JP2015073314A - Server device group and network system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve the problem that a load distribution device itself becomes a bottleneck and a single fault point on performance when a request from a terminal is distributed to a plurality of service provision devices by a load distribution device in a system where the request is transmitted from the terminal to a service provision device group and that a special function for appropriately selecting the service provision device is required on a terminal side when a load is distributed in the service provision device group without using the load distribution device.SOLUTION: A service provision system includes: a name solution device that has an address management table storing a plurality of IP addresses, selects the address from the address management table in response to a request from a terminal to make a response; and a plurality of service provision devices to which zero (0) or more IP address, of the IP addresses stored in the address management table, is assigned without duplication. Thus load is distributed by assigning the number of IP addresses corresponding to processing capacity of the service provision device.

Description

  The subject matter disclosed herein relates to a server system that provides a service via a network in response to a request from a terminal.

  A form of receiving a certain service by accessing a service providing apparatus on a network such as a Web service from a terminal has become common. The service providing system side needs to enhance the system as the number of service users increases.

  There are two approaches to enhancing the performance of a system: an approach that increases the performance of components (scale-up) and an approach that increases the number of components (scale-out). In the former approach, the performance cannot be expanded when the upper limit of the performance enhancement of the component is reached. Therefore, the latter approach is mainly required in a large-scale system.

  When scale-out of a service providing system that receives services from terminals, a special appliance called a load balancer is conventionally used in front of a plurality of service providing apparatuses. In this method, the terminal accesses the load balancer placed in the previous stage. The load distribution apparatus that has received an access from the terminal selects an appropriate service providing apparatus in consideration of the load of the service providing apparatus under its control, and transfers a request from the terminal. When the processing capacity is insufficient as a system, the scale-out type processing capacity can be improved by increasing the number of subordinate service providing apparatuses.

  In addition, the method disclosed in Patent Document 1 uses that name resolution is performed before the terminal accesses the service providing apparatus, and is appropriate in the same manner as the load distribution apparatus when responding to a name resolution request from the terminal. An appropriate service providing apparatus is selected and information necessary for access to the service providing apparatus as a distribution destination is returned.

  Further, in Non-Patent Document 1, when a storage service is provided in a scale-out type service providing device, one or more data is assigned to each device in order to adjust the amount of data in charge according to the capability of each service providing device. A method is disclosed in which a virtual node is allocated and the data storage amount handled by each service providing apparatus is adjusted by changing the allocation.

U.S. Pat.

Giuseppe DeCandia et al., “Dynamo: Amazon's Highly Available Key-Value Store”, in the Proceedings of the 21st ACM Symposium on Operating Systems Principles, Stevenson, WA, October 2007.

  The range in which the scale-out using the load balancer described above is effective is limited to the case where the processing amount in one service providing device is much larger than the processing amount required for distribution in the load balancer. In other words, when the processing at the service providing device is lightweight or when the performance of the service providing device alone is large, the distribution processing itself at the load balancer itself becomes a bottleneck, and the upper limit of the performance of the entire system is specified. Resulting in. In addition, when a failure occurs in the load balancer, access to all the service providing devices under the failure becomes impossible, and a large cost is required for handling the failure in the load balancer such as duplex.

  In load distribution using name resolution disclosed in Patent Document 1, access to a service providing apparatus after name resolution is possible even when a failure occurs in the name resolution apparatus. However, in order to perform name resolution appropriately, it is necessary to collect information on the route between the terminal and the service providing apparatus and hold the state. For this reason, when a failure occurs in the part holding this information, subsequent name resolution becomes difficult and access to the entire system becomes difficult.

  Further, in the load balancing method disclosed in Non-Patent Document 1, the data storage amount handled by each service providing apparatus can be adjusted by the number of virtual nodes, but the terminal side is aware of where the data is stored. In order to access any one of the service providing devices, the process transfer is actually performed again to access the data. For this reason, the transfer of processing can be a bottleneck, similar to the load distribution by the load distribution method mentioned at the beginning. In order to avoid this, it is possible to embed a processing unit for determining the data storage destination on the terminal side. However, a storage destination determination program is distributed to all terminals, and it is necessary depending on the situation of the service providing device group. Problems remain, such as the need to notify all terminals of information.

  What is disclosed in this specification is that a load distribution function (Feature) can be a bottleneck or a failure point, and that a load distribution function is required on the terminal side for load distribution. And a service providing system for executing the method.

  The disclosed service providing system requests the name resolution device to resolve the identifier of the service providing device when accessing the service providing device from the terminal, and requests the IP address obtained as a result of the name resolution. It is premised on a system that performs the above, and has the following features.

  A name resolution device that has an address group management table that stores a plurality of IP addresses, performs a process of selecting an address from the address group management table and responding to a request from the terminal, and is stored in the address group management table A plurality of service providing devices to which 0 or more IP addresses are assigned without duplication are provided, and load distribution is realized by assigning a number of IP addresses corresponding to the processing capabilities of the service providing devices to each service providing device.

  Each service providing device that accepts a request from a terminal measures the load of another service providing device that is a monitoring destination, and if the load of the monitoring destination is larger than the load of the own service providing device that is the monitoring source, the load is large By changing the load difference by moving the IP address assigned to the monitoring destination device to the own device having a small load, the load variation is dealt with.

  Also, in the name resolution device, addresses in the address group management table are divided into two or more groups and managed, and a function that selects one or more groups from the date and time is a response target for a request from the terminal By having the function of switching the address to be changed for each date and time, it is possible to secure an unused address necessary for the load transfer with a high probability.

A more specific aspect of the disclosed service providing system is a service providing system that includes a plurality of service providing devices and one or more name resolution devices, and one of the service providing devices responds to a service request from a terminal. There,
Name resolution device
managing an address group consisting of n (k is n or more) IP addresses assigned to a service providing device group consisting of k service providing devices;
In response to a name resolution request from a terminal, a name resolution request processing unit that responds by selecting any one IP address from a group of addresses to be managed, and
Each of the plurality of service providing devices
Of the n IP addresses managed by the name resolution device, m (m is 0 or more) are assigned to the service providing device without overlapping with other service providing devices,
A terminal request processing unit that provides a service in response to a service request from a terminal based on a response to the name resolution request;
A load distribution function unit that performs load distribution with other service providing devices,
As the monitoring destination service providing device, the load is monitored from other monitoring source service providing devices,
The load distribution function unit of the monitoring source service providing apparatus moves the assignment of the IP address from the monitoring destination service providing apparatus to the own service providing apparatus when the load of the monitoring destination service providing apparatus is high.

In the disclosed service providing system,
The load distribution function part of the monitoring source service providing device moves the IP address assignment.
If the load of the monitoring destination service providing device is higher than the load of the own service providing device by a certain value or more, send an address takeover request to the monitoring destination service providing device,
When an address takeover response including an IP address that is an assignment transfer target is received from the monitoring destination service providing apparatus, the included IP address may be registered as an IP address assigned to the own apparatus.

  According to the above aspect, it becomes possible to distribute the load to the service providing apparatus group without having a single point of failure and a bottleneck apparatus such as a load distribution apparatus between the terminal and the service providing apparatus. Will improve.

  In addition, the name resolution device required for name resolution does not retain dynamic data such as load information of each service providing device, but retains static data such as IP address groups as states. It is possible to easily recover from a failure by storing such data in stable storage.

  In addition, even if multiple name resolution devices are deployed, it is only necessary to provide static data to each name resolution device, so complicated procedures such as maintaining consistency with dynamic update of shared data Is no longer necessary.

  In addition, by adjusting the load on each service providing device by assigning the number of IP addresses according to the processing capacity, it is possible to guide the request to the appropriate server using the communication mechanism with the standard IP of the terminal Even when the service providing apparatus having a certain IP address is changed by the transfer of the IP address, the new request is appropriately redirected by the routing function that the IP network has as a standard.

  Further, by changing the address group to be returned in the name resolution device in time series, it becomes possible to secure the address on the service providing device side where no session is established between the terminal and the service providing device. This makes it possible to select an unused address when assigning an IP address in the service providing apparatus. When some kind of session management is required for a service provider such as TCP / IP protocol, complicated processing is required to transfer the IP address used by the session during session establishment. It is desirable to transfer the IP address. With this method, it is possible to improve the possibility of securing an unused IP address and reduce the processing cost for load equalization.

  According to the disclosure, it is possible to realize a service providing system that has a load distribution function but is less likely to become a bottleneck or a failure point, and that does not require a load distribution function on the terminal side.

The figure which illustrates the whole structure of a service provision system Diagram illustrating the functional configuration of the DNS server The figure which illustrates the structure of an address group management table Example of DNS server processing flowchart The figure which illustrates the functional structure of a service provision apparatus The figure which illustrates the structure of a service provision apparatus list Figure illustrating the configuration of the assigned IP address list The figure which illustrates the structure of a monitoring destination service provision apparatus list The figure which illustrates the structure of the monitoring origin service provision apparatus list Flowchart illustrating processing for establishing relationship between monitoring of service providing apparatus and monitored object Flowchart illustrating processing of load distribution function unit of service providing apparatus Flowchart illustrating processing of load distribution function unit of service providing apparatus The figure which illustrates the sequence structure of a service provision system The figure which illustrates the sequence structure of a service provision system Fig. 3 is a diagram illustrating an example of a service providing device list configuration in the case of N + M configuration Flowchart illustrating the processing of the DNS server when switching the IP group that responds according to the time of day The figure which illustrates the whole structure of a service provision system in case two or more DNS servers exist The figure which illustrates the hardware constitutions of a DNS server and a service provision apparatus

(Example 1)
FIG. 1 is an overall configuration diagram of the first embodiment.

  The service providing device 104 and the DNS server 103 are connected to a router / switch 102 constituting an external network, and the router / switch 102 is connected to the terminal 101 via a mobile core network / Internet or the like. Note that the connection topology described here is an example, and this embodiment can be applied if the terminal 101, the DNS server 102, and the service providing apparatus 104 are connected to each other with IP reachability.

  FIG. 2 shows a functional configuration of the DNS server 103 which is an example of the name resolution device of the present embodiment. The DNS server 103 converts a domain name into an IP address as an example of name resolution. The DNS server 103 is connected to the router / switch 102 via the packet transmission / reception unit 204. A name resolution request from the terminal 101 is processed by a name resolution request processing unit (hereinafter referred to as DNS request processing unit) 202. Also, the management function unit 203 performs processing such as registration and deletion of addresses in the address group management table 201.

  FIG. 18 shows an example of the hardware configuration of the DNS server 103 and the service providing apparatus 104 whose functional configuration will be described later.

  These devices include a main storage device 1802, a central processing unit 1803, a secondary storage device 1805 such as a hard disk drive, an input / output device 1807 such as a display and a keyboard, a communication device 1806 for connecting to the router switch 102, and those This can be realized by a general computer 1801 provided with an internal communication line 1804 such as a bus for connecting the devices.

  For example, the address group management table 201 in the DNS server is placed in the main storage device 1802 and in the secondary storage device 1805 in order to store data permanently.

  Each process described below is realized by a program stored in the secondary storage device 1805 being copied to the main storage device 1802 at the time of execution and then interpreted and executed by the central processing unit 1803. The transmission / reception processing is performed via the communication device 1806.

  FIG. 3 is an example of the address group management table 201 that the DNS server 103 has. In the address group management table 201 shown here, IPv6 addresses are registered in a plurality of entries in subnet units as addresses for names to be resolved (domain names in this embodiment). In FIG. 3, the address group is represented by an IPv6 address, but the address group may be represented by an IPv4 address. In each entry of the address group management table 201, an address assigned to all the service providing apparatuses 104 is registered.

  Next, how the DNS server 103 responds to a request from the terminal 101 using the address group management table 201 will be described.

  FIG. 4 shows a processing flow when the DNS server 103 receives a request from the terminal 101. In step 401, the name resolution request requested by the terminal 101 is checked for which name. For example, if the address group management table having the form shown in FIG. 3 is provided and if name resolution of www.example.com is requested, the process proceeds to step 402.

  In step 402, one of the addresses included in the address group management table is selected and returned. At this time, the method of selecting an address to be returned from the address group is not limited as long as the probabilities of returning each address are substantially equal. As an example, there are a method of selecting one by one in order from the top, and a method of selecting from a group of all addresses using uniform random numbers.

  In step 401, if the request is for a name that is not managed as the address group management table 201, the request is transferred to another DNS server, and the result obtained by inquiring to the other DNA server of the transfer destination is returned. (Step 403). This is a countermeasure when a name resolution request for a name unrelated to the service providing apparatus 104 is received. If a request for a name not related to this service providing system is not supported, a name resolution error may simply be returned.

  Next, the service providing apparatus 104 will be described.

  FIG. 5 shows a configuration diagram of the service providing apparatus 104. The service providing apparatus 104 is connected to the router / switch 102 via the packet transmitting / receiving unit 502. Processing of services provided by the service providing apparatus 104 to the terminal 101 is processed by the terminal request processing unit 501. Further, load distribution is processed by the load distribution function unit 503. The load distribution function unit 503 has four tables, which are referred to as a service providing device list 601, a monitoring destination service providing device list 801, a monitoring source service providing device list 901, and a responsible IP address list 701, respectively.

  FIG. 6 shows an example of the service providing apparatus list 601.

  The service providing apparatus list 601 manages the identifiers of n service providing apparatuses 104 that distribute loads. The value of the service providing apparatus list 601 is set by the administrator when the service providing apparatus 104 is started, and is updated by the load distribution function unit when the service providing apparatus 104 is added to or deleted from the system.

  FIG. 7 shows an example of the assigned IP address list 701.

  The assigned IP address list 701 stores m IP addresses assigned to the own service providing apparatus 104 having this list. The assigned IP address list 701 is set by the administrator when the service providing apparatus 104 is activated, and is updated by the load distribution function unit when the IP address is moved to or transferred from another node, that is, transfer and transfer.

  FIG. 8 shows an example of the monitoring destination service providing apparatus list 801.

  The monitoring destination service providing apparatus list 801 indicates information of other service providing apparatuses 104 for which the service providing apparatus 104 having this list is to monitor the load.

  Here, the service providing apparatus 104 targets the service providing apparatus 104 having the identifier “serv2” as the monitoring target, and the IP address to be assigned when the monitored service providing apparatus becomes heavily loaded or a failure occurs. The date and time when the assignment of the IP address is moved to the monitored service providing apparatus 104 by the candidate list and the previous load adjustment is recorded as the load adjustment date and time. The load adjustment date and time is referred to in order to prevent the address assignment movement due to the load from occurring continuously in a short period of time.

  In this example, the movement of IP address assignment and the movement of load are not synchronized. Therefore, if the movement of IP address assignment is always performed in a state where there is a difference in load, there will be more IP addresses than necessary for the movement of the load. Because it moves, IP address movement in a short period is restricted by referring to the load adjustment date and time.

  FIG. 9 shows an example of the monitoring source service providing apparatus list 901.

  The monitoring source service providing apparatus list 901 records other service providing apparatuses 104 that are monitoring the service providing apparatus 104 having this list.

  In FIG. 9, the own service providing apparatus 104 is monitored by another service providing apparatus 104 having the identifier “serv4”, and when the own service providing apparatus 104 becomes heavily loaded or when a failure occurs. , Address candidates to be transferred to the service providing apparatus 104 having the identifier “serv4” are stored in the takeover IP list.

  When the service providing apparatus 104 is monitored from a single service providing apparatus 104, the contents of the takeover IP list are the same as the contents of the assigned IP address list 701 of the service providing apparatus 104. When monitored from a plurality of service providing apparatuses, the addresses included in the assigned IP address list 701 are shared.

  Further, when the service providing apparatus 104 updates the assigned IP address list 701 by moving the IP address (transferred from another or transferred to another), the takeover IP list is also updated.

  A method for constructing the relationship between monitoring and monitored between all the service providing apparatuses 104 at the time of initial setting will be described below. FIG. 10 shows a construction procedure executed by the load distribution function unit 503.

  In step 1001, some form of ordering is performed on the identifiers of all service providing apparatuses 104 included in the service providing apparatus list 601. Since it is only necessary to define the order relationship for all identifiers, methods such as calculating a hash value and determining the order based on the magnitude of the number, or determining the order in a dictionary order, and the like are conceivable. Here, as an example, calculation is performed using a hash value.

  In step 1002, the list of service providing apparatuses that can be ordered is sorted in ascending order. Next, in step 1003, the identifier of the own device is searched from the service providing device list 601.

  Finally, in step 1004, the service providing apparatus 104 next to the entry of the own apparatus is set as the monitoring destination, and the previous service providing apparatus 104 is set as the monitoring source. Further, it is assumed that the service providing device list 601 has a ring structure, and the monitoring destination of the device at the end of the list is the device at the top of the list, and the monitoring source of the device at the top of the list is the device at the end of the list. By this processing, all the service providing devices 104 can be handled as being virtually connected with one ring topology.

  Next, a processing flow of the load distribution function unit 503 of one service providing apparatus 104 is shown in FIG.

  In step 1101, when the service providing apparatus 104 is activated, the service providing apparatus list 601 and the assigned IP address list 701 are set as initial settings. As an example of setting of the assigned IP address list 701, assuming that the performance of the service providing apparatus 104 is uniform in the initial state, a method of assigning the same number of addresses to all the service providing apparatuses 104, or each service providing apparatus 104 There is a method of assigning a large number of addresses to the service providing apparatus 104 with high performance by weighting according to the performance difference.

  Next, in step 1102, a monitoring destination service providing apparatus is determined according to the processing procedure shown in FIG. 10, and an apparatus information acquisition request message is transmitted. In step 1103, a message reception is waited. In the following, processing is performed according to the content of the received message.

  When the device information acquisition request message is received, the entry of the transmission source service providing device is added to the monitoring source service providing device list 901, and the takeover IP address is reconfigured for the entire monitoring source. Then, a device information message including the monitoring source service providing device list 901 is transmitted to all the monitoring source service providing devices (step 1104).

  When the device information message is received, the monitoring destination service providing device list 801 is updated with the device information included in the received message (step 1105).

  When the load information request message is received, the load information of the current device is transmitted as a load information message (step 1106). The load information includes CPU load, transmission / reception traffic volume per unit time, I / O load, memory usage, and the like. In a process in which access to the disk is a bottleneck, the service providing apparatus 104 selects the service according to the type of service provided to the terminal, such as using the I / O load as load information.

  When an address takeover request message is received from another service providing apparatus 104 described in the monitoring source service providing apparatus list 901, one address that is not currently used is selected from the assigned IP address list 701. When all addresses are used, such as when there is a heavy load, select one at random from the addresses currently in use.

  Then, the IP address selected from the assigned IP address list 701 is deleted, the IP address is unassigned from the own device, and an address takeover response message including the IP address is notified to the transmission source device (step 1107) . In addition, the IP address is deleted from the takeover IP address list of the monitoring source service providing apparatus list 901.

  As can be seen from the above, in this specification, an address takeover request means a request to move an address assignment destination from a service providing apparatus as a request transmission destination to a service providing apparatus as a request transmission source.

  When an address takeover request message is transmitted from the own device and an address takeover response message is received from another service providing device 104 that is a transmission partner, the address included in the received IP address list 701 is registered, and the own device This address is assigned to. Then, a gratuitous ARP is transmitted from the packet transmitting / receiving unit in order to notify other devices in the system that a new address has been assigned to the own device. In addition, the takeover IP address list in the monitoring source service providing apparatus list 901 is updated, and a device information message including the newly registered IP address information is transmitted to the monitoring source service providing apparatus monitoring the own apparatus ( Step 1108).

  When the service providing device addition request message or the deletion request message is received, the service providing device list 601 is updated. Then, the relationship between the monitoring source and the monitoring destination is recalculated according to the flow shown in FIG. The monitoring source service providing apparatus list 901 is updated. Then, a device information acquisition request message is transmitted to the service providing device 104 that has newly become a monitoring destination (step 1109).

  When the service requesting apparatus subscription request message is received, the source service providing apparatus is added to the service providing apparatus list 601. Then, in the service requesting device 104 that is the transmission source of the subscription request message, an addition request message of all the service providing devices 104 excluding the source service providing device is sent to the other service providing devices. An addition request message is transmitted (step 1110).

  When the address addition request message is received, the address included in the received message is added to the assigned IP address list 701, and the address is assigned to the own apparatus. Further, the address is added to the takeover IP address list of the monitoring source service providing apparatus list 901, and a device information message including the apparatus information is transmitted to the monitoring source service providing apparatus 104 that has been changed (step 1111).

  When an address deletion request message is received, the corresponding address is deleted from the assigned IP address list 701. Then, the address is deleted from the takeover IP address list of the monitoring source service providing apparatus list 901, and a device information message including the device information is transmitted to the monitoring source service providing apparatus 104 that has been changed (step 1112).

  The load distribution function unit 503 of the service providing apparatus 104 performs processing for monitoring the operation information of other service providing apparatuses 104 in parallel with the procedure illustrated in FIG. The procedure of this process is shown in FIG.

  When monitoring is started, the service providing apparatus 104 repeatedly (for example, periodically) transmits a load information request message to the monitoring destination service providing apparatus 104 to be monitored (step 1201). If the load information message is received within a certain time, the monitored service providing apparatus 104 recognizes that no failure has occurred.

  If a certain period has elapsed from the load adjustment date / time in the monitoring destination service providing apparatus list 801 before reception, the load difference from the own apparatus is calculated (step 1202). If the load of the monitoring destination service providing apparatus is higher than the load of the own apparatus by a certain value or more, an address takeover request is transmitted to the monitoring destination service providing apparatus 104 (1203). If the load difference is less than a certain value, or if a certain period has not elapsed since the previous load adjustment, a load information request is transmitted again after a certain period of suspension.

  If the load information message cannot be received within a certain period of time, it is determined that a failure has occurred in the monitored service providing apparatus, and the service is provided to the other n service providing apparatuses 104 that distribute the load. A deletion request message for requesting deletion of the service providing apparatus is transmitted from the providing apparatus list 601.

  Then, a takeover IP address list inherited from the service providing apparatus is extracted from the entry of the apparatus in the monitoring destination service providing apparatus list 801 and added to the assigned IP address list 701 (step 1205).

  Next, the device is deleted from the service providing device list 601, and the monitoring destination service providing device 104 is determined by the procedure shown in FIG. Next, a device information acquisition request message is transmitted to the monitoring destination service providing device 104 (step 1206). Then, for example, after a certain period of rest, the procedure from step 1201 is repeated.

  FIG. 13 and FIG. 14 show an example of a series of sequences generated by the procedure described so far.

  First, initial setting and request processing from the terminal will be described.

  In step 1301 of FIG. 13, the administrator sets the service providing device list 601 and the assigned IP address list 701 for each service providing device 104, and builds the relationship between the monitoring source and the monitoring destination according to the flow of FIG. ing.

  Then, each service providing apparatus 104 serving as a monitoring source sends a device information acquisition request to the service providing apparatus 104 serving as a monitoring destination (step 1302).

  Receiving the device information acquisition request, the service providing device 104 returns a device information message according to step 1104 in FIG. 11 (step 1303). Thus, the initial setting of each service providing apparatus 104 is completed.

  The terminal 101 sends a name resolution request to the DNS server 103 to receive the service (step 1304).

  The DNS server 103 returns an address to the terminal 101 according to the flow of FIG. 4 (step 1305).

Terminal 101 sends a processing request to the IP address indicated in the reply (step 1306),
The service providing apparatus 104-3 in charge of the address responds to the request (step 1307).

  Next, the load equalization process when the load is unbalanced between the service providing apparatuses 104-1 and 104-2 will be described.

  The service providing apparatus 104-1 repeatedly transmits, for example, a load information request to the service providing apparatus 104-2 as a monitoring destination periodically in Step 1201 of the flow shown in FIG. 12 (Step 1308).

  Receiving the load information request, service providing apparatus 104-2 returns the load information of the current apparatus in accordance with step 1106 of FIG. 11 (step 1309).

  Here, it is assumed that the service providing apparatus 104-1 that has received the load information determines that the load difference is greater than or equal to a certain value in step 1202, and transmits an address takeover request to the service providing apparatus 104-2 according to step 1203. (Step 1310).

  Receiving the address takeover request, service providing apparatus 104-2 selects a takeover address as shown in step 1107, and returns an address takeover response to service providing apparatus 104-1 (step 1311).

  The service providing apparatus 104-1 that has received the address takeover response updates the assigned address list 701 of the own apparatus and transmits the apparatus information to the service providing apparatus 104-4 that is the monitoring source of the service providing apparatus 104-1 To do. The service providing apparatus 104-4 updates the monitoring destination service providing apparatus list 801 according to step 1105 (step 1312).

  Next, a case where a failure occurs in the monitoring destination service providing apparatus 104-2 will be described.

  The service providing apparatus 104-1 periodically transmits a load information request to the monitoring destination service providing apparatus 104-2 according to the flow in FIG. 12 (step 1313).

  When the reception of load information times out for a predetermined R times or more, the monitoring destination apparatus determines that a failure has occurred according to step 1204. Then, according to the service providing apparatus list 601, a request for deleting the service providing apparatus 104-2 is transmitted to all the service providing apparatuses 104 (step 1314).

  The service providing apparatuses 104-3 and 104-4 that have received the deletion request update the service providing apparatus list 601 of each apparatus in accordance with step 1109. Then, the relationship between the monitoring source and the monitoring destination is reestablished according to the flow of FIG. 10, and the monitoring destination service providing device list 801 and the monitoring source service providing device list 901 are updated.

  Here, since the monitoring destination service providing apparatus 104-1 has been updated to the service providing apparatus 104-3, the monitoring destination service providing apparatus 104-1 transmits an information acquisition request to the service providing apparatus 104-3 that has newly become the monitoring destination. (Step 1315).

  The service providing apparatus 104-3 that has received the information acquisition request returns apparatus information in accordance with step 1104 (step 1316).

  Next, FIG. 14 shows a case where the service providing apparatus 104-2 is recovered from the failure and added to the system again.

  The administrator enters the information of the service providing apparatus 104-2 in the service providing apparatus list 601 of the service providing apparatus 104-2. The service providing apparatus 104-2 arbitrarily selects the service providing apparatus 104 that is providing the service (here, the service providing apparatus 104-4), and transmits a subscription request (step 1402).

  The service providing apparatus 104-4 that has received the subscription request transmits an additional request including the contents of the service providing apparatus list 601 to the source service providing apparatus 104-2 in accordance with Step 1110 of FIG. Further, an addition request is transmitted to service providing apparatus 104-2 to service providing apparatuses 104-1 and 104-3 other than the transmission source (step 1403).

  Each service providing device 104 that has received the addition request updates the service providing device list 601 according to step 1109, reconstructs the relationship between the monitoring source and the monitoring destination according to the flow of FIG. 10, and the monitoring destination is changed. Transmits an information acquisition request to the monitoring destination service providing apparatus (step 1404).

  Receiving the information acquisition request, the service providing apparatus 104 returns apparatus information according to step 1104 (step 1405).

  When a failure occurs in the DNS server 103 itself, it is necessary for some external server to detect it and restart the DNS server 103 or start another DNS server 103. At this time, if the DNS server manages the load information and operation information of the service provider group, the information must be saved in some form of stable storage during operation and restored on restart. . This greatly increases the cost required to improve fault tolerance. On the other hand, in this embodiment, the DNS server itself has the address group management table 201 that is invariant information regardless of the operation status of the service providing apparatus 104 group, and the load information of the service providing apparatus group. Since the DNS server 103 can always be started with the same setting. For this reason, the cost of dealing with a failure of the DNS server 103 can be kept very low.

(Example 2)
In the first embodiment, all service providing apparatuses 104 are managed in an equal relationship, and all normal service providing apparatuses 104 accept requests from the terminal 101.

  In the second embodiment, of the N + M service providing apparatuses 104, N service providing apparatuses 104 process a request from the terminal 101, and the M service providing apparatuses 104 are in a standby state. An example is shown. Since there are many common points between the configuration and the processing procedure, the difference will be described here.

  FIG. 15 shows a service providing apparatus list 1501 in this embodiment. Compared to the service providing apparatus list 601 of the first embodiment, a column indicating whether each service providing apparatus 104 is in an operating state (ACT) or in a standby state (SBY) is added.

  In the state set in this way, the monitoring in the first embodiment is performed between the service providing apparatuses 104 in the operating state. When a failure occurs in the monitored service providing apparatus 104, select one service providing apparatus 104 whose status is SBY from the service providing apparatus list 1501, change the server status to ACT, The information is transmitted to the service providing apparatus 104 as a message. Further, the takeover IP address of the service providing apparatus 104 in which the failure has occurred is transmitted to the new service providing apparatus as an address takeover response message. As a result, load distribution can be realized in the same manner even in the N + M configuration.

(Example 3)
In Examples 1 and 2, an address that is not used at the time of IP address takeover is selected. However, when one session lasts for a long time, there is a possibility that all addresses are in use and an address that can be safely taken over cannot be secured.

  In this embodiment, the address group returned by the DNS server 103 is divided into a plurality of lower groups, and the reply target group is switched according to the date and time, so that an IP address that is not in use can be secured when taking over the IP address. How to do it.

  FIG. 16 shows the operation of the DNS server 103 in this embodiment. Compared to FIG. 4 which is a corresponding procedure in the first and second embodiments, the step 1602 is different.

  In step 1602, one or more address groups are selected using a predetermined function from the current date and time to the address group number, and an address is selected from the group and a reply is made.

  This procedure creates a group of addresses that are not distributed at a certain date and time. Assuming that the continuation of the session between the terminal 101 and the service providing apparatus 104 does not continue for a certain period of time, the probability that there is a terminal 101 using the address after a certain period of time has elapsed after the address has been distributed is Low enough. This makes it possible to secure an unused address at the time of address takeover.

(Example 4)
In the first, second, and third embodiments, the DNS server 103 is described as one. However, the DNS server 103 according to the present embodiment has a static address group management table 201 and a function from the date and time to the address group number as a state. There is no information to change.

  For this reason, even when a plurality of DNS servers 103 are placed in the system and these DNS servers are switched and used on the terminal side, state synchronization between the DNS servers 103 is not necessary. FIG. 17 shows an overall configuration diagram of the fourth embodiment. With this configuration, the performance scalability and fault tolerance of the DNS server 103 can be improved.

101: Terminal, 102: Router / switch, 103: DNS server, 104: Service providing device, 201: Address group management table, 202: DNS request processing unit, 203: Management function unit, 204: Packet transmission / reception unit, 501: Terminal Request processing unit, 502: packet transmission / reception unit, 503: load distribution function unit, 601: service providing device list, 701: assigned IP address list, 801: monitored service providing device list, 901: monitoring source service providing device list, 1501 : Service providing device list.

Claims (7)

In a service providing system comprising a plurality of service providing devices and one or more name resolution devices, and any of the service providing devices responds to a service request from a terminal,
The name resolution device
managing an address group consisting of n (k is n or more) IP addresses assigned to a service providing device group consisting of k service providing devices;
A name resolution request processing unit that responds by selecting any one IP address from the address group to be managed in response to a name resolution request from the terminal;
Each of the plurality of service providing devices includes:
M (m is 0 or more) of the n IP addresses managed by the name resolution device are allocated to the service providing device without overlapping with the other service providing devices,
A terminal request processing unit for providing a service to a service request from the terminal based on the response to the name resolution request;
A load distribution function unit that performs load distribution with other service providing devices,
As a monitoring destination service providing device, the load is monitored from another monitoring source service providing device, and the load distribution function unit of the monitoring source service providing device is configured to IP when the load on the monitoring destination service providing device is high. A service providing system, wherein address allocation is moved from the monitoring destination service providing apparatus to the own service providing apparatus. The service providing system according to claim 1,
The load distribution function unit of the monitoring source service providing apparatus, as a process of moving the IP address assignment,
When the load of the monitoring destination service providing apparatus is higher than the load of the own service providing apparatus by a certain value or more, an address takeover request is transmitted to the monitoring destination service providing apparatus,
Upon receiving an address takeover response including an IP address that is an assignment movement target from the monitoring destination service providing apparatus, the included IP address is registered as an assigned IP address to the own apparatus. . The service providing system according to claim 2,
The load distribution function unit of the monitoring destination service providing device is:
In response to the address takeover request, select one of the assigned IP addresses assigned to the device itself,
A service providing system, wherein the selected IP address is deleted from assignment to the own device and notified to the monitoring source service providing device by transmitting the address takeover response. In the service provision system as described in any one of Claim 1 to 3,
The IP address to be moved of the allocation is an unused address,
The name resolution device
Divide the address group to be managed into two or more subordinate groups,
The service providing system, wherein the lower group to which an address to be selected in response to a name resolution request from the terminal belongs is selected according to date and time according to date and time. In the service provision system as described in any one of Claim 1 to 4,
The load distribution function unit of the monitoring source service providing device is:
Managing the previous date and time when the IP address assignment was moved from the monitoring destination service providing apparatus to the monitoring source service providing apparatus,
A service providing system, wherein when the difference between the current date and time and the previous date and time is greater than or equal to a predetermined value, the IP address assignment is moved. In the service provision system as described in any one of Claim 1 to 5,
The load distribution function unit of the monitoring source service providing device is:
Monitor other service-destination service provider devices in operation,
Manage IP address candidates to be transferred from the monitored service providing apparatus to the own apparatus,
If the monitoring determines that a failure has occurred in the monitoring destination service providing apparatus, the IP address candidate to be managed is registered as an IP address assigned to the own apparatus to be managed. Offer system. The service providing system according to claim 6,
The k service providing devices are divided into two groups of active and standby,
The load distribution function unit of the monitoring source service providing apparatus in operation monitors other monitoring destination service providing apparatuses in operation,
When confirming a failure of the monitoring destination service providing device, select one of the standby service providing devices as a new monitoring destination service providing device,
A service providing system, wherein an address takeover response including the IP address assigned to the monitoring destination service providing apparatus in which a failure has occurred is transmitted to a new monitoring destination service providing apparatus.

Images