band value increased by the band increase coefficient exceeds the maximum band allowed for the line, the maximum band allowed for the line is replaced.

Therefore, it is possible to avoid the possibility that the device at the output destination is discarded when the maximum bandwidth allowed for the line is exceeded, so that the invalid allocation can be reduced. (Second Embodiment) One embodiment of a band allocation device and a band allocation method of the present invention will be described with reference to FIGS. This form corresponds to claims 2 to 4, claim 10 to claim 12, claim 31, claim 43, claim 45 and claim 46.

FIG. 5 is a block diagram showing the configuration of the band allocating device of this embodiment. FIG. 6 is a flowchart showing an operation example of the linear prediction unit. FIG. 7 is a flowchart showing the operation of the selecting unit of this form. FIG. 8 is a state transition diagram showing the state transition of the band prediction unit of this embodiment. FIG. 9 is a time chart showing an operation example (1) of this embodiment. Figure 10
Is a time chart showing an operation example (2) of this mode.

In this embodiment, the band predicting means, the first predicting means, the second predicting means, the used band discriminating means, the excess discriminating means and the selecting means of claim 2 are the band predicting section 30 and the linear predicting section, respectively. 31 corresponds to the staircase prediction unit 32, the unused band identification unit 35, the predicted value excess identification unit 34, and the selection unit 33. In this embodiment, it is assumed that the traffic observing unit 10, the band request receiving unit 20, the band predicting unit 30, and the band allocating unit 40 are provided in the band allocating device as shown in FIG. The configurations and operations of the traffic observing unit 10, the band request receiving unit 20, and the band allocating unit 40 are the same as those in the first embodiment, but the band predicting unit 30 is different.

In this embodiment, the band predictor 30 has a linear predictor 31, a staircase predictor 32 and a selector 3 as shown in FIG.
Equipped with 3. Further, the selection unit 33 includes a predicted value excess identification unit 34 and an unused band identification unit 35. The linear predictor 31 uses a predetermined constant band allocation period (T).
Each time, the used bandwidth is observed and the requested value of the allocated bandwidth is monitored, and these results are retained as history for a certain period. Then, the band to be allocated next (current band) is linearly predicted using the history of the used band or the allocation request band in the past several band allocation cycles (T). Here, the prediction value of the linear prediction unit 31 is called a linear prediction value.

In the i-th cycle of each band allocation cycle (T), x (i-2): used band observed in the previous allocation cycle x (i-1): observed in the previous allocation cycle Used band f (x (i-1), x (i-2), ...): Function used in linear prediction y (i + 1): Predicted band (number of cells) in the next allocation period to be predicted Then, the linear prediction value can be expressed as y (i + 1) = f (x (i-1), x (i-2), ...) (1).

However, the band is represented by the number of cells per hour.
When performing linear prediction using a linear function, at least two time values (x (i-1), x (i-2)) are required as inputs. For example, the following function can be used as the linear prediction function. y (i + 1) = (2x (i-1) -x (i-2)) × α (2) y (i + 1) = (3x (i-1) -2x (i-2) )) × α (3) α: constant The above equations (2) and (3) are obtained by using a linear function of the current use band (Y (i + 1) = α (X ( i-1) -X (i-2)) + X (i-1)) is effective because it is predicted to increase a little when it increases and a little when it decreases.

A specific example of the case where linear prediction is performed using the equation (2) will be described. For example, assuming the case where the number of passing cells in the previous two times is 5 and the number of passing cells in the previous time is 10 and the band increase coefficient (α) is 1.2, the next allocated band (the number of cells) y by the linear prediction is obtained as follows. .

Y = [number of cells by linear prediction] × [band increase coefficient] = ([number of previously passed cells] × 2- [number of previously passed cells]) × [band increase coefficient] = (10 × 2-5) × 1.2 = 18 (4) If the result of multiplying the band of linear prediction in the above formula (4) by the band increase coefficient α is equal to the band of linear prediction, the band of linear prediction is The result of adding the bandwidth for one cell per unit time (T) is predicted as the value of the current allocation bandwidth request.

Also, when the number of cells becomes a negative number, it is predicted as a band for one cell. However, when the observation band (the number of passing cells of the previous time) is 0 continuously (at least twice or more), the allocation request band is predicted as 0. here,
In this form, α may be 1 when the observation by the traffic observation unit is not used.

Further, the linear prediction unit 31 compares the above-mentioned straight line prediction with the declared value, and outputs the one with the larger one as the linear predicted value of the allocation request band. On the other hand, the staircase prediction unit 32 obtains a predicted value using a staircase function (non-linear function). In this example, the staircase prediction unit 32 determines the maximum value among the band (Vmin) such as the guaranteed band or the maximum band of the line, the current used band, the current allocated band, and the linear prediction value output by the linear prediction unit 31. Select and output as the current predicted value. Further, the predicted value is compared with the declared value of the bandwidth allocation request, and if the declared value is larger, the declared value is output as the predicted value. Here, the prediction value output by the staircase prediction unit 32 is referred to as a staircase prediction value.

The predictive value excess discriminating unit 34 included in the selecting unit 33 shown in FIG. 5 determines whether the declared value received by the band request receiving unit 20 exceeds the linear predictive value predicted by the linear predicting unit 31. Identification is made for each allocation period (T). Also, the selection unit 3
The unused band identification unit 35 included in No. 3 determines whether or not an unused band exists in the allocated band based on the observation value output from the traffic observation unit 10 in the band allocation period (T).
Identify each.

The selection unit 33 executes the processing shown in FIG. 7 for each band allocation period (T), and uses the identification result of the predicted value excess identification unit 34 and the identification result of the unused band identification unit 35 to perform linear processing. Either the linear prediction value output by the prediction unit 31 or the stairs prediction value output by the stairs prediction unit 32 is selected and output as the final prediction value. Next, the operation of the selection unit 33 will be described with reference to FIG.

At least during the period of two times of the cycle T, the linear predicting unit 31 proceeds from step S51 to S57 and unconditionally selects the staircase predicted value. Bandwidth request value and / or
Alternatively, immediately after the observed value of the used band has changed from 0 to a finite value other than 0 (at least twice in the cycle T), the linear prediction unit 3
Since there is not enough past data in 1 for linear prediction, it cannot be denied that the linear prediction value may be unreasonably low.
Therefore, in order to make a prediction on the safe side, in that case, the process proceeds from step S51 to S57, and the stairs prediction value is selected.

In step S52, the present predictive value excess discriminating unit 34 for the linear predictive value output from the linear predictor 31 is used.
Enter the identification result of. In step S53, the past history of the identification result input in step S52 is examined, the number of excesses A in the past N3 times of the bandwidth allocation period (T) is obtained, and it is compared with a predetermined threshold N4. That is,
When adopting a linear predicted value, identify whether the rate at which the declared value exceeds the predicted value is large.

If "A≤N4", the process advances from step S53 to S56 to select a linear prediction value. Also, "A>
In the case of "N4", the process proceeds from step S53 to S54. In step S54, the identification result of the unused band identification unit 35 for this time with respect to the staircase prediction value output by the staircase prediction unit 32 is input. In step S55, the past history of the identification result input in step S54 is examined, and the band allocation cycle (T)
The excess number B in the past L1 times is calculated and compared with a predetermined threshold L2. That is, when the staircase prediction value is adopted, it is determined whether or not the frequency of occurrence of an unused band (the used band is smaller than the staircase prediction value) is high.

If "B≥L2", the process proceeds from step S55 to S56, and a linear prediction value is selected. In addition, "B <
In the case of "L2", the process proceeds from step S55 to S57 and the stairs prediction value is selected. Therefore, the state transition regarding the selection of the linear prediction value of the linear prediction unit 31 and the staircase prediction value of the staircase prediction unit 32 in the selection unit 33 is as shown in FIG.

Whether the linear predicted value or the staircase predicted value is selected, the declared value is compared, and if the declared value is larger, the declared value is adopted as the final predicted value. To do. When using the band prediction unit 30 of this form,
For example, a band is assigned to each line as shown in FIGS.

The reason why the linear predicted value or the staircase predicted value is selected under the conditions shown in FIG. 7 is as follows. (1) When the usage state exceeds the prediction band, the band is insufficient in the band prediction using the linear prediction value ((1) The allocated band request exceeding the prediction is increasing), so the staircase prediction value is adopted. As a result, the allocated bandwidth can be increased rapidly and the arrival time to the target bandwidth can be shortened.

(2) When the ratio of the used band exceeding the linear prediction value is small (eg, when the linear prediction value exceeds the value obtained by multiplying the band increase coefficient or the like), the band usage state is in the linear prediction value. Since it is considered to be within the range, the linear prediction value is adopted. (3) When a band more than necessary is allocated by staircase prediction, the frequency of invalid bands that are not actually used increases. Therefore, when the frequency of occurrence of the unused band with respect to the staircase prediction value becomes equal to or higher than a predetermined value, the linear prediction value is adopted. Further, when the staircase prediction value is limited with the guaranteed bandwidth as the upper limit, the staircase prediction is changed to the linear prediction in order to allocate a band exceeding the guaranteed bandwidth.

(4) When the frequency of the unused band appearing in the band allocated by the staircase prediction value is small, it is considered that the state of the large allocation request band is continuing, so the selection of the staircase prediction value is continued. In this mode, since the bandwidth can be allocated based on the staircase prediction value, the guaranteed bandwidth can be promptly allocated to the line whose currently allocated bandwidth is less than or equal to the guaranteed bandwidth, and the guaranteed bandwidth can be guaranteed. Is possible.

When the unused bandwidth increases after allocating the guaranteed bandwidth in the staircase prediction, the linear prediction value is adopted and the observed used bandwidth or the declared allocation required bandwidth of the previous and the second previous line is adopted. Since it is possible to predict the linearly proportional band as the current allocation request band and allocate the band, it is possible to suppress the invalid band allocation. It should be noted that for customers' lines that do not have a guaranteed bandwidth, it is desirable to evenly divide the surplus bandwidth and replace it.

In this way, when the customer requests the bandwidth allocation, the guaranteed bandwidth can be allocated even if the requested bandwidth is less than the guaranteed bandwidth. Therefore, it is possible to avoid the guarantee band breakage. Further, when a band exceeding the customer's request is allocated, a wasteful unused band is generated, which is not preferable from the viewpoint of effective use of the shared band. In such a case, it is possible to reduce a wasteful unused band by adopting a linear prediction value and allocating a band.

In this embodiment, the linear predictor 31 compares the predicted value with the declared value and outputs the one with the larger value as the linear predicted value of the allocation request band. This is for the following reason. For example, in the case of B-PON, the declared value is displayed as the cell accumulation amount (queue length) of the ONT buffer at the time of declaration. Here, a description will be given of the allocation request bandwidth by displaying the queue length. The declared queue length is the number of cells already stored in the buffer, and is the number of cells that can be transmitted immediately by allocating a band. Therefore, as long as there is a free band, at least the band of the number of cells corresponding to the queue length can be effectively used.

In this mode, it is possible to give a guaranteed band to a user who is performing flow control between end terminals such as TCP, and the unused band is effective among other users. It can be utilized. In addition, since this embodiment has the staircase prediction unit as compared with the 1.4th embodiment, there is an effect that the corner calculation becomes simple without dynamically changing the coefficient of the band prediction function.

(Second Embodiment) One embodiment of the band allocating apparatus and the band allocating method of the present invention will be described. This form corresponds to claims 25 and 37. This form is a modification of the second embodiment, and the difference from the second embodiment lies in the band request receiving unit 20 and the traffic observing unit 10.

In this embodiment, the band request receiving unit 20 normalizes the number of cells to be received by dividing it by the time to receive the declared value for each line. Further, the traffic observing unit 10 normalizes by dividing the number of observed cells by the time of the identification interval of the used band for each line. Therefore, it is possible to correctly predict and allocate the band even if the cycle of band allocation or the like changes.

(Embodiment 2.2) An embodiment of the bandwidth allocating apparatus and the bandwidth allocating method of the present invention will be described. This form corresponds to claims 26 and 38. This form is a modification of the second embodiment. Portions different from the second embodiment will be described. In this mode, the Vmin used by the staircase prediction unit 32 is the entire usable shared band, or the usable shared band divided by the number of lines whose predicted value of the band allocation request is positive, or the usable shared band. The value is obtained by multiplying the total by the guaranteed bandwidth of each line and dividing by the sum of the guaranteed bandwidth of the line whose predicted bandwidth request is positive, or by multiplying the available shared bandwidth by the guaranteed bandwidth of each floor and the sum of the guaranteed bandwidth of all lines. It is characterized by the value divided by.

Therefore, in this mode, it is possible to give a band to a line whose used band is increased more rapidly than in the second embodiment. Here, the size of the shared bandwidth that can be used is the sum of the original surplus bandwidth, which is the remainder of the bandwidth allocation that should be dynamically guaranteed. Alternatively, for example, when DBA of B-PON is performed as the used band, the line (T-
"Fixed BandwidT" to allocate bandwidth to CONT)
A value excluding "h" is used. In the latter case, the calculation of the usable shared band is fixed, which has the effect of simplifying the calculation.

(Third Embodiment) FIG. 1 shows an embodiment of the band allocating apparatus and the band allocating method of the present invention.
This will be described with reference to FIG. FIG. 11 is a flowchart showing the operation of the linear predictor of this embodiment. This form is the second
It is a modification of the embodiment. In this form, the operation of the linear prediction unit 31 is changed as shown in FIG. In FIG. 11, steps corresponding to those in FIG. 6 are shown with the same numbers.

That is, step S43 in FIG.
The processing of B and S47B is different from that of FIG. For the same parts as those in the second embodiment, the following description will be omitted. In the second embodiment, if the maximum value of the queue length of the declared value is significantly smaller than the buffer length and there is a deviation between the declared cycle and the allocated cycle, the predicted allocation request bandwidth will fluctuate. Can occur.

The reason for this is as follows. (1) When the input traffic exceeds the guaranteed band and rapidly increases, (2) the allocated band increases with a cycle delay (the allocated bands for the first two cycles are the same), and eventually (3) the input band overshoots.
Here, (input band <output band), and (4) input traffic and the accumulated amount in the buffer are declared. After transmitting the buffer accumulated cells, (input band = output band), and (5) the number of transmitted cells is reduced by the buffer accumulated cells as compared with the previous allocation period.

(6) This decreasing tendency is applied to the next band prediction, a band smaller than the input band is allocated, and (input band> output band), the cells are accumulated in the buffer. Vibration continues after that. In this form, the generation of vibration can be suppressed. In this form, in order to avoid overestimating the allocation request bandwidth unnecessarily even though the input traffic has not decreased, the processing (S47B) when the allocation bandwidth has decreased from the previous allocation bandwidth is changed. There is.

In order to calculate the allocated bandwidth, the linear predicting unit 31 predicts the bandwidth by a linear function based on the number of cells used in the previous and previous allocation cycles. If the slope of the linear function is negative, the step is performed. In S47B, the number of cells used in the last-preceding allocation period is adopted as the predicted band for the next allocation period. Therefore, as compared with the case of the second embodiment, the reduction of the allocated bandwidth is delayed, so that the excessive reduction of the allocated bandwidth can be suppressed.

On the other hand, if the input bandwidth is reduced, the used bandwidth is still reduced, so the bandwidth calculated based on this number of used cells is reduced, and the predicted value of the allocation request bandwidth is appropriately set. Can be reduced. Note that FIG.
1, v (0): the number of OLT reception cells in the previous allocation cycle v (-1): the number of OLT reception cells in the allocation cycle before last α: a band increase coefficient.

In this mode, if the bandwidth of the input traffic is constant, stable output can be performed. (Fourth Embodiment) FIGS. 12 and 13 show one embodiment of a band allocation device and a band allocation method of the present invention.
Will be described with reference to. FIG. 12 is a block diagram showing the configuration of the band prediction unit of this embodiment. FIG. 13 is a state transition diagram showing the state transition of the band prediction unit in this form.

This embodiment is a modification of the second embodiment. The configuration of the band prediction unit 30 is changed as shown in FIG. 12, and the operation regarding the state transition of the band prediction unit is changed as shown in FIG. The following description will be omitted for the unchanged parts. As shown in FIG. 12, in this embodiment, the all-use identifying unit 36 is added to the selecting unit 33 of the band predicting unit 30.

The all-use identifying unit 36 determines the band allocation period (T).
In the past N1 times, the number of times that all of the allocated band has been used or the number of times that a report exceeding the allocation has occurred is compared with the threshold value N2 to identify the frequency of use of all the bands for each line. Therefore, the selection state of the selection unit 33 transits as shown in FIG. That is, N2 during the past N1 allocation cycles
A line that has used all of its allocated bandwidth in an allocation cycle of more than or equal to one, or a line that has received a declared value exceeding the allocation request bandwidth predicted by the linear prediction means in N4 or more of the declaration cycles of the past N3 allocation cycles, or all L during the past L1 allocation period after using the band
The staircase prediction value is adopted for a line having an unused band with an allocation period of two times or more, and a linear prediction value is adopted for other lines.

In this mode, since all use of the allocated band is confirmed, it is possible to avoid the occurrence of vibration. (Fifth Embodiment) FIGS. 14 and 15 show one embodiment of a band allocation device and a band allocation method of the present invention.
Will be described with reference to. FIG. 14 is a block diagram showing the configuration of the band prediction unit of this embodiment. FIG. 15 is a state transition diagram showing the state transition of the band prediction unit in this form.

This embodiment is a modification of the second embodiment. The configuration of the band prediction unit 30 is changed as shown in FIG. 14, and the operation regarding the state transition of the band prediction unit is changed as shown in FIG. The following description will be omitted for the unchanged parts. As shown in FIG. 14, in this embodiment, a total use identifying unit 36 and a continuation number identifying unit 37 are added to the selecting unit 33 of the band predicting unit 30.

The all-use identifying unit 36 determines the band allocation period (T).
In the past N1 times, the number of times that all of the allocated band has been used or the number of times that a report exceeding the allocation has occurred is compared with the threshold value N2 to identify the frequency of use of all the bands for each line. The continuation frequency identification unit 37 compares the continuous prediction frequency with the threshold value M while the selection unit 33 selects the staircase prediction value for each line, and outputs the result.

The selecting unit 33 utilizes the identification results of the prediction value excess identifying unit 34, the unused band identifying unit 35, the all-use identifying unit 36, and the continuation number identifying unit 37, as shown in FIG. Select the predicted value. That is, as shown in FIG. 15, the line that fully uses the allocated bandwidth in the N2 or more allocation cycles in the past N1 allocation cycles, or the linear prediction means in the N4 or more declaration cycles in the past N3 allocation cycles A line that has received a declared value that exceeds the estimated allocation request bandwidth, or a line that has been continuously allocated bandwidth M with the predicted step value selected is M or less, or the past after using all bandwidth The staircase prediction value is adopted for a line having no unused band in the L1 or more allocation periods during the L1 allocation period, and the linear prediction value is adopted for other lines.

In this embodiment, when selecting a staircase prediction value, the selected state of the staircase prediction value is maintained for a period corresponding to M times. Therefore, even if a deviation occurs between the reporting cycle, the observation cycle, and the band allocation cycle, it is possible to suppress oscillation in the band allocation control.

(Sixth Embodiment) FIG. 1 shows an embodiment of the band allocating apparatus and the band allocating method of the present invention.
This will be described with reference to FIG. This form corresponds to claims 5 and 13. FIG. 16 is a block diagram showing the configuration of the staircase prediction unit of this embodiment. In this form, the guaranteed bandwidth identification of claim 5 corresponds to the guaranteed bandwidth allocation identifying unit 323.

This embodiment is a modification of the second embodiment. In this embodiment, the staircase prediction unit 32 is modified as shown in FIG. 16, but the other configurations and operations are the same as those in the second embodiment. For the same parts as those in the second embodiment, the following description will be omitted. The staircase prediction unit 32 shown in FIG. 16 includes a below-guaranteed band prediction unit 321, a above-guaranteed band prediction unit 322, and a guaranteed band allocation identification unit 323.

The guaranteed bandwidth allocation identifying unit 323 compares the current allocated bandwidth with the threshold value of the guaranteed bandwidth of the line to identify whether the allocated bandwidth is less than or equal to the guaranteed bandwidth. When the allocated bandwidth is less than or equal to the guaranteed bandwidth, the guaranteed bandwidth allocation identifying unit 323 selects the allocation request bandwidth output by the below-guaranteed bandwidth predicting unit 321 and outputs it as the staircase prediction value. When the allocated band exceeds the guaranteed band, the guaranteed band allocation identifying unit 323 selects the allocation request band output from the guaranteed band or higher prediction unit 322 and outputs it as the step prediction value.

The below-guaranteed-band predicting unit 321 selects the maximum value among the guaranteed band of the line, the currently used band, the current allocated band, and the linear prediction value of the linear prediction unit 31, and predicts the allocation request band. Output as a value. Guaranteed bandwidth or higher prediction unit 3
22 is a band obtained by dividing the usable shared band by the guaranteed bandwidth ratio of all lines, or a band obtained by dividing the usable shared band by the total number of lines, the currently used band, the currently allocated band, and the linear prediction unit 31. The maximum value among the linear prediction values is selected and is output as the prediction value of the allocation request bandwidth.

Here, the usable shared band is the sum of the original surplus bands, which is the remainder of the band allocation to be dynamically guaranteed, or the band excluding the band to be fixedly allocated to the line. Therefore, if the previously predicted value is less than or equal to the guaranteed bandwidth of the line, the maximum bandwidth among the guaranteed bandwidth of the line, the current used bandwidth, the current allocated bandwidth, and the linear prediction value is predicted as the allocation request bandwidth, and the previous prediction If the bandwidth exceeds the guaranteed bandwidth of the line, the usable shared band is divided by the guaranteed bandwidth ratio of all lines, or the usable shared band is divided by the total number of lines, the current used band, the current The maximum band among the allocated band and the linear prediction value is predicted as the allocation request band.

According to this embodiment, even when a bandwidth exceeding the guaranteed bandwidth is requested for allocation, it is possible to deal with the allocation request bandwidth that increases in bursts, and a more fair surplus bandwidth allocation can be realized. . (Embodiment 6.1) An embodiment of the bandwidth allocating apparatus and the bandwidth allocating method of the present invention will be described. This form corresponds to claims 27 and 39. This form is a modification of the sixth embodiment.

The difference between this embodiment and the sixth embodiment is that
This is in the process when the preceding prediction value is 0. In the sixth embodiment, when shifting from linear prediction to staircase prediction, if the preceding prediction value is less than or equal to the guaranteed band, the below-guaranteed band prediction unit 3
Transition to 21 predictions. In this form, the preceding prediction value is 0.
Only when the preceding prediction value is less than or equal to the guaranteed band, instead of transitioning to the prediction of the less than or equal to guaranteed band prediction unit 321, the prediction is changed to more than or equal to the guaranteed band prediction unit 322.

Therefore, compared to the case where the initially allocated band for the line that has not used the band is transited to the prediction of the less than guaranteed band prediction unit 321 as in the sixth embodiment, the predicted band is more prompt. Can be increased rapidly. (Seventh Embodiment) One embodiment of a band allocation device and a band allocation method of the present invention will be described with reference to FIG. This form includes claims 6 to 8 and claim 1.
4 to claim 16 are supported. FIG. 17 is a block diagram showing the configuration of the linear prediction unit of this form.

This embodiment is a modification of the second embodiment. In this embodiment, the linear prediction unit 31 is modified as shown in FIG. 17, but the other configurations and operations are the same as those in the second embodiment. For the same parts as those in the second embodiment, the following description will be omitted. The linear prediction unit 31 illustrated in FIG. 17 includes a linear proportional prediction unit 311 and a difference prediction unit 312.
And an adder 313.

The linear proportional predictor 311 calculates a band that is linearly proportional to the observed used band or the declared allocation request band of the line of the previous time and the line before the previous time. Difference prediction unit 312
Calculates a value obtained by multiplying the previously declared allocation request bandwidth by a coefficient, or a value obtained by multiplying the increase (difference) of the previous declared bandwidth with respect to the last-minute declared bandwidth by a coefficient. The addition unit 313 outputs the result of adding the output of the linear proportional prediction unit 311 and the output of the difference prediction unit 312 as a linear prediction value.

In this embodiment, the result obtained by adding the value output from the difference prediction unit 312 to the value linearly proportional to the observed value or the declared value is output as the linear prediction value, and therefore the selection unit 33 selects the linear prediction value. Even if it is, the increase in the allocation request bandwidth can be predicted more quickly than in the second embodiment. Here, for the coefficient used in the difference prediction unit 312, the declared value is OL after the declared value is transmitted by the ONT.
In proportion to the time required to be reflected in the bandwidth allocation at T,
It is determined by the coefficient of the control period such as the band allocation period (T).

For example, if a delay of (3T) cycles occurs after the declared value is transmitted from the ONT and is reflected in the bandwidth allocation in the OLT, the coefficient is set to 3. Therefore, when the declaration information can be used, it becomes possible to deal with the allocation request bandwidth that increases in bursts as compared with the second embodiment, and a more fair surplus bandwidth allocation can be realized. You can

(Embodiment 7.1) An embodiment of the band allocating apparatus and the band allocating method of the present invention will be described. This form corresponds to claims 32 and 44. This form is a modification of the seventh embodiment. The difference between this form and the seventh embodiment is that the difference prediction unit 312
Is in operation. In this mode, the difference prediction unit 312 converts a value linearly proportional to the declared value of the allocation request band into the number of cells per cycle of the allocation request declaration, and when the buffer length of the allocated apparatus exceeds, It is characterized in that it is used by replacing it with the buffer length of the allocated device per cycle of the allocation request declaration. Therefore, it is possible to suppress an invalid bandwidth allocation that exceeds the buffer length of the allocated device.

(Eighth Embodiment) FIG. 18 shows an embodiment of the band allocating apparatus and the band allocating method of the present invention.
Will be described with reference to. This form is claim 24, claim 3
Corresponds to 6. This form is a modification of the second embodiment. In this mode, as shown in FIG. 18, the linear prediction unit 31 includes a linear proportional prediction unit 311, an addition unit 313, and a band usage rate observation unit 315.

The band usage rate observing section 315 calculates the band usage rate for the previously allocated band. The linear proportional prediction unit 311 finds a coefficient by which the used band or the allocation request band is multiplied based on the calculated band usage rate, and outputs the result as a linear predicted value. Here, the coefficient is given by a function for the usage rate, and the smaller the usage rate, the smaller the coefficient.

In this embodiment, since the coefficient varies depending on the band usage rate, the increase of the predicted value is suppressed when the band usage rate is low. Therefore, excessive band allocation can be prevented as compared with the case where the coefficient is a constant. (Ninth Embodiment) An embodiment of a band allocation device and a band allocation method of the present invention will be described with reference to FIG. FIG. 19 is a state transition diagram showing the state transition of the band prediction unit in this form. This form is a modification of the second embodiment.

In this mode, the state transition from the linear prediction to the staircase prediction is such that the used band or the declared value is when the traffic observing unit uses all the allocated band or when the predicted value by the linear predicting unit is 0. This occurs when at least one of the cases where the traffic observation unit identifies that it is even one cell is true. The transition from the staircase prediction to the linear prediction occurs when at least one of the case where the traffic observing unit does not use all the allocated band or the case where the used band does not exceed the predicted value by the staircase predicting unit is true.
The excess identification unit used here does not identify excess / non-excess of the declared value in the other prediction bands with respect to the prediction band, except when the prediction band is 0.

Therefore, in this embodiment, the band prediction formula in the band in which the band for observation is added to the prediction band can be expressed as follows. alloc (t) = max (recv.idle (t) x α, recv.data (t) + que (t) x β, γ) (5) where recv.data (t) is the bandwidth allocation Use band detected by the traffic observation unit that can be used for processing in the band prediction unit at cycle t (display of the number of cells per hour), recv.i
dle (t) is the number of unused cells that can be used for processing in the band predicting unit in the band allocating cycle t, and que (t) is used in processing in the band predicting unit in the band allocating cycle t. Declaration value (queue length) of the bandwidth allocation request that becomes possible, max (a, b, ...) is the maximum value, and alloc (t) is output as the result prediction value of the processing of the bandwidth prediction unit at the bandwidth allocation cycle t. Represents the allocated bandwidth (display of the number of cells per hour).

In this case, (recv.data (t)),
The subscript (t) of (recv.idle (t)), (que (t)), and (alloc (t)) is shown focusing on the prediction process in the band prediction unit, and the band allocation cycle The input and output of the processing in the band prediction unit at t are shown as the same subscript.
That is, when performing the prediction processing in the band prediction unit in the t-th period #t of the band allocation cycle, the detection and band request reception unit in the traffic observation unit up to the period # (t-1) of the band allocation cycle. It means each value when predicting the allocated bandwidth in the period of # (t + 1) of the bandwidth allocation cycle by using the declared value in. Therefore, in this case, (recv.data (t)), (recv.idle (t))
Each value of indicates the use status of the bandwidth actually allocated by (alloc (t-2)). Therefore, although the same subscript is attached, (recv.data (t)), (recv.idle (t)) and (al
The timing of loc (t)) is shifted by two periods of the band allocation period.

The term α in this band prediction formula is for predicting the next band based on the band usage rate. recv.idl
If e (t) is 0, it is determined that the bandwidth is increasing, and α is set to a large value (for example, 1.5), and if recv.idle (t) is positive, it is in a steady state or a decreasing state. It is determined that α is a small value (for example, 1.1). The term β is for reacting rapidly to a sudden increase in the input band. β
Is a parameter for correcting the allocated band when the input cells are accumulated in the buffer of the band allocation device such as ONT in BPON. The fact that cells are accumulated in the buffer means that (input band <allocated band), so that the shortage of the allocated band is estimated from the queue length of the declared value and added to the allocated band.

If the input band and the allocated band continuously have the same value as they are, three band allocation cycles corresponding to the allocation delay (reflected after the cells start to accumulate) until the changed allocated band can be reflected. It is appropriate to allocate a band of about two cycles that have been accumulated before the start of the process and one cycle that is an excess of the band allocation period). γ is a predetermined band P. Bandwidth allocation for a predetermined bandwidth such as a guaranteed bandwidth is performed when the input bandwidth tends to increase below GUB (for example, a guaranteed bandwidth, a maximum bandwidth, a bandwidth in which fixed allocation is fixedly excluded from the guaranteed bandwidth). Even if γ is set to 0, if a band for observation is required, the band corresponding to the band is set.

As an example of dynamically changing α, β, γ,
For example, the following combinations are used. If “(recv.data (t))> 0” and “(recv.idle (t)) = 0”, α = 1.5, β = 2.5, γ = (P.GUB) “(recv .data (t))> 0 ”and“ (recv.idle (t))> 0 ”, α = 1.1, β = 2.5, γ = 0“ (recv.data (t)) = If it is “0”, α = 1.5, β = 2.5, γ = 0 In this mode, since the prediction of the increase of the next band is adjusted based on the band usage rate, the wasteful allocation of the band is performed. Can be reduced. In addition, by correcting the allocated bandwidth with the queue length that is the declared value, there is an effect that the delay time until the bandwidth allocation is reduced with respect to a sudden increase in the allocation request bandwidth.

In this embodiment, as compared to the second embodiment, when the excess identifying section identifies that the used band or the declared value is even one cell, the transition to stair prediction is performed, so that there is no unused band. It is also possible to deal with the case where the allocated bandwidth of 1 increases and the queue length increases. (Tenth Embodiment) FIG. 20 and FIG. 2 show one embodiment of the band allocating apparatus and the band allocating method of the present invention.
2, with reference to FIG. This form is claim 1.
7. It corresponds to claim 18. FIG. 20 is a flowchart showing an operation example of the band prediction unit in this embodiment. Figure 2
2 is a block diagram showing the configuration of the band allocation device of this embodiment.

In this embodiment, the traffic observing means, the band request receiving means, the band predicting means and the band allocating means of claim 17 are the traffic observing section 10, the band request receiving section 20, the band predicting section 30 and the band allocating section 40, respectively. Equivalent to.

Similar to the first embodiment, the bandwidth allocating device of this embodiment has a traffic observing section 10 and a bandwidth request receiving section 2
0, a band prediction unit 30, and a band allocation unit 40.
The band allocating unit 40 allocates a band for each line according to the size of the band predicted by the band predicting unit 30 (the band currently required). The traffic observing unit 10 observes the band allocated by the band allocating unit 40 for each line and the amount of traffic actually flowing through the line, that is, the use condition of the band.

The band request receiving section 20 receives the declared value of the band requested for each line, which is input by the customer or the like. The band prediction unit 30 uses at least one of the information on the used band obtained by the observation of the traffic observation unit 10 and the declared value of the allocation request band received by the band request reception unit 20, and the band allocation unit 4
The current allocation request bandwidth is predicted based on the past allocation bandwidth of 0.

The band predicted by the band predicting unit 30 is allocated to each line by the band allocating unit 40. The difference between this form and the first embodiment lies in the band prediction process. As a specific operation of the band prediction unit 30 in this form, an operation as shown in FIG. 20 can be considered. In this operation example, it is assumed that the band prediction is repeatedly performed at regular intervals. In addition, it is assumed that the operation used is simplified so that the value used for prediction is switched at every predetermined cycle (bandwidth allocation cycle).

In this embodiment, control is performed so as to prevent the predicted band of the band predicting unit 30 from unnecessarily changing when the band allocated by the band allocating unit 40 temporarily loses even one cell. Hereinafter, description will be made with reference to FIG. The band is represented by the number of cells per unit time. Also, this process is performed for each line.

The initial value of the flag alloc.flag is set to 0, and this value is inherited between adjacent periods of each band allocation cycle. When the allocated bandwidth (alloc (t)), which is the predicted value of the bandwidth used for the line, is predicted to be other than 0 and the band is allocated, the bandwidth is allocated in step S70 of FIG. 20 (alloc.flag = 1). Then, when the condition of step S62 is satisfied, (alloc.flag = 0) is set. Also, (alloc
If (recv.data (t) = 0), the value of the flag alloc.flag is not changed because there is no bandwidth allocation in the bandwidth allocation cycle two times before, which is the predicted allocation bandwidth (t-2)). Therefore, the used bandwidth in the previous bandwidth allocation cycle, which is the allocated bandwidth predicted by (alloc (t-1)), is positive, and the flag alloc.flag = 1
In this case, bandwidth allocation is performed according to the branching condition of step S64.

The band prediction formula in FIG. 20 is as follows. alloc (t) = max (recv.data (t) × α1 + α2, recv.data (t) + que (t) t × β, γ) (6) recv.data (t): Bandwidth allocation period t Used by the bandwidth observing unit for processing in the bandwidth observing unit (displaying the number of cells per hour) recv.idle (t): Processing in the bandwidth Predicting unit at band allocation cycle t Number of unused cells that can be used for que (t): Declared value (queue length) of bandwidth allocation request that can be used for processing in the bandwidth prediction unit at bandwidth allocation cycle t max (a, b , ...): Maximum value alloc (t): Allocated bandwidth output as a predicted value of the result of the processing of the bandwidth predictor in the bandwidth allocation cycle t (display of the number of cells per time) In this mode, (recv.data (t)), (recv.idle
The subscript (t) of (t)), (que (t)), and (alloc (t)) is shown focusing on the prediction processing in the band prediction unit, and the band prediction in the band allocation cycle t is shown. The input and output of the process in the section are shown as the same subscript. That is,
When performing the prediction process in the band prediction unit in the t-th period #t of the band allocation period, the detection in the traffic observation unit and the band request reception unit in the period # (t-1) of the band allocation period It means each value when predicting the allocated bandwidth in the period # (t + 1) of the bandwidth allocation cycle using the declared value. Therefore, in this case, the values of (recv.data (t)) and (recv.idle (t)) actually indicate the usage status of the bandwidth allocated by (alloc (t-2)). There is. Therefore, although the same subscript is attached, the timings of (recv.data (t)), (recv.idle (t)) and (alloc (t)) are shifted by two band allocation periods.

It should be noted that FIG. 20 is an example of the case where prediction is performed in cell units. In S69, the right side of the equation (6) is processed by ceil (), which means rounding up, in order to round up. Here, the setting values of the coefficients α1, α2, β, γ differ between the case of linear prediction and the case of staircase prediction. For example, in linear prediction, (α1, α2, γ) is set to (1, 2, 0), and in staircase prediction, (α1 = 2, α2 = 10, γ = P.GUB)
And

The variable γ is a predetermined band P.P.
It is used to perform bandwidth allocation for a predetermined bandwidth such as a guaranteed bandwidth when the input bandwidth tends to increase below GUB (for example, a guaranteed bandwidth, a maximum bandwidth, a bandwidth in which fixed allocation is fixedly excluded from the guaranteed bandwidth). The transition conditions of linear prediction and staircase prediction regarding the determination of the coefficients α1, α2, β, γ are shown below.

(Recv.data (t) = 0) and (que (t) = 0) and
In case of (alloc.flag = 0), in case of alloc (t) = 0 (recv.data (t)> 0), in case of (recv.idle (t) = 0), (α1 = 2, α2 = 10, β =
2.5, γ = P.GUB) (recv.idle (t)> 0) (α1 = 1, α2 = 2, β =
2.5, γ = 0) (recv.data (t) = 0), (α1 = 1, α2 = 10, β = 2.5, γ = P.GUB) In this embodiment, in the ninth embodiment, Unlike the form, (γ =
When it is set to 0), it is possible to reduce the appearance frequency of the allocation period to which the band for observation is added.

It is assumed in FIG. 20 that the band expected to be used by the line is predicted as the allocated band alloc (t). Therefore, when the reported value of the allocation request bandwidth of the effective line has not arrived and the used bandwidth is 0, the allocation request bandwidth is increased at a predetermined interval as in the case of the second embodiment. It is necessary to give a band for monitoring to the output of the allocated band alloc (t) shown in FIG.

Therefore, as shown in FIG. 23, the band expected to be used by the line is the band P.3 for monitoring the increase of the allocation request band at a predetermined interval. If it is less than UM, the allocated bandwidth is set to P. Replace with UM. When the value of the allocated bandwidth alloc (t) is explicitly changed, the alloc.flag is not changed, that is, as shown in FIG.
Between step S72 and "end" of step S72
1, S722 is inserted, and the allocated bandwidth alloc (t) is the bandwidth P. U
If it is less than M, the allocated bandwidth is P.M. Replace with the value of UM. P. In this embodiment, it is assumed that the value of UM is changed according to the interval at which the increase of the allocation request bandwidth is monitored.

For example, when the interval for monitoring the increase of the allocation request bandwidth is four times the bandwidth allocation cycle, “1,0,0,0,1,0,0,0, ... , One “1”, and three “0” are used to sequentially extract the values of the sequence one by one for each bandwidth allocation period. The value of UM may be determined. In this form, the allocated bandwidth alloc (t)
The used bandwidth recv.data (t) is used to obtain It is also assumed that the input value for each cycle is updated for each allocation cycle for the purpose of simple conditional branching. Therefore, the used band recv.data (t) becomes 0 if no band is allocated, and as a result, the allocated band alloc (t) becomes 0.

That is, when there is a period in which the band is not allocated and the period in which the band is allocated is a multiple of 2 of the band allocation period, the band is allocated based on the used band two cycles before. If there is no band allocation in the period two cycles before, the used band becomes 0, and a period in which no band is allocated occurs. As a result, "allocated" and "unallocated" appear alternately in each band allocation cycle.

Therefore, in this embodiment, the used band recv.dat is used.
When a (t) is 0, it is determined whether the reason is unallocated bandwidth or there is no cell to be transmitted, based on the input from the bandwidth allocating unit 40 as the presence or absence of bandwidth allocation for the bandwidth allocation cycle. To do.

In the line using the allocated band, the used band is 0 because the band is not allocated in the band allocation period.
If the used bandwidth is non-zero, the bandwidth prediction unit 30 performs the required bandwidth prediction. The flag alloc.flag is used for this identification. The flag alloc.flag records whether or not bandwidth allocation was performed in the previous bandwidth allocation cycle, and is used to identify whether or not there is a requested bandwidth.

This flag alloc.flag is the 9th
It is also possible to apply it to the band prediction formula shown in the embodiment. Further, in this embodiment, as compared with the band prediction formula of the ninth embodiment, instead of the value obtained by multiplying the number of used CSs by the coefficient α1, the value obtained by multiplying the number of used CSs by the coefficient α1 and a predetermined constant α2. Since the sum of and is used, it is possible to predict a large band when the required band increases and prevent excessive prediction in the part where the predicted band converges. Therefore, there is also an effect that the required bandwidth is not unnecessarily excessively estimated.

As described above, in this embodiment, the value can be calculated without taking over the value for each predetermined allocation period, and the band can be predicted without error even when there is a period with no allocated band. (Eleventh Embodiment) FIGS. 1 and 21 show one embodiment of a band allocation device and a band allocation method of the present invention.
Will be described with reference to. This form corresponds to claims 19 and 20. FIG. 21 is a flowchart showing an operation example of the band prediction unit in this embodiment.

In this embodiment, the traffic observing means, the band request receiving means, the band predicting means and the band allocating means of claim 19 are the traffic observing section 10, the band request receiving section 20, the band predicting section 30 and the band allocating section 40, respectively. Corresponding to. As shown in FIG. 1, the bandwidth allocating device of this embodiment includes a traffic observing unit 10, a bandwidth request receiving unit 20, and a bandwidth predicting unit 3.
0 and the band allocation unit 40.

The band allocating unit 40 allocates a band for each line according to the size of the band predicted by the band predicting unit 30 (the band currently required). Traffic observation section 10
Observes the bandwidth allocated by the bandwidth allocating unit 40 for each line and the amount of traffic actually flowing through the line, that is, the usage status of the band. The bandwidth request receiving unit 20 receives a declared value of an allocation request bandwidth for each line, which is input by a customer or the like.

The band predicting unit 30 includes the traffic observing unit 10
The current allocation request band is predicted based on at least one of the information on the used band obtained by the observation and the declared value of the allocation request band received by the band request receiving unit 20. The band predicted by the band prediction unit 30 is allocated to each line by the band allocation unit 40.

The operation of the traffic observation section 10 of this embodiment is slightly different from that of the tenth embodiment. In this form, when there is no allocated band for a specific period (bandwidth allocation period) observed by the traffic observation unit 10, the traffic observation unit 10 does not update the observation value to be output, and the past observation value is not updated. Is output as is as the latest observed value. In addition, the band prediction is repeatedly performed in a predetermined cycle as in the tenth embodiment. Further, the value used for prediction is switched in a predetermined cycle unit.

In this embodiment, since the observed value output from the traffic observing section 10 is corrected in advance, it is not necessary to discriminate whether or not the allocated band is 0 for band prediction. That is,
In the bandwidth prediction operation of this form, it is not necessary to use the flag alloc.flag, and the bandwidth prediction operation is performed as shown in FIG.
It is simplified compared to. The band prediction formula in FIG. 21 is as follows.

Alloc (t) = max (recv.data (t) × α1 + α2, recv.data (t) + que (t) t × β, γ) (7) However, (recv.data (t ) = 0) and (que (t) = 0), alloc (t)
= 0 (recv.data (t)> 0), if (recv.idle (t) = 0) then (α1 = 2, α2 = 10, β =
2.5, γ = P.GUB) (recv.idle (t)> 0) (α1 = 1, α2 = 2, β =
2.5, γ = 0) (recv.data (t) = 0), (α1 = 2, α2 = 10, β = 2.5, γ = P.GUB) where alloc (t): bandwidth allocation period Allocated bandwidth (cell number display per time) output as a predicted value of the processing of the bandwidth prediction unit at t recv.data (t): Can be used for processing by the bandwidth prediction unit at bandwidth allocation cycle t Bands detected by the traffic observation unit (display of the number of cells per hour) recv.idle (t): Number of unused cells that can be used for processing in the band prediction unit at the band allocation period t que (t ): Declared value (queue length) of bandwidth allocation request that can be used for processing in the bandwidth prediction unit at bandwidth allocation cycle t max (a, b, ...): Maximum value recv.data (t)), (recv.idle
The subscript (t) of (t)), (que (t)), and (alloc (t)) is shown focusing on the prediction processing in the band prediction unit, and the band prediction in the band allocation cycle t is shown. The input and output of the process in the section are shown as the same subscript. That is,
When performing the prediction processing in the band prediction unit in the t-th period #t of the band allocation period, the detection in the traffic observation unit and the band request detection unit in the period # (t-1) of the band allocation period It means each value when predicting the allocated bandwidth in the period # (t + 1) of the bandwidth allocation cycle using the declared value. Therefore, in this case, the values of (recv.data (t)) and (recv.idle (t)) actually indicate the usage status of the bandwidth allocated by (alloc (t-2)). There is. Therefore, although the same subscript is attached, the timings of (recv.data (t)), (recv.idle (t)) and (alloc (t)) are shifted by two band allocation periods.

Also, the variable γ is a predetermined band in the step prediction.
P.GUB (Guaranteed bandwidth, maximum bandwidth, guaranteed bandwidth from which bandwidth is fixedly removed) Used to allocate a certain bandwidth such as a guaranteed bandwidth when the input bandwidth tends to increase below There is. In the example of FIG. 21, the band predicted to be used by the line is predicted as the allocated band (alloc (t)). Therefore, when the reported value of the allocation request bandwidth of the effective line has not arrived and the used bandwidth is 0, the increase of the allocation request bandwidth is monitored at a predetermined interval as in the tenth embodiment. 21 must be added to the output of the allocated bandwidth (alloc (t)) in FIG. Therefore, it is possible to add the same processing as that of FIG. 23 to FIG. 20, but α which is the constant term of steps S85 and S86.
P. 2 rather than 2. If UM is not large, “alloc (t) ← 0” is replaced with “alloc (t) ← PU in step S83 of FIG.
Replace with "M".

In this embodiment, similarly to the tenth embodiment, when γ is set to 0, the frequency of appearance of the allocation period to which the observation band should be added can be reduced. As described above, in this embodiment, the same effect can be obtained by the operation of the band predicting unit 30 which is simpler than that of the tenth embodiment. (Twelfth Embodiment) Another embodiment of the present invention will be described. This mode is a modification of the eleventh embodiment, and only the operation of the traffic observation unit 10 is changed as follows.

In this embodiment, in the period T (0) in which the allocated bandwidth is 0 (the period of one allocation cycle), the past period T (-n) to the current period T ( The traffic observation unit 10 outputs (Bx / elapsed time) as a used band using the elapsed time up to (0) (n × allocation period). However, when the value delivered to the band prediction unit 30 is limited to an integer number of cells or the value used by the band prediction unit 30 is the number of cells and the used band is less than 1 cell, 1
Output as a cell.

Therefore, the tenth embodiment and the eleventh embodiment
As in the embodiment described above, the true used band is used instead of the used band value that is larger than the actual state, so that there is an effect of reducing the possibility that the predicted value becomes excessive. (Thirteenth Embodiment) Another embodiment of the present invention will be described. This form is a modification of the tenth embodiment, and the judgment formula of step S66 in FIG. 20 is changed. In the tenth embodiment, the determination formula in step S66 is (recv.idle (t) = 0) or (recv.data (t) = 0), (α1 =
2, α2 = 10, β = 2.5, γ = P.GUB) In other cases, (α1 = 1, α2 = 2, β = 2.5, γ = 0)
Met.

Instead of this, in this form, (recv.idle (t) = 0), (α1 = 2, α2 = 10, β = 2.5,
γ = P.GUB) (recv.idle (t)> 0), (α1 = 1, α2 = 2, β = 2.5, γ
= 0. For this reason, unlike the tenth embodiment, staircase prediction is not always performed when there is no bandwidth allocation in the period of the next bandwidth allocation cycle in which the use of the band is used, and there is an effect that unnecessary bandwidth allocation can be reduced. .

(Fourteenth Embodiment) Another embodiment of the present invention will be described. This form is a modification of the tenth embodiment, and the judgment formula corresponding to step S66 in FIG. 20 or FIG. 23 is changed. FIG. 24 shows an operation example of this form. In FIG. 24, the portion corresponding to the judgment formula of step 66 of FIG. 23 is step S66.
A, S66B, and the results of those determinations are changed to three steps S67A, S67B, S68.

In the tenth embodiment, step S66.
The judgment formula of (recv.idle (t) = 0) or (recv.data (t) = 0) is (α1 =
2, α2 = 10, β = 2.5, γ = P.GUB) In other cases, (α1 = 1, α2 = 2, β = 2.5, γ = 0)
Met.

Instead of this, in this form, the combination of (recv.idle (t) = 0) and (recv.data (t)> 0), (α1 =
2, α2 = 10, β = 2.5, γ = P.GUB) (recv.idle (t) = 0) and (recv.data (t) = 0), (α1 =
1, α2 = 1, β = 2.5, γ = 0) (recv.idle (t)> 0) and (recv.data (t)> 0), (α1 =
1, α2 = 2, β = 2.5, γ = 0).

Therefore, in this embodiment, staircase prediction is not always performed when there is no band allocation in the period of the next band allocation cycle in which the use of the band is used, and the prediction is performed less than in the thirteenth embodiment. This has the effect of reducing invalid bandwidth allocation. (Fifteenth Embodiment) Another embodiment of the present invention will be described. This form is a modification of the fourteenth embodiment. FIG. 25 shows an operation example of this form.

In FIG. 25, step S67C and the subsequent processing part are changed as follows. That is, in step S67C of this embodiment, step S6
Instead of performing the calculation of 9 (Equation (6)), the allocated bandwidth (alloc (t-1)) in the period of the immediately preceding bandwidth allocation cycle is used as it is. That is, the allocated band (alloc (t-1)) is set as the allocated band (alloc (t)) in the period of the current band allocation cycle.

Therefore, in this embodiment, the staircase prediction is not always performed when there is no band allocation in the period of the next band allocation cycle in which the band is used. Further, since the allocation is performed according to the allocation value (alloc (t)) in the immediately preceding prediction, the bands to be allocated in the adjacent periods of the band allocation cycle are largely different from those in the thirteenth and fourteenth embodiments. It also reduces the possibility of vibration.

(Sixteenth Embodiment) One embodiment of the band allocating apparatus and the band allocating method of the present invention will be described with reference to FIGS. 1 and 26. This form is a modification of the first embodiment. FIG. 26 is a flowchart showing an operation example of the band prediction unit in this embodiment. The band prediction process is different between this mode and the first embodiment. As a specific operation of the band prediction unit 30 in this mode, the operation shown in FIG. 26 can be considered. In this operation example, it is assumed that band prediction is repeatedly performed at a constant cycle (bandwidth allocation cycle).

In this embodiment, the information that can be used for prediction in the j-th allocation period (BAP # (j)) of the band allocation cycle is for each allocation period (BAP # (j-1)) one cycle before. Used bandwidth (recv.data (k)) in the period (BAP # (k)), unused bandwidth that was allocated but not used (recv.idle (k)), current allocation period (BAP # (k) )) Is unallocated at the end of the next allocation period (BAP
The remaining queue length information (que (k)) carried over to # (k + 1)) and declared as the declared value, and the predicted bandwidth predicted in the allocation period (BAP # (j)) is the allocation period after one cycle. It is assumed that it will be reflected as the allocated bandwidth (alloc (j + 1)) predicted by (BAP # (j + 1)).

Therefore, the predicted input band INPUT (j + 1) predicted to be newly input in the (j + 1) th allocation period (BAP # (j + 1)) of the band allocation cycle is the latest input available for prediction. It is predicted that the bandwidth will be almost balanced with the input bandwidth in the immediately preceding allocation period (BAP # (j-1)). Allocation period (BAP # (j
The input bandwidth in (-1)) is equal to the sum of the bandwidth used in the allocation period (BAP # (j-1)) and the growth rate of the queue length. This means that when the queue length decreases, the used bandwidth in the allocation period (BAP # (j-1)) becomes larger than the input bandwidth by the decrease in the queue length. If the queue length increases,
Since the queue length is increased by the shortage of the actual allocated bandwidth (Determ (j-1)) with respect to the input bandwidth, the used bandwidth is the used bandwidth plus the increased queue length.

However, the next allocation period (BAP # (j +
1)) in the current allocation period (BAP # (j)), the used bandwidth (recv.data (j-
1)) and the declared value (que (j-1)) are used for prediction, so an error occurs. Therefore, after the declared value (que (k)) starts to increase from 0 to a value other than 0, the queue after the lapse of time Th after subtracting one cycle from the allocation delay (two cycles of the bandwidth allocation cycle in this embodiment) Prediction is performed by adding the length (que (k)) to the prediction band.

Therefore, the band prediction formula in FIG. 26 is as follows. alloc (j) = [recv.data (j-1) + que (j-1) -que
(j-2)] × α1 + α2 + β × que (j-1) where α1 and α2 are the allocation periods immediately after the prediction (BAP # (j-1)) from the input band in the allocation period immediately before the prediction (BAP # (j-1)). BAP # (j + 1)) is a proportional coefficient and an additional component for increasing and predicting the band, and is a constant value in this example. In FIG. 26, the value of β is set to 1 only when Th has passed since the declared value was set to 0 using the inherited value X, and is set to 0 otherwise.

As described above, in this embodiment, assuming that the same input as the immediately preceding allocation period of the bandwidth allocation cycle is input to the line also in the immediately following allocation period, the fixed residual queue length due to the allocation delay is Since the bandwidth is allocated according to the residual queue length in the allocation period in consideration of the delay, it is possible to reduce the excessively allocated bandwidth and quickly allocate the bandwidth. (Seventeenth Embodiment) FIGS. 22 and 2 show one embodiment of the band allocating apparatus and the band allocating method of the present invention.
This will be described with reference to FIG. This form is a modification of the second embodiment. FIG. 27 is a flowchart showing an operation example of the band prediction unit in this embodiment.

The band prediction processing is different between this embodiment and the second embodiment. As a specific operation of the band prediction unit 30 in this form, the operation shown in FIG. 27 can be considered. In this operation example, it is assumed that band prediction is repeatedly performed at a constant cycle (bandwidth allocation cycle).

In this mode, the band IN predicted to be input to the line during the period of the allocation cycle which is the object of allocation calculation
The sum of the predicted value of the residual queue length in the buffer of the allocated device that is not allocated before the period of the allocation cycle and is carried over to the period of the allocation cycle is predicted as the band to be allocated. In this form, the information that can be used for prediction in the j-th allocation period (BAP # (j)) of the band allocation cycle is the period (BAP # (j-1)) one cycle before (BAP # (j-1)). #
(k)) used bandwidth (recv.data (k)), unused bandwidth that was allocated but not used (recv.idle (k)), end of current allocation period (BAP # (k)) Residual queue length information (que (k)) that is sometimes unallocated and carried over to the next allocation period (BAP # (k + 1)) and declared as the declared value, allocation period (BAP # (k))
Actually allocated bandwidth (Datem
(k)), and the predicted bandwidth predicted in the allocation period (BAP # (j)) is the allocated bandwidth predicted in the allocation period (BAP # (j + 1)) one cycle later (alloc (j + 1)) It is supposed to be reflected as.

In this embodiment, similarly to the sixteenth embodiment, the predicted input band INPUT (j + 1) predicted to be newly input to the line in the allocation period (BAP # (j + 1)) is Latest available input bandwidth, that is, the immediately preceding allocation period (BA
It is predicted that the input band at P # (j-1)) is almost balanced. However, in the allocation period (BAP # (j)), the allocated bandwidth (al
actual allocated bandwidth (detem (j)) actually allocated to loc (j))
May not be met and the input band prediction may not match.

Therefore, in this embodiment, the queues remaining after the growth rate of the queue length has stopped are assigned all together. Therefore, the band prediction formula in FIG. 27 is as follows. alloc (j) = [recv.data (j-1) + que (j-1) -que (j-2)] ×
α1 + α2 + β × que (j-1) where α1 and α2 are the allocation period (BAP # (j + 1) immediately after prediction using the input band in the allocation period (BAP # (j-1)) immediately before prediction. )) Is a proportional coefficient and an additional component for increasing and predicting the band, and is a constant value in this example.

In FIG. 27, the immediately preceding allocation period (BAP # (j-
1)) and the declared value que (-in the allocation period (BAP # (j-2)) before that)
If 1) and que (-2) are compared and they match (β = 1), otherwise, it is considered that the input band and the allocated band do not match (β = 2). As described above, in this embodiment, it is assumed that the same input as the allocation period immediately before the bandwidth allocation cycle is input to the line in the allocation period immediately after, but the residual queue length due to allocation delay or line contention is Since the expansion is stopped and then the blocks are collectively allocated, the excessively allocated band can be reduced and the band can be quickly allocated.

(Eighteenth Embodiment) One embodiment of the band allocating apparatus and the band allocating method of the present invention will be described with reference to FIGS. 1 and 28. This form is a modification of the second embodiment. FIG. 28 is a flowchart showing an operation example of the band prediction unit in this form. In the tenth embodiment described above, the prediction is performed based on only the usage rate of the allocated band to determine whether the allocated band is in the increasing state or in the stable state. Also,
The observation values (recv.data (t)) and (recv.idle (t)) of the traffic observation section are used to judge the increase state or the stable state.

However, due to fluctuations in the input band and shifts in the observation timing, the observed values (recv.data (t)) and (rev.data (t))
The value of cv.idle (t) may fluctuate. If the additional bandwidth (corresponding to α2) allocated to detect that the requested bandwidth is increasing is small, it is determined that the requested bandwidth is increasing due to fluctuations in the observed values even if the bandwidth is not increasing. It may happen. The increase in the prediction band due to such an erroneous determination results in useless band allocation, which causes a decrease in band use efficiency.

Observation values (recv.data (t)) and (recv.i
It is conceivable to increase the excess bandwidth as a means to prevent erroneous judgment due to dle (t)). However, increasing the excess band itself leads to a decrease in band use efficiency. Therefore, a condition used for determining whether the requested bandwidth is in the increasing state or the stable state is added to suppress unnecessary transition to the increasing state.

In this embodiment, the queue length increase amount (que.inc (t)) is used to suppress the transition to the increase state. The queue length increase amount (que.inc (t)) is a value obtained by correcting the difference between the past queue length and the latest queue length declared by the ONT with the allocated bandwidth or the used bandwidth. The operation of this form is as shown in FIG.

As in the tenth embodiment, in step S182 of FIG. 28, it is determined whether the required bandwidth is increasing or stable. When it is determined in step S182 that the requested bandwidth is in the increasing state, a determination is made using the queue length increase amount (que.inc (t)) in the next step S183. When the queue length increase amount (que.inc (t)) is equal to or greater than a predetermined threshold Th defined as a value of 0 or more in step S183, it is determined that the requested bandwidth is in an increasing state, and is less than the threshold Th. Is judged to be in a stable state.

Due to the fluctuation of the observed value, almost no unallocated cells are retained in the buffer for the line existing in the band allocation device. For this reason, the queue length increase amount (que.inc (t)) does not exceed the threshold value. Therefore, by making a determination in step S183, it is possible to prevent excessive bandwidth prediction due to erroneous determination. (Nineteenth Embodiment) One embodiment of the band allocating apparatus and the band allocating method of the present invention will be described with reference to FIGS. 1, 29 and 30. This form is a modification of the second embodiment. FIG. 29 is a flowchart showing an operation example of the band prediction unit in this embodiment. FIG. 30 is a schematic diagram showing transition conditions regarding the suppression flag of linear prediction and staircase prediction in this embodiment.

In the eighteenth embodiment, the queue length increase amount (que.inc (t)) is used to judge whether the requested bandwidth is in the increasing state or in the stable state. However, the increase in queue length (que.
inc (t)) cannot be used.

Therefore, in this embodiment, as shown in FIG. 30, control is performed so as to suppress the transition from the stable state to the increase state in the prediction process. The state transition to be suppressed is from stable state (target of linear prediction) to increasing state (target of staircase prediction)
To 0 and the transition from the 0 allocation state where there is no band allocation to the increase state. In the eighteenth embodiment, the queue length increase amount (que.inc
Although the judgment is made in (t)), since two state transitions can be classified in this form, a judgment is made by giving a threshold value for suppression to each state transition individually.

The operation of this mode is as shown in FIG. 29, and the change of the value of the suppression flag at the time of state transition is shown in FIG.
As shown in. Here, a suppression flag (restrict.flag) is provided in order to suppress the transition from the stable state to the increased state in the prediction process, and the suppression flag (restrict.fl
The increasing and steady state treatments shall be selected according to the value of ag).

Further, Th1 and Th2 are set to values of 0 or more as the thresholds referred to by the restriction flag (restrict.flag). The threshold Th1 determines the number of cycles that suppress the transition from the stable state to the increase state, and the threshold Th2 determines the number of cycles that suppress the transition from the 0 allocation state to the increase state. When the prediction process is performed in the increasing state, the value (-1) indicating the increasing state is substituted in the restriction flag (restrict.flag) in step S193 of FIG. This value is also used in step S182 to determine the band allocation state in the previous band allocation cycle. When it is determined that the increase state continues from the previous band allocation cycle, the increase state can be continuously maintained. However, once in the stable state, the threshold value Th1 is assigned to the restriction flag (restrict.flag) in step S1942. As a result, step S is performed after the next band allocation period.
The identification of 191 suppresses the increase. For this reason, when there is a band allocation request, band prediction is performed using the stable state parameters (α1, α2, γ). The suppression of transition to the increasing state is continued until the value of the restriction flag (restrict.flag) is subtracted and becomes 0 in step S196.

Further, when the band allocation changes from a certain state to the 0 allocation state, the transition from the 0 allocation state to the increase state is suppressed. When it is in the 0 allocation state, the threshold value Th2 is assigned to the restriction flag (restrict.flag) in step S200 of FIG. When a bandwidth allocation request is generated after the next bandwidth allocation cycle (when it is determined Yes in both the identifications of S62 and S181), the increase state is set until the restriction flag (restrict.flag) becomes 0. Suppress the state transition of. To suppress the transition to the increasing state,
At 196, the value of the restriction flag (restrict.flag) is subtracted, and the operation is continued until it becomes zero.

With this form of control, it is possible to prevent an unnecessary transition to the increased state and excessive band prediction in the prediction process. (Twentieth Embodiment) FIGS. 22 and 31 show one embodiment of the band allocating apparatus and the band allocating method of the present invention.
Will be described with reference to. This embodiment corresponds to claim 22, claim 30, claim 34, and claim 42. This form is a modification of the second embodiment. FIG. 31 is a flowchart showing an operation example of the band prediction unit in this mode.

The traffic observing means, the band request receiving means, the band predicting means and the band allocating means of claim 25 correspond to the traffic observing section 10, the band request receiving section 20, the band predicting section 30 and the band allocating section 40, respectively. . As shown in FIG. 22, the bandwidth allocating device of this embodiment is the traffic observing unit 1
0, a band request receiving unit 20, a band predicting unit 30, and a band allocating unit 40.

The obijo allocation unit 40 allocates a band for each line according to the size of the band predicted by the band prediction unit 30 (the band required at the present time). The traffic observation unit 10
The bandwidth allocated by the bandwidth allocating unit 40 for each line and the amount of traffic actually flowing through the line, that is, the usage status of the band are observed. The bandwidth request receiving unit 20 receives a declared value of an allocation request bandwidth for each line, which is input by a customer or the like.

The band predicting unit 30 includes the traffic observing unit 10
The present allocation request band is predicted based on at least one of the information of the used band obtained by the observation of the above and the reported value of the allocation request band received by the band request receiving unit 20 and the past allocation band of the band allocation unit 40. To do. The band allocation unit 40 allocates the band predicted by the band prediction unit 30 to each line. The band prediction process is different between this mode and the second embodiment. As a specific operation of the band prediction unit 30 in this form, the operation shown in FIG. 31 can be considered.

In the operation example of FIG. 31, it is assumed that the band prediction is repeatedly performed at a constant cycle (bandwidth allocation cycle). In this mode, the band INPUT predicted to be input to the line in the band allocation cycle for which the allocation is calculated.
Then, the sum of the predicted values of the residual queue lengths in the buffers of the allocated devices that are not allocated before the corresponding allocation period of the band allocation cycle and are carried over to the allocation period is predicted as the allocation target band.

[0209] In this embodiment, the information usable for prediction in the j-th period (BAP (#j)) of the band allocation period is up to the band allocation period (BAP (# j-1)) one period before. Used bandwidth recv.data (k) in each period (BAP (#k)), unused bandwidth recv.idle (k) allocated but not used, at the end of the corresponding period (BAP (#k)) The remaining queue length information que (k) that has not been allocated and is carried over to the next band allocation cycle (BAP (# k + 1) and declared as the declared value and the band allocation cycle period (BAP (#k)) are actually used. It is assumed that the bandwidth is allocated to the line by the bandwidth allocator, and the predicted bandwidth predicted in the period (BAP (#j)) of the bandwidth allocating period is the period (BAP # (j +) after one period.
It is reflected as the allocated bandwidth alloc (j + 1) predicted in 1)).

Also, in this embodiment, the observed use band re
It is assumed that the input INPUT to the line equivalent to cv.data (j-1) is in the period (BAP (# j + 1)) of the band allocation cycle. However, the period (BAP (#
In j)), the actual allocated bandwidth d is allocated to the allocated bandwidth alloc (j) predicted to be allocated due to the competition between the lines.
In some cases, the item (j) may not be satisfied, and the prediction of the input band may not match.

Therefore, in this embodiment, the data is not allocated in the current band allocation cycle (BAP (#j)) and remains in the buffer of the allocated apparatus in the next band allocation cycle (BAP (# j + 1)). The queue length to be allocated is set to the previous bandwidth allocation period (B
AP (# j-1)), the current bandwidth allocation period (BAP (# j-1))
j)) The declared value que, which is the queue length carried over to
(J-1) is the allocated bandwidth al used in the current bandwidth allocation cycle (BAP (#j)) predicted in the previous bandwidth allocation cycle.
A value obtained by subtracting the actual allocation bandwidth detem (j) actually allocated in the current bandwidth allocation cycle (BAP (#j)) from loc (j) is added, and the predicted queue length Preque is calculated using one of the following expressions. Predict (j).

Preque (j) = max (buff, min (0, que (j-1) + alloc (j) -detem (j))) (A 1) Preque (j) = max (buff, min (0, que (j-1) + max (0, alloc (j) -detem (j))) ... (A2) where (buff) is the line to be used by the assigned device This means the maximum queue length. The smaller value compared with (buff) is used because the queue length is not predicted to be excessively large beyond the maximum value of the queue length that can be used. Further, the reason why it takes a larger value compared to 0 is that the queue length is not a negative number.

In the latter expression (A2), (alloc
The maximum value of (j) -detem (j)) and 0 is selected
This is because the queue length is not underestimated when the predicted bandwidth is underestimated. Valid declared value que (j-1)
Is received by the request receiving unit, que (j-1) is input to the equation (A1) or (A2) to predict the current predicted queue length Preque (j).

When the request receiving unit does not receive the valid declared value que (j-1), the value q is predicted using the formula (A1).
Predict (j-1) is input instead of ue (j-1) for prediction. This is (Preque (j) in the formula (A2).
This is because ≦ Preque (j + 1)) will continue to be maintained and the queue length will not be reduced, so that an excessively large queue length will be continuously predicted. If the request receiving unit does not receive the valid declared value que (j-1), que (j-1) is changed to Preque (j-1).
If the value of Preque (j) according to the formula (A1) replaced with is no longer positive (cumulative value becomes 0) or the number of unused cells recv.Idle (j-1) exceeding a predetermined number is traffic When observed by the observation section (recv.idle (j-1)> Th.preq),
(Preque (j) = 0)

Therefore, the band prediction formula in FIG. 31 is as follows. alloc (j) = [recv.data (j-1)] x α1 + α2 + β x Preque (j)
＋ γ = [recv.data (j-1)] × α1 ＋ α2 ＋ max (buff, min (0, qu
e (j-1) + alloc (j) -detem (j)) + γ where α1 and α2 are the band allocation cycle (B
AP (# j-1)) is a proportional coefficient and an addition component for additional prediction of the band of the band allocation cycle (BAP (# j + 1)) immediately after prediction using the input band of AP (# j-1). β is a proportional coefficient for bandwidth prediction related to queue length. The variable γ is a predetermined band P.P. It is used to perform bandwidth allocation for a predetermined bandwidth such as a guaranteed bandwidth when the input bandwidth tends to increase below GUB (for example, guaranteed bandwidth, maximum bandwidth, bandwidth in which fixed allocation is fixedly removed from the guaranteed bandwidth). These have different values in linear prediction and in staircase prediction. P. MAB means the maximum bandwidth allowed for the line.

(Α1, α2, γ) is, for example, (1, 0, 0) in the linear prediction and (2, 10, P.
GUB). The transition conditions between the linear prediction and the staircase prediction with the set values of (α1, α2, β, γ) are shown below. (Acquisition of valid declared value que (j-1)) and (que (j-1) -Preque (j-1)> Th.sr) ... (A3) or (valid declared value que (j- (No acquisition of 1)) and (Preque (j-1) = 0) and ((recv.idle (j-1)) = 0 times (IdleCnt)> Th.nsr) ..... (A4) Above Case staircase prediction: (α1, α2, β, γ) = (2,1
0,1, P.GUB) Transition to linear prediction other than the above (α1, α2, β, γ) = (1,0,
1,0) or (1,1,1,0) where Th.preq: threshold value for clearing the predicted queue length to 0; 2
Cell degree Th.sr: threshold value for predicting stairs related to queue length; 10
Cell degree Th.nsr: threshold value for staircase prediction based on the number of unused cells; about 3 cells.

The condition of the expression (A3) is used to perform swiftly assigned castles by the staircase prediction because the queues that cannot be predicted by the linear prediction are extended. Expression (A
The condition 4) is used for the following reason. When the valid reception is received by the request receiving unit and the input according to the predicted value is input to the line, it is an ideal allocation state that the number of unused cells becomes 0, so linear prediction To do. If a valid declaration has not been received by the request receiving unit, accumulation of estimated bandwidth to be allocated past the bandwidth allocation cycle (BAP (# j-1) (Σalloc (k) [k = 1 to j-1] ],here
[k = 1 to j-1] means the total sum of the subscript k from 1 to (j-1)) and the accumulated actual allocated bandwidth (Σ
detem (k) [k = 1 to j-1]) is other than 0, the unallocated portion should remain as the queue, or the predicted queue length Preque (j) is It is appropriate that the number of unused cells becomes 0 until it becomes 0.

Therefore, when (Preque (j)> 0),
Even if (recv.idle (j) = 0), staircase prediction is not performed.
If this is (Preque (j)> 0) then (recv.data (j-
Stairway prediction is not necessary because the allocations of 1) and (Preque (j)) and above are made.

With such a transition of the prediction, it is possible to realize a less wasteful band allocation without erroneously predicting the staircase in a situation where the staircase prediction should not be performed, such as an ideal allocation situation. (Twenty-first Embodiment) One embodiment of the band allocating apparatus and the band allocating method of the present invention will be described with reference to FIGS. This form is claim 22, claim 3
It corresponds to 0, claim 34, and claim 42.

This embodiment is a modification of the twentieth embodiment. The difference between this form and the twentieth embodiment lies in the method of acquiring the actual allocated bandwidth detem (j). FIG. 31 shows
It is a flowchart which shows the operation example of the band estimation part in this idea. In this mode, the traffic observing means, the band request receiving means, the band request predicting means and the band borrowing means according to claim 22 are respectively the traffic observing section 10, the band request receiving section 20, the band request predicting section 30 and the band allocating section 40. Equivalent to.

As shown in FIG. 1, the band allocating apparatus of this embodiment comprises a traffic observing section 10, a band request receiving section 20, a band prediction section 30 and a band allocating section 40. The band allocating unit 40 allocates a band for each line according to the size of the band predicted by the band predicting unit 30 (the band currently required). The traffic observing unit 10 observes the band allocated by the band allocating unit 40 for each line and the amount of traffic actually flowing through the line, that is, the use condition of the band.

The band request receiving section 20 receives the declared value of the band requested for each line, which is input by the customer or the like. The band prediction unit 30 predicts the current allocation request band based on at least one of the information on the used band obtained by the observation of the traffic observing unit 10 and the declared value of the allocation request band received by the band request requesting unit 20. . The band predicted by the band prediction unit 30 is allocated to each line by the band allocation unit 40.

The difference between this embodiment and the twentieth embodiment lies in the band prediction processing. In the twentieth embodiment, the actual allocated bandwidth detem (j) is acquired from the Obihiro allocation unit, but in this embodiment, it is possible to allocate the sum of the allocated bandwidths sub-measured by the bandwidth prediction unit and the entire line to which the bandwidth is allocated. The actual allocated band castle is predicted by comparing it with another band and calculating the excess band. In this mode, there is an effect that it is possible to predict the allocated band without the band predicting unit acquiring information about the actual allocated band of the band allocating unit. Therefore, this form is particularly suitable when there is no priority of bandwidth allocation between lines.

(Twenty-second Embodiment) Another embodiment of the band allocating apparatus and the band allocating method of the present invention will be described. This form corresponds to claims 29 and 41. This form is a modification of the second embodiment. The difference between this form and the second form of smile is the band prediction process.

In this mode, when the band constantly input to the line does not become an integral multiple of the number of cells per band allocation period, it is predicted that the invalid band allocation will be reduced and the necessary band will be allocated. There is a feature in doing it. for that reason,
In this form, the used band recv.data (j-1) in the band allocation cycle that is earlier than the immediately preceding band allocation cycle (BAP (# j-1)) in the band allocation cycle (BAP (#j)) is smoothed. The converted value ave (j-1) is calculated and used in place of the used band recv.data (j-1) in the prediction formula of the twentieth embodiment.

Ave (j-1) is calculated by one of the following expressions. ave (j) = ave (j-1) × ε1 + (1-ε) × (recv.data (j-2) + ave (j-2) −detem (j-2) ・ ・ ・ (B1) ave ( j) = ave (j-1) + ε2 × (recv.data (j-2) −detem (j) −ave (j-1) ··· (B2) where ε1: smoothing coefficient; for example ((( Number of cycles to be smoothed-1) / (number of cycles to be smoothed)) ε2: Smoothing coefficient considering the average deviation; eg ((number of cycles to be smoothed-1) / (number of cycles to be smoothed) )) Note that the term (ave (j-2) -detem (j-2) is the actual allocated bandwidth de
This is a correction when tem (j) is less than the predicted allocated bandwidth, but when the residual queue length is received by the request receiving unit as the declared value of the effective bandwidth request, it is replaced by the growth rate of the queue length. good. The smoothed value is represented by a finer band granularity than the band granularity according to the number of cells in each band allocation cycle.

Therefore, the allocated bandwidth in a plurality of bandwidth allocation periods
As the average value of alloc (t), in order to match the smoothed value and reduce the invalid allocation, the following prediction formula is used to average and match the predicted band. alloc (j + 1) ＝ [ave (j-1)] + mod (ave (j-1), UNIT) + ave.rest
(j-1)] × α1 + α2 + β × γ ave.rest (j-1) = ave.rest (j-2) + recv.data (j-1) -ave (j-
2) Here, UNIT: cumulative value of a band ave.rest: ave (j + 1) corresponding to one cell per band allocation period, which is not allocated because it does not become 1 cell in the band allocation period [A ]: Integer part of A Mod (A, B): A is the remainder when divided by B.

This prediction formula is the allocated bandwidth alloc by prediction.
It is assumed that (input band = recv.data (j)) can be predicted unless (j) or the actual allocated band detem (j) is less than the input band. Preque (j) becomes positive when the predicted queue length occurs when the predicted allocated bandwidth alloc (j) or the actual allocated bandwidth detem (j) is less than the input bandwidth.
Here, α1 and α2 are band allocation periods (B
AP (# j-1)) is a proportional coefficient and an addition component for extra-prediction of the band of the band allocation period (BAP (# j + 1)) immediately after sub-measurement with respect to the input band of AP (# j-1). β is a proportional coefficient for band prediction related to the queue length. The variable γ is a predetermined obi castle P.GUB (for example, a guaranteed band,
It is used to allocate a certain amount of bandwidth, such as the guaranteed bandwidth, when the input bandwidth tends to increase below the maximum bandwidth and the bandwidth that is fixed bandwidth less the guaranteed bandwidth. P.MAB means the maximum bandwidth allowed for the line.

This form has a value ave smoothed at the time of prediction.
The case where (j-1) is used is divided into the case where ave (j-1) is not used. When the fluctuation of the smoothed value ave (j-1) is less than or equal to a value obtained by multiplying a predetermined threshold value or deviation by a predetermined value, the prediction is performed using ave (j-1), and the band allocated according to the prediction. When the actual allocation due to competition is insufficient due to the difference from the input bandwidth to the line for (ave (j-2) −
detem (j-2)), if the value exceeds a predetermined threshold value or exceeds the deviation multiplied by a predetermined value, a
If the input fluctuates from the band predicted using ve (j-1), recv.data (j-1) is used as it is instead of ave (j-1).

That is, -Th.ave <recv.data (j-1) -av
From the state where e (j-1) <Th.ave recv.data (j-1) is used as it is to the state where ave (j-1) is used, (-Th.ave≤recv.data (j-1)- ave (j-1) ≤ + Th.
ave) is satisfied a predetermined number of times (about 1 / ε). This state transition is shown in FIG. Where Th.av
e is a threshold value or deviation that should be satisfied by the difference between the smoothed value ave (j-1) and the expected value.

An example of the smoothed average deviation in the case of the formula (B2) multiplied by a coefficient is shown in the following formula. Smoothed average deviation: D (j) = D (j-1) + h × (| recv.dat
a (j) -ave (j) | -D (j-1)) h: coefficient As described above, in this form, when the input is constant, the number of cells corresponding to the band allocation period It is possible to reduce the invalid allocation regardless of the granularity of the band.

[0232]

As described above, according to the present invention, compliance with guaranteed bandwidth allocation for traffic that changes in bursts,
It is possible to prevent buffer overflow in ONT, suppress increase in round trip time due to allocation delay, suppress unreasonably small bandwidth allocation for sporadic users of bandwidth, and suppress generation of invalid bandwidth.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a band allocation device according to a first embodiment.

FIG. 2 is a flowchart showing an operation example (1) of a band prediction unit in the first embodiment.

FIG. 3 is a flowchart showing an operation example (2) of the band predicting unit in the first embodiment.

FIG. 4 is a flowchart showing an operation example (3) of the band predicting unit in the first embodiment.

FIG. 5 is a block diagram showing a configuration of a band allocation device according to a second embodiment.

FIG. 6 is a flowchart showing an operation example of a linear prediction unit.

FIG. 7 is a flowchart showing an operation of a selection unit according to the second embodiment.

FIG. 8 is a state transition diagram showing state transitions of a band predicting unit according to the second embodiment.

FIG. 9 is a time chart showing an operation example (1) of the second embodiment.

FIG. 10 is a time chart showing an operation example (2) of the second embodiment.

FIG. 11 is a flowchart showing the operation of the linear prediction unit according to the third embodiment.

FIG. 12 is a block diagram showing a configuration of a band predicting unit according to a fourth embodiment.

FIG. 13 is a state transition diagram showing state transitions of the band predicting unit according to the fourth embodiment.

FIG. 14 is a block diagram showing a configuration of a band predicting unit according to a fifth embodiment.

FIG. 15 is a state transition diagram showing state transitions of the band predicting unit according to the fifth embodiment.

FIG. 16 is a block diagram showing a configuration of a staircase prediction unit according to a sixth embodiment.

FIG. 17 is a block diagram showing a configuration of a linear prediction unit according to the seventh embodiment.

FIG. 18 is a block diagram showing a configuration of a linear prediction unit according to the eighth embodiment.

FIG. 19 is a state transition diagram showing state transitions of the band predicting unit according to the ninth embodiment.

FIG. 20 is a flowchart showing an operation example of a band predicting unit in the tenth embodiment.

FIG. 21 is a flowchart showing an operation example of a band prediction unit in the eleventh embodiment.

FIG. 22 is a block diagram showing a configuration of a band allocation device according to a tenth embodiment.

FIG. 23 is a flowchart showing an operation example of a band predicting unit in the tenth embodiment.

FIG. 24 is a flowchart showing an operation example of a band predicting unit in the fourteenth embodiment.

FIG. 25 is a flowchart showing an operation example of a band prediction unit in the fifteenth embodiment.

FIG. 26 is a flowchart showing an operation example of a band prediction unit in the sixteenth embodiment.

FIG. 27 is a flowchart showing an operation example of a band prediction unit in the seventeenth embodiment.

FIG. 28 is a flowchart showing an operation example of a band prediction unit in the eighteenth embodiment.

FIG. 29 is a flowchart showing an operation example of a band prediction unit in the nineteenth embodiment.

FIG. 30 is a state transition diagram showing state transitions in the nineteenth embodiment.

FIG. 31 is a flowchart showing an operation example of a band prediction unit in the twentieth embodiment.

FIG. 32 is a state transition diagram showing state transitions according to the 22nd embodiment.

[Explanation of symbols]

10 Traffic Observation Department 20 Band request receiver 30 Band Prediction Unit 31 Linear predictor 32 Stairs prediction unit 33 Selector 34 Prediction value excess discriminator 35 Unused band identification section 36 All-use identification part 37 Continuation Count Identification Unit 40 Band Allocation Unit 311 Linear proportional predictor 312 Difference Prediction Unit 313 Adder 321 Predicted Band Below Guaranteed Band 322 Guaranteed band or higher prediction unit 323 Guaranteed Bandwidth Allocation Identification Unit

   ─────────────────────────────────────────────────── ─── Continued front page    F-term (reference) 5K030 HA10 JL03 LC05 LE16 MB09                       MB15                 5K051 AA01 BB02 CC02 DD09 DD13                       DD14 FF02 FF03 FF12

Claims (46)

[Claims] 1. A band allocating device for dynamically allocating at least a part of a shared band shared by a plurality of lines to each line, and traffic observing means for observing a use state of a band allocated to each line. Bandwidth request receiving means for receiving the declared value of the allocation request bandwidth input for each line, and at least the used bandwidth of each line detected by the traffic observation means and the declared value of the allocation request bandwidth received by the bandwidth request receiving means A band allocating device comprising: a band predicting unit for predicting an allocation request band of each line based on one side; and a band allocating unit for allocating a band corresponding to a prediction result of the band predicting unit for each line. . 2. The band allocating device according to claim 1, wherein the band predicting means uses a first predicting means for performing linear prediction, a second predicting means for performing non-linear prediction, and all allocated bands. The used bandwidth identification means for obtaining the above-mentioned ratio for each line based on the history of past observed values and comparing it with the first threshold, and the history of at least one of the declared value of the input bandwidth and the usage status of the assigned bandwidth. On the basis of the above, the excess identification means for obtaining the ratio of the declared value exceeding the prediction value of the first prediction means for each line and comparing it with the second threshold value; the identification result of the used band identification means and the excess identification means. A band allocating device further comprising: a selection unit that selects one of the predicted value of the first prediction unit and the predicted value of the second prediction unit based on the identification result. 3. The band allocating device according to claim 2, wherein the selecting unit identifies the used band identifying unit and the excess identifying unit until a predetermined period or a predetermined number of times elapses after the prediction is started. A band allocation device, characterized in that the prediction value of the second prediction means is selected independently of the above. 4. The bandwidth allocating device according to claim 2 or 3, wherein the second predicting means is a guaranteed bandwidth of a line, a current used bandwidth, a current allocated bandwidth, and a predicted value of the first predicting means. A bandwidth allocation device characterized by selecting and outputting the maximum value among the predicted values. 5. The band allocating device according to claim 2 or 3, further comprising: guaranteed band identification means for comparing a previously predicted advance predicted value with a threshold of a guaranteed band of the line, and the second prediction. Means, when the preceding predicted value is less than or equal to the guaranteed bandwidth of the line,
The maximum value among the guaranteed bandwidth of the line, the currently used bandwidth, the current allocated bandwidth and the predicted value of the first prediction means is selected as the predicted value, and when the preceding predicted value exceeds the guaranteed band of the line. As a result of dividing the usable shared band by the total number of lines or the guaranteed band ratio of all lines, the current used band, the current allocated band, and the maximum value among the predicted values of the first prediction means are used as predicted values. A bandwidth allocating device characterized by selecting. 6. The bandwidth allocating device according to claim 2 or 3, wherein the first predicting means assigns a predetermined coefficient to a declared value of a used bandwidth or allocation request bandwidth observed at at least two past time points. A bandwidth allocating device, characterized by predicting a sum or difference of multiplied values as a current allocation request bandwidth. 7. The bandwidth allocating device according to claim 2 or 3, wherein the first predicting means is a value linearly proportional to a used bandwidth observed at at least two past times or a declared allocation request bandwidth. And a value obtained by adding a value obtained by multiplying a previously declared allocation request bandwidth by a coefficient or a value obtained by multiplying a value obtained by multiplying a previously declared bandwidth by a coefficient with respect to the last-minute report bandwidth to a current predicted value. Allocation device. 8. The bandwidth allocating device according to claim 7, wherein the coefficient is the number of control cycles corresponding to the time required from the transmission of the declared value in the ONT to the reflection of the declared value in the bandwidth allocation in the OLT. A bandwidth allocating device characterized by being applied as. 9. A traffic observing procedure for observing a use state of a band allocated to each line in a band allocation method for dynamically allocating at least a part of a shared band shared among a plurality of lines to each line, Based on at least one of the bandwidth request receiving procedure for receiving the declared value of the allocation request bandwidth input for each line, and the detected used bandwidth of each line and / or the received declared value of the allocation request bandwidth, the allocation request bandwidth of each line And a band allocation procedure for allocating a band corresponding to a prediction result of the band prediction procedure for each line. 10. The band allocation method according to claim 9, wherein, as the band prediction procedure, a first prediction procedure for performing linear prediction, a second prediction procedure for performing non-linear prediction, and all allocated bands are used. The used bandwidth identification procedure for obtaining the ratio of each line based on the history of past observed values and comparing it with the first threshold, and the history of at least one of the declared value of the input bandwidth and the usage status of the assigned bandwidth. On the basis of the excess identification procedure of obtaining the ratio of the declared value exceeding the predicted value of the first prediction procedure for each line and comparing it with the second threshold value, the identification result of the used bandwidth identification procedure and the excess identification procedure A band allocation method, further comprising: executing a selection procedure for selecting one of a predicted value of the first prediction procedure and a predicted value of the second prediction procedure based on an identification result. 11. The bandwidth allocation method according to claim 10,
In the selection procedure, the predicted value of the second prediction procedure is set regardless of the identification results of the used band identification procedure and the excess identification procedure during a predetermined period or a predetermined number of times after the prediction is started. A band allocation method characterized by selecting. 12. The bandwidth allocating method according to claim 10, wherein in the second prediction procedure, a guaranteed bandwidth of a line, a current used bandwidth, a current allocated bandwidth, and a predicted value of the first prediction procedure. A bandwidth allocation method characterized by selecting and outputting the maximum value among the predicted values. 13. The bandwidth allocation method according to claim 10 or 11, further comprising performing a guaranteed bandwidth identification procedure of comparing a previously predicted advance prediction value with a threshold of a guaranteed bandwidth of a line, and In the prediction procedure, if the preceding prediction value is less than or equal to the guaranteed bandwidth of the line,
The maximum value among the guaranteed bandwidth of the line, the currently used bandwidth, the current allocated bandwidth and the predicted value of the first prediction means is selected as the predicted value, and when the preceding predicted value exceeds the guaranteed band of the line. As a result of dividing the usable shared band by the total number of lines or the guaranteed band ratio of all lines, the current used band, the current allocated band, and the maximum value among the predicted values of the first prediction means are used as predicted values. A band allocation method characterized by selecting. 14. The bandwidth allocation method according to claim 10 or 11, wherein in the first prediction procedure, a predetermined coefficient is added to a declared value of a used bandwidth or an allocation request bandwidth observed at at least two points in the past. A bandwidth allocation method, characterized in that the sum or difference of the multiplied values is predicted as the current allocation request bandwidth. 15. The bandwidth allocation method according to claim 10 or 11, wherein in the first prediction procedure, a value that is linearly proportional to a used bandwidth observed at at least two past times or a declared allocation request bandwidth. And a value obtained by adding a value obtained by multiplying a previously declared allocation request bandwidth by a coefficient or a value obtained by multiplying a value obtained by multiplying a previously declared bandwidth by a coefficient with respect to the last-minute report bandwidth to a current predicted value. Allocation method. 16. The bandwidth allocation method according to claim 15,
A bandwidth allocation method, wherein the number of control cycles corresponding to the time required from the transmission of the declared value in the ONT to the reflection of the declared value in the bandwidth allocation in the OLT is applied as the coefficient. 17. The band allocating device according to claim 1, wherein the band predicting unit identifies whether or not there is band allocation of the band allocating unit for an observation period of a line observed by the traffic observing unit, A band allocating device, characterized in that if there is no band allocation, the band usage status of the corresponding line is considered to be other than 0 and the prediction is performed. 18. The bandwidth allocation method according to claim 9, wherein in the bandwidth prediction procedure, it is identified whether or not there is bandwidth allocation in the bandwidth allocation procedure for an observation period of a line observed in the traffic observation procedure. If there is no bandwidth allocation, the bandwidth allocation method is characterized by assuming that the usage status of the bandwidth of the corresponding line is other than 0 and performing the prediction. 19. The bandwidth allocating device according to claim 1, wherein the traffic observing means is based on a bandwidth usage condition observed before that while the bandwidth allocation of the bandwidth allocating means to the observed line is 0. A band allocation device, which outputs an observation value other than 0. 20. The bandwidth allocation method according to claim 9, wherein, in the traffic observation procedure, while the bandwidth allocation in the bandwidth allocation procedure for the observed line is 0, it is based on the usage status of the previously observed bandwidth. A bandwidth allocation method characterized by outputting an observation value other than 0. 21. The band allocating device according to claim 1, wherein the band allocating unit allocates a band of 1 cell or more in addition to the band corresponding to the prediction result during a period for detecting an increase in a band requested to be allocated. Alternatively, a band allocating device selectively allocating a smaller band of a result of adding a band of one cell or more to a band corresponding to the prediction result and a guaranteed band of the line. 22. The band allocating device according to claim 1, wherein the band predicting unit selects from the predicted band when the actual band allocated by the band allocating unit is less than a predicted band predicted as a band to be allocated to a line. It is considered that the amount of information corresponding to the band obtained by subtracting the actual allocated band is accumulated in the buffer in the allocated device as a queue, and corresponds to the difference between the actual allocated band and the predicted band from the band used by the line. A band allocation device, which predicts that a band will decrease. 23. The band allocating device according to claim 1, wherein the band predicting means outputs a second prediction band obtained by multiplying the first prediction band of the prediction result by a predetermined band prediction coefficient, and at the same time, When the second prediction band does not have a band increase of 1 cell or more per first period for detecting an increase in the allocation request band as compared with the first prediction band, the first prediction band is set to the above-mentioned A bandwidth allocating device, which outputs a result of adding one cell per first period as a prediction result. 24. The band allocating device according to claim 1, wherein the band predicting means multiplies a plurality of coefficients by which at least one of a used band and a reported value of the allocated band observed in the past is multiplied in order to predict a change in the band. While holding, whether the usage rate of the bandwidth observed by the traffic observation means is a predetermined value or more, whether the bandwidth observed by the traffic observation means is an increasing tendency or a decreasing tendency, the increasing tendency of the used bandwidth and According to a result of identifying at least one condition of whether the current allocated bandwidth exceeds a predetermined value and whether the declared value is 0, the plurality of coefficients are selected and the prediction is performed. Characteristic band allocation device. 25. The bandwidth allocating device according to claim 1, wherein the bandwidth request receiving unit normalizes a reported value for each line according to a reporting cycle, and the traffic observing unit responds an observed value to an observation cycle or a band allocation cycle. The bandwidth allocating device predicts an allocation request bandwidth of each line based on the normalized declared value and the observed value. 26. The bandwidth allocating device according to claim 2, wherein the second predicting unit divides the entire usable shared band or the entire usable shared band by the number of lines whose predicted value of the bandwidth allocation request is positive. Bandwidth, or the entire usable shared bandwidth, the guaranteed bandwidth ratio of the line for which the predicted value of the bandwidth request is positive, or the entire usable shared bandwidth, the guaranteed bandwidth ratio of all the lines,
A bandwidth allocation device, which selects and outputs the maximum value among the current allocated bandwidth, the currently used bandwidth, and the predicted value of the first prediction means. 27. The bandwidth allocating apparatus according to claim 5, wherein the second predicting means sets the same predicted value as when the preceding predicted value exceeds the guaranteed bandwidth of the line when the preceding predicted value is 0. A bandwidth allocating device characterized by selecting. 28. The band allocating apparatus according to claim 1, wherein when the predicted value of the band predicting means or the predicted value that is increased exceeds a maximum band allowed for a line, the predicted value is replaced with the maximum band. A band allocation device characterized by the above. 29. The band allocating device according to claim 1, wherein the band predicting means takes an average value of the used band when the present used band is within a range of a predetermined deviation from an average of past used bands. A band allocation device that predicts a band as described above. 30. The bandwidth allocating device according to claim 2, wherein the predicting means determines an actual allocated bandwidth allocated by the allocating means,
When the estimated bandwidth that the predicting means should allocate is less than the total, the value corresponding to the difference between the actual allocated bandwidth and the predicted allocated bandwidth is added to the declared value received by the requesting band receiving means. And a predicted value using at least one of a value obtained by subtracting a value corresponding to the difference between the actual allocated band and the predicted allocated band from the usage status of the line observed by the traffic observing means, A bandwidth allocation device, characterized in that the predicted value by the first predicting means and the predicted value by the second predicting means are selected by comparing the predicted value by the first predicting means. 31. The band allocating device according to claim 2, wherein the first predicting means uses a coefficient as a linear predictive function used by the linear predicting means from a value obtained by multiplying a coefficient used for a previous used band to a used band two times before. A bandwidth allocating device, wherein a value obtained by subtracting the multiplied value is multiplied by a bandwidth increase coefficient is used. 32. The bandwidth allocating apparatus according to claim 1, wherein the bandwidth predicting unit converts a calculated value obtained by multiplying the allocation request bandwidth by a predetermined coefficient into the number of cells per cycle of the allocation request declaration, and the allocated apparatus. A bandwidth allocating device, wherein the calculated value is replaced with the buffer length of the allocated device per cycle of allocation request declaration when the length of the buffer of the allocated device is exceeded. 33. The bandwidth allocating method according to claim 9, wherein in the bandwidth allocating procedure, in addition to the bandwidth corresponding to the prediction result, 1 is added during a period for detecting an increase in the allocation request bandwidth.
A band of cells or more is allocated, or a result of adding a band of one cell or more to a band corresponding to the prediction result and a guaranteed band of the line, whichever is smaller, is selectively allocated. Bandwidth allocation method. 34. The bandwidth allocation method according to claim 9, wherein, in the bandwidth prediction procedure, if the actual bandwidth allocated by the bandwidth allocation procedure is less than the predicted bandwidth predicted as the bandwidth to be allocated to the line, It is considered that the amount of information corresponding to the band obtained by subtracting the actual allocated band is accumulated in the buffer in the allocated device as a queue, and corresponds to the difference between the actual allocated band and the predicted band from the band used by the line. A band allocation method characterized by predicting that the band will decrease. 35. The band allocation method according to claim 9, wherein the band prediction procedure outputs a second prediction band obtained by multiplying the first prediction band of the prediction result by a predetermined band prediction coefficient, and further outputs the second prediction band. The first prediction band is used to detect an increase in the allocation request band as compared with the first prediction band.
When there is no band increase of 1 cell or more per period, the result of adding 1 cell per the first period to the first prediction band is output as a prediction result. 36. The bandwidth allocating method according to claim 9, wherein a plurality of coefficients by which at least one of a used bandwidth and a reported value of an allocated bandwidth observed in the past are multiplied to hold in order to predict a change in the bandwidth. In the procedure, whether the usage rate of the band observed in the traffic observation procedure is a predetermined value or more, whether the band observed in the traffic observation procedure is an increasing trend or a decreasing trend, the increasing trend of the used band and the current Of the plurality of coefficients are selected according to a result of identifying at least one condition of whether or not the allocated bandwidth exceeds a predetermined value and whether or not the declared value is 0. Band allocation method. 37. The bandwidth allocating method according to claim 9, wherein in the bandwidth request receiving procedure, the declared value for each line is normalized according to a declaration cycle, and in the traffic observation procedure, the observed value is responsive to an observation cycle or a bandwidth allocation cycle. The bandwidth allocation method is characterized in that the allocation request bandwidth of each line is predicted based on the normalized declared value and the observed value in the bandwidth prediction procedure. 38. The bandwidth allocation method according to claim 10, wherein in the second prediction procedure, the entire usable shared band or the entire usable shared band is divided by the number of lines whose predicted value of the bandwidth allocation request is positive. Bandwidth, or the entire available shared bandwidth, the guaranteed bandwidth ratio of the line for which the predicted value of the bandwidth request is positive, or the entire available shared bandwidth, the guaranteed bandwidth ratio of all the lines, the current allocated bandwidth, and the current usage A bandwidth allocation method, characterized in that the maximum value is selected and output from the bandwidth and the predicted value of the first prediction means. 39. The bandwidth allocation method according to claim 13, wherein in the second prediction procedure, when the preceding prediction value is 0, the same prediction value as when the preceding prediction value exceeds the guaranteed bandwidth of the line is used. A band allocation method characterized by selecting. 40. The bandwidth allocating method according to claim 9, wherein when the predicted value in the bandwidth prediction procedure or the increased predicted value exceeds a maximum bandwidth allowed for a line, the predicted value is set to the maximum bandwidth. A band allocation method characterized by replacement. 41. The bandwidth allocation method according to claim 9, wherein in the bandwidth prediction procedure, an average value of the used bandwidth is obtained when the current used bandwidth is within a predetermined deviation range from the average of the past used bandwidth. A bandwidth allocation method which predicts bandwidth as described above. 42. The bandwidth allocation method according to claim 10, wherein
In the prediction procedure, when the actual allocation bandwidth allocated in the allocation procedure is less than the allocation bandwidth predicted to be allocated in the prediction procedure, the difference between the actual allocation bandwidth and the predicted allocation bandwidth is A value obtained by adding a value corresponding to the declared value received in the requested bandwidth reception procedure, and a value corresponding to the difference between the actual allocated bandwidth and the predicted allocated bandwidth in the traffic observation procedure. The predicted value using at least one of the values subtracted from the usage situation and the predicted value by the first prediction procedure are compared to determine the predicted value by the first prediction procedure and the predicted value by the second prediction procedure. A band allocation method characterized by selecting. 43. The band allocation method according to claim 10, wherein the linear prediction function used in the linear prediction procedure is a value obtained by subtracting a value obtained by multiplying a previously used band by a coefficient from a value obtained by multiplying a previously used band by a coefficient. A bandwidth allocation method using a value obtained by multiplying by a bandwidth increase coefficient. 44. The bandwidth allocation method according to claim 9, wherein, in the bandwidth prediction procedure, a calculated value obtained by multiplying the allocation request bandwidth by a predetermined coefficient is converted into the number of cells per cycle in which the allocation request is declared, and the allocated apparatus. A bandwidth allocation method, wherein the calculated value is replaced with the buffer length of the allocated apparatus per cycle of the allocation request declaration when the length of the buffer of the allocation request exceeds. 45. The band allocating device according to claim 2, wherein the first predicting unit compares the predicted value with a declared value of the allocation request band and outputs the larger one as a linear predicted value. Allocation device. 46. The bandwidth allocating method according to claim 10, wherein in the first prediction procedure, the predicted value is compared with a declared value of the allocation request bandwidth, and the larger one is output as a linear predicted value. Allocation method.