JP5440200B2 - Relay device and bandwidth control method - Google Patents

Description

  The present invention relates to a bandwidth control technique.

A wide area network (WAN) is a communication system that connects remote locations with a network.
It is called a communication system. Hereinafter, such a network connecting remote bases is referred to as a WAN, a WAN line, or the like. In such a WAN communication system, there is a network device that speeds up or optimizes a WAN line. The speed-up or optimization of the WAN line is realized by each network device that connects each base network to the WAN, such as a VPN (Virtual Private Network) router, a WAN speed-up device, and a WAN optimization device.

  In such a network device, bandwidth control of the WAN line is performed. For example, when link congestion is detected due to buffer full or the like in the network device, priority band control is performed such that a band is preferentially allocated to a communication requiring high speed over other communication.

JP 2008-5078 A JP 2007-243300 A

Generally, a WAN as described above is often a black box network that is not managed by an edge node manufacturer or the like, such as the Internet. In such a black box network, congestion may occur in the network, and a band necessary for communication between bases may not be obtained. In order to avoid such a problem, it is necessary that RSVP (Resource Reservation Protocol), priority control, and the like are implemented in each relay device such as a router included in the black box network. However, since each user cannot manage the inside of the black box network, it often cannot depend on the mounting function of each relay device included in the black box network.

  As a result, the conventional edge node as described above cannot detect congestion in a link that is not directly connected to itself among each link that connects a plurality of relay devices included in a specific network such as a black box network. . As described above, the conventional edge node has a problem in that it cannot execute the bandwidth control to cope with the congestion in the specific network.

  In view of such a problem, an object according to one aspect of the present invention is to provide a bandwidth control technique that avoids congestion in a specific network.

  Each aspect of the present invention employs the following configurations in order to solve the above-described problems.

A 1st aspect is related with the relay apparatus which relays a packet from a 1st network to a 2nd network. The relay device according to the first aspect generates congestion among a plurality of communication units specified by at least a destination address and a transmission source address set in each packet relayed from the first network to the second network. The at least one specific link in the second network is shared between the congestion detection unit that detects the communication unit and the communication unit in which congestion is detected detected by the congestion detection unit among the plurality of communication units. A specifying unit for specifying another communication unit, and a band for limiting a transmission band to the second network of the other communication unit specified by the specifying unit and the communication unit in which congestion occurs among the plurality of communication units. Control means.

  Another aspect of the present invention may be a method for realizing the above configuration, a program for causing a computer to realize the above configuration, or a computer that records such a program. It may be a recording medium.

  According to each aspect described above, it is possible to provide a bandwidth control technique that avoids congestion in a specific network.

The figure which shows schematic structure of a WAN communication system. 1 is a block diagram showing a schematic configuration of an edge node 10 in Embodiment 1. FIG. FIG. 6 is a diagram illustrating an example of a session management table in the first embodiment. The figure which shows the example of the route information table. The figure which shows the example of the link set table 27. FIG. 6 is a flowchart illustrating an operation example of the bandwidth control apparatus according to the first embodiment. 6 is a flowchart illustrating an operation example of the bandwidth control apparatus according to the first embodiment. FIG. 5 is a block diagram illustrating a schematic configuration of an edge node 10 according to a second embodiment. FIG. 10 is a diagram illustrating an example of a session management table in the second embodiment. The image figure which shows a shared session group. The figure which shows the example of the shared session table 73. FIG. 10 is a flowchart illustrating an operation example of inspection of a shared session group in the second embodiment. 9 is a flowchart illustrating an operation example of transmission band control in the second embodiment.

  Hereinafter, a specific example of the bandwidth control device as one embodiment will be described. The bandwidth control apparatus as an embodiment acts as an edge node that connects bases via a WAN. Hereinafter, the bandwidth control device in this embodiment is also referred to as an edge node.

This embodiment does not limit the meaning of WAN, and WAN may be a network including a plurality of relay devices such as routers. According to the edge node in the present embodiment, even Ru Oh applicable as the WAN was unmanaged network to edge node provider. Each example given below is an example, and the present embodiment is not limited to the configuration of each example below.

  Hereinafter, a first example of the bandwidth control device (edge node) as the embodiment will be described.

[System configuration]
FIG. 1 is a diagram showing a schematic configuration of a WAN communication system. The WAN communication system includes base networks 2, 3 and 4 that are respectively connected via the WAN1. The WAN 1 is a public network such as the Internet, for example. Each base network is, for example, a LAN (Local Area Network) provided in each of the base A, the base B, and the base C.

  The edge node 10 according to the first embodiment is arranged at the boundary between the base network 2 and the WAN 1 and connects the base network 2 to the WAN 1. The edge node 10 in the first embodiment controls the bandwidth of communication transmitted from the base network 2 to the WAN 1 and does not particularly limit communication transmitted from the WAN 1 to each base. Therefore, the other edge nodes 8 and 9 that receive packets whose bandwidth is controlled by the edge node 10 in the first embodiment and are transmitted in the WAN direction may be general relay apparatuses. Therefore, in the first embodiment, the edge nodes 8 and 9 that connect the base networks 3 and 4 to the WAN 1 are general routers. Note that the edge nodes in the first embodiment may be used for the edge nodes 8 and 9 as well.

[Device configuration]
FIG. 2 is a block diagram illustrating a schematic configuration of the edge node 10 according to the first embodiment. As shown in FIG. 2, the edge node 10 according to the first embodiment includes a base side receiving unit 11, a session analyzing unit 12, a session management table 13, a session queue 14, an output controller 15, a WAN side transmitting unit 16, and a WAN side receiving. Section 21, packet analysis section 22, route search section 23, response confirmation processing section 24, bandwidth control section 25, route information table 26, link set table 27, base side transmission section 28, and the like. Each of these processing units is realized as a software component, a hardware component, or a combination thereof (see [Others]).

  The site-side receiving unit 11 and the site-side transmitting unit 28 operate as a communication interface with the site network 2. The site-side receiving unit 11 receives a packet from the site network 2. If the site side receiving unit 11 determines that the received packet should be sent to the WAN 1, the site side receiving unit 11 notifies the session analyzing unit 12 of reception of this packet. When receiving the packet from WAN 1 from the packet analysis unit 22, the site side transmission unit 28 transmits the packet to the site network 2.

The session analysis unit 12 analyzes the packet received by the base side reception unit 11. Specifically, the session analysis unit 12 determines the destination IP (Internet Protocol) address, source IP address, protocol data, service type, destination port number, source port number, sequence number, packet length from the received packet. Extract code bits and the like.

The session analysis unit 12 identifies the session to which the packet belongs based on the extracted destination IP address, source IP address, destination port number, and source port number. The session used here indicates a communication unit specified by a destination IP address, a source IP address, a destination port number, and a source port number. Thus, for example, one TCP (Transmission Control Protocol) connection corresponds to one session. UDP
Packets exchanged by (User Datagram Protocol) also constitute one session.

  When session analysis unit 12 finishes identifying the session to which the received packet belongs, session analysis unit 12 places the packet in the session queue corresponding to the identified session.

  In addition, the session analysis unit 12 manages the session identified according to the received packet with the session management table 13. FIG. 3 is a diagram illustrating an example of a session management table in the first embodiment. The session management table 13 stores, for each session, a session ID (Identification), session identification information, state, congestion detection information, route information, speed information, queue information, and the like.

The session ID is identification information given by the session analysis unit 12 in order to identify each session within the edge node 10. The session specifying information includes a destination IP address, a source IP address, a destination port number, and a source port number for specifying each session. When the destination IP address, source IP address, destination port number, and source port number extracted from the received packet are not stored in the session management table 13 as session identification information, the session analysis unit 12 Is generated. The session analysis unit 12 adds a new record in which the new session ID, the destination IP address extracted from the received packet, the source IP address, the destination port number, and the source port number are set to the session management table 13. .

When the protocol data extracted from the received packet indicates TCP, the session analysis unit 12 updates the status field in the session management table 13 according to the similarly extracted code bit. For example, when the SYN (Synchronize) flag of the code bit is set, a value (SYN_RCVD) indicating that SYN is received is set in the status field. When the FIN flag of the code bit is set, a value (FIN_WAIT) indicating a FIN packet waiting state is set in the status field. If there is a session in which FIN_WAIT is set in the status field, if the packet analysis unit 22 determines that the FIN packet of the session has been received, the record of the session is deleted from the session management table 13.

The congestion detection information is information used in congestion detection using a response confirmation packet (hereinafter referred to as an ACK packet) in TCP. Congestion detection information includes latest transmission time, maximum transmission sequence number, minimum unacknowledged response sequence number, and average RTT (Round Trip Time)
, Including RTT mean deviation. The latest transmission time indicates the time when the packet received from the base network 2 is transmitted to the WAN 1. In the first embodiment, when the session analysis unit 12 puts a received packet into the session queue, the time information at that time is set as the latest transmission time.

  The maximum transmission sequence number indicates the maximum sequence number in the same session regarding the packet transmitted from the edge node 10 to WAN1. When the sequence number extracted from the received packet is larger than the maximum transmission sequence number set in the record indicating the session in the session management table 13, the session analysis unit 12 performs session management on the extracted sequence number. This is reflected in the table 13.

The minimum unconfirmed response sequence number, average RTT (Round Trip Time), and RTT average deviation are updated by the response confirmation processing unit 24, respectively. In the route information field, a pointer to a record related to the session in the route information table 26 is stored. Since the route information table 26 is managed by the route search unit 23, this route information field is also updated by the route search unit 23.

  The speed information is information used for bandwidth control of each session. The speed information includes priority order, actual transmission bandwidth, and the like. When adding a new session record to the session management table 13, the session analysis unit 12 determines the priority of the session. The session analysis unit 12 stores the determined priority in the priority field. The priority of the session is determined from, for example, the service type and port number extracted from the received packet.

  The actual transmission bandwidth indicates the amount of data per unit time transmitted from the edge node 10 to the WAN 1 for the target session. The session analysis unit 12 accumulates the packet lengths of the packets placed in the session queue, and acquires a value obtained by dividing the accumulated value by a unit time (for example, 10 milliseconds (ms)) as the latest data. The session analysis unit 12 updates the actual transmission bandwidth field with a value obtained by taking a weighted average of the latest data and the data already stored in the actual transmission bandwidth field of the session management table 13.

  The queue information field stores a pointer to a session queue in which packets of the target session are queued. When the session analysis unit 12 puts the packet in the session queue, the session analysis unit 12 stores a pointer to the session queue in the queue information field.

  The session queue 14 is a buffer that is provided for each session and temporarily stores packets to be sent to the WAN 1. As described above, the packet received from the base network 2 is stored in the session queue corresponding to the session identified by the session analysis unit 12.

  The output controller 15 selects each session queue 14 in a predetermined order, extracts the packet stored in the selected session queue 14 by FIFO (First-IN First-OUT), and extracts the extracted packet as WAN. To the side transmitter 16. When receiving an instruction for priority control from the bandwidth controller 25, the output controller 15 extracts a packet and discards the packet in accordance with the instruction. Details of this priority control will be described later.

  The WAN side transmission unit 16 and the WAN side reception unit 21 operate as a communication interface with the WAN 1. The WAN side transmission unit 16 sends the packet extracted by the output controller 15 to the WAN 1. The WAN side receiving unit 21 receives a packet from the WAN 1 and sends the received packet to the packet analyzing unit 22.

  The packet analysis unit 22 analyzes the packet received by the WAN side reception unit 21. When the received packet is an ACK packet, the packet analysis unit 22 notifies the response confirmation processing unit 24 that the ACK packet has been received. When the received packet is an ICMP (Internet Control Message Protocol) Time Exceeded message (hereinafter simply referred to as an “time exceeded message”), the packet analysis unit 22 confirms that the time exceeded message has been received. The route search unit 23 is notified. The packet analysis unit 22 instructs the base side transmission unit 28 to transmit the packet received by the WAN side reception unit 21 to the base network 2.

  The route search unit 23 periodically collects route information of each session stored in the session management table 13. As the route information of each session, information of each relay device (router or the like) on the route is collected. Similarly, when a new record is added to the session management table 13 due to the occurrence of a new session, and when an instruction is received from the bandwidth control unit 25, the route search unit 23 also receives a session route. Get information.

  The route information of the session is collected using an ICMP time exceeded message, for example, in the same manner as a command called “traceroute”. The route search unit 23 sequentially updates predetermined lifetime IP packets for route search with the destination IP address of the session stored in the session management table 13 as the destination while updating the lifetime (TTL) field of the IP header in order from 1. Send. When the route search unit 23 receives a notification of reception of the time excess message from the packet analysis unit 22, the route search unit 23 analyzes the time excess message to acquire route information.

  The route search unit 23 manages the route information collected in this way using the route information table 26. When the route search unit 23 collects route information related to a new destination IP address, it adds a new record to the route information table 26, and stores pointer information for the new record in the session management table 13 described above. Store in the route information field.

FIG. 4 is a diagram illustrating an example of the route information table 26. In the route information table 26,
For each destination IP address, the IP address, average RTT, and RTT average deviation of each router that passes between the edge router 10 and the destination IP address are stored. The route search unit 23 acquires the IP address of each router that passes from the received time excess message to the destination IP address, and stores the acquired IP address in the route information table 26.

  Further, the route search unit 23 measures the response time (RTT) from the transmission time of the IP packet for route search to the reception time of the overtime message for each router. The route search unit 23 calculates a weighted average between the measured RTT of each router and the RTT of each router already stored in the route information table 26, and calculates the average RTT field of the route information table 26 with the calculated value. Update. The weighted average can be expressed by the following equation.

Current average RTT = 1 / n × (currently measured RTT + (n−1) × previous average RTT)
The weighting coefficient n is set to 8, for example. Note that the route search unit 23 may receive a time excess message a plurality of times from the same router by transmitting a predetermined IP packet for a route search in which the same TTL value is set a plurality of times. The route search unit 23 may calculate the average RTT of each router based on the time excess message received a plurality of times.

  When the average RTT is updated, the route search unit 23 calculates the RTT average deviation using the average RTT and the average RTT already stored in the route information table 26 using the following formula.

Current RTT average deviation = 1 / m × (| Current RTT−Existing average RTT | + (m−1) × Previous RTT average deviation)
The weighting coefficient m is set to 4, for example. The route search unit 23 stores the acquired RTT average deviation in the RTT average deviation field of the route information table 26.

  The route search unit 23 updates the link set table 27 when a new destination IP address record is added to the route information table 26. Further, the session analysis unit 12 updates the link set table 27 when detecting the occurrence of a new session.

  FIG. 5 is a diagram illustrating an example of the link set table 27. The link set table 27 stores link information included in each route stored in the route information table 26. Since the link here indicates a connection between routers, each record of the link set table 27 stores a list of IP addresses of two adjacent routers and a session ID using the link. Each IP address of the adjacent router is acquired from the information in the route information table 26. The session ID of the session using the link is extracted from the session management table 13 based on the pointer stored in the route information field of the session management table 13.

  When receiving a notification of reception of the ACK packet from the packet analysis unit 22, the response confirmation processing unit 24 updates the minimum unconfirmed response sequence number, the average RTT, and the RTT average deviation in the session management table 13. The minimum unacknowledged response sequence number field stores the minimum sequence number in the same session in the ACK packet that has not yet been received from WAN1.

The average RTT field stores a weighted average value of time (round-trip response time) from when a packet is transmitted until the ACK packet of this packet is received. In the first embodiment, the response confirmation processing unit 24 receives the difference (time) between the time when the ACK packet is received and the time stamp set in the ACK packet, and the average RTT already stored in the session management table 13. Is stored as an average RTT. The weighted average can be expressed by the above-described formula.

  Normally, the same data as the time stamp set in the packet to be acknowledged is set in the time stamp of the ACK packet. Therefore, strictly speaking, the difference between the time (time stamp) at which the transmission source node transmitted the packet and the time at which the ACK packet for the packet is received at the edge node 10 is acquired as the RTT.

  The average deviation of the average RTT is stored in the RTT average deviation field. When the response confirmation processing unit 24 newly calculates the average RTT, the response confirmation processing unit 24 calculates the RTT average deviation using the above-described equation using the new average RTT.

  When the response confirmation processing unit 24 updates the minimum unconfirmed response sequence number, the average RTT, and the RTT average deviation in the session management table 13, it determines whether congestion has occurred in each session based on the session management table 13. . Specifically, the response confirmation processing unit 24 detects the occurrence of congestion from the increase in RTT and the packet loss rate.

  When congestion occurs in a certain line, packets are accumulated in a buffer of a router connected to the line, and eventually the buffer overflows and packet loss occurs. As a result, an increase in RTT indicates a sign of congestion, and a non-delivery of an ACK packet indicates the occurrence of a packet loss. However, since the range of values that the RTT can take is large depending on the network status (long distance or short distance) and the state (quality), it is difficult to determine the increase in RTT.

  Therefore, the response confirmation processing unit 24 uses the average deviation of the RTT to determine an increase in the RTT when the following condition is satisfied. The predetermined coefficient k is set to 2, for example. By using the average deviation, it is possible to detect an increase in the RTT in consideration of the range of values that the RTT can take in the normal state.

RTT measured this time> Current average RTT + Current RTT average deviation × Predetermined coefficient k
In addition, the response confirmation processing unit 24 determines a packet loss when an ACK packet is not received even after a packet loss determination time indicated by the following equation after a certain packet is sent to the WAN 1. The response confirmation processing unit 24 acquires the packet loss rate by packet loss determination. The packet loss rate is calculated as the number of packet losses that occurred within a predetermined unit time. The response confirmation processing unit 24 determines that the packet loss rate is increased when the calculated packet loss rate is larger than a predetermined threshold. This predetermined threshold is set to an appropriate value by simulation or the like, and is held in a memory or the like so as to be adjustable in advance.

Packet loss judgment time = average RTT + 2 × RTT average deviation × predetermined coefficient k
The response confirmation processing unit 24 determines that congestion of the corresponding session has occurred when an increase in RTT is detected as described above and an increase in the packet loss rate is detected. When the occurrence of congestion is detected, the response confirmation processing unit 24 notifies the bandwidth control unit 25 of the occurrence of congestion together with a session ID indicating a session in which congestion has occurred.

When the bandwidth control unit 25 receives a notification of occurrence of congestion of a session together with the session ID, the bandwidth control unit 25 detects a bottleneck link on the route of the session. Here, the bottleneck link means a link that is highly likely to cause congestion in the session. The bandwidth control unit 25 uses the link set table 27 and the session management table 13 to specify a session that includes the detected bottleneck link on the route. Based on the information stored in the session management table 13, the bandwidth control unit 25 adds up the actual transmission bandwidths of all the specified sessions, and determines the target transmission bandwidth from the total transmission bandwidth. The bandwidth control unit 25 controls the output controller 15 so that the bandwidth used in the session including the bottleneck link on the route becomes the target transmission bandwidth. In this control, a packet of a session having a high priority is taken out preferentially within the target transmission bandwidth for transmission from the session queue 14 to the WAN 1, and a packet exceeding the target transmission bandwidth is discarded from the session queue 14. Detailed processing of the bandwidth control unit 25 will be described later in the section of the operation example.

[Operation example]
Hereinafter, an operation example of the bandwidth control apparatus according to the first embodiment will be described with reference to FIGS. 6A and 6B. 6A and 6B are flowcharts illustrating an operation example of the bandwidth control apparatus according to the first embodiment.

  The response confirmation processing unit 24 monitors the occurrence of congestion in each session based on the session management table 13. In response to the ACK packet received from WAN1, the response confirmation processing unit 24 updates the minimum unconfirmed response sequence number, the average RTT, and the RTT average deviation in the session management table 13, and whether the RTT for each session has increased. It is monitoring whether or not. In addition, the response confirmation processing unit 24 monitors the non-delivery of the ACK packet for each session using the latest transmission time and the packet loss determination time of the session management table 13, and as a result, whether or not the packet loss rate has increased. Determine whether.

  When the response confirmation processing unit 24 detects a session in which the RTT has increased and the packet loss rate has increased, the response confirmation processing unit 24 determines that congestion has occurred in the session (S61). When the response confirmation processing unit 24 detects a session in which congestion occurs, the response confirmation processing unit 24 notifies the bandwidth control unit 25 of the occurrence of congestion together with a session ID for identifying the session.

  When the bandwidth control unit 25 receives the notification of the occurrence of the congestion of the session together with the session ID, the route search so as to newly collect the route information of the session in which the congestion has occurred in order to check the bottleneck link on the route of the session. The unit 23 is instructed.

  The route search unit 23 extracts a destination IP address corresponding to the session ID from the session management table 13 according to an instruction from the bandwidth control unit 25, and collects route information up to this destination IP address (S62). As a result of this collection processing, the route search unit 23 obtains router information (R (i); i is the number of routers) and RTT information (M (i); each router R (i) passed to the destination IP address. RTT). The route search unit 23 sends the acquired router information and RTT information to the bandwidth control unit 25.

  When receiving the router information and the RTT information from the route search unit 23, the bandwidth control unit 25 extracts a pointer to the route information table corresponding to the session ID from the session management table 13. The bandwidth controller 25 extracts the route information indicated by the extracted pointer from the route information table 26 (S63). That is, route information related to the destination IP address of the congestion occurrence session is extracted from the route information table 26. The extracted route information includes router information (R (i); i is the number of routers) passed to the destination IP address and average RTT information (A (i); each router R (i). Average RTT) and RTT average deviation information (D (i); RTT average deviation of each router R (i)).

The bandwidth control unit 25 uses the new route information sent from the route search unit 23 and the existing route information extracted from the route information table 26, and the bottleneck link on the route of the session in which congestion occurs. Is detected (S64). Specifically, the bandwidth control unit 25 first determines that the newly collected RTT (M (i)) among the routers (R (i)) on the route of the session in which the congestion has occurred is the average RTT so far. A router that is significantly different from (A (i)) is extracted. For example, among the routers (R (i)), the router (R (x); i = x) having the maximum value obtained by | A (i) −M (i) | / D (i) is specified. The Next, the bandwidth control unit 25, based on the hop information included in the route information acquired from the route information table 26, the router (R (x−) adjacent to the router (R (x)) in the direction away from the destination. 1)) is specified. As a result, the bandwidth control unit 25 detects the link between the router (R (x)) and the router (R (x−1)) as a bottleneck link.

  The bandwidth control unit 25 specifies a session including the detected bottleneck link (R (x-1), R (x)) on the route (S65). Specifically, the bandwidth control unit 25 extracts a record indicated by a pair of the router (R (x−1)) and the router (R (x)) from the link set table 27, and the session included in this record A list of IDs (a set of sessions (S)) is extracted.

  The bandwidth control unit 25 calculates the sum of the actual transmission bandwidths of the extracted set of sessions (S) (S66). Specifically, the bandwidth control unit 25 extracts the actual transmission bandwidth corresponding to each extracted session ID from the session management table 13, and totals the extracted actual transmission bandwidths.

  The bandwidth control unit 25 determines a target transmission bandwidth from the totaled actual transmission bandwidth (S67). For example, the bandwidth control unit 25 determines the result of multiplying the total actual transmission bandwidth by a predetermined coefficient (for example, 0.5) as the target transmission bandwidth. The bandwidth control unit 25 determines the target transmission data amount by multiplying the target transmission bandwidth by a unit time (for example, 10 (ms)) used for calculation of the actual transmission bandwidth (S68).

  The bandwidth control unit 25 rearranges the set of sessions (S) in descending order of priority stored in the session management table 13 (S70). Subsequently, the bandwidth control unit 25 extracts the session ID indicating the highest priority session from the session set (S), and deletes the extracted session ID from the session set (S) (S71).

  The bandwidth control unit 25 extracts a pointer to the session queue corresponding to the extracted session ID from the session management table 13, and performs output control so as to send a packet of the session queue 14 specified by the extracted pointer. Instruct the device 15. In response to this instruction, the head packet of the session queue 14 is sent to the WAN 1 and deleted from the session queue 14 (S72).

  The bandwidth controller 25 receives the packet length of the transmitted packet from the output controller 15 and accumulates this packet length (S73). The bandwidth control unit 25 determines whether or not the accumulated packet length is larger than the target transmission data amount (S74). When the accumulated packet length exceeds the target transmission data amount (S74; YES), the bandwidth control unit 25 collects the packets remaining in the specified session queue 14 and other sessions ( Packets remaining in the session queues 14 of all sessions included in S) are discarded from each session queue 14 (S75).

  Thereby, according to the bandwidth control apparatus 10 in the first embodiment, the bottleneck link in the network from the edge node 10 including the unmanageable WAN 1 to the destination terminal can be detected, and the bottleneck link is used. The transmission bandwidth of all sessions can be limited. As a result, according to the first embodiment, it is possible to eliminate session congestion.

  When the accumulated packet length is equal to or less than the target transmission data amount (S74; NO), the bandwidth control unit 25 determines whether or not a packet remains in the specified session queue 14 ( S76).

  If the packet remains (S76; NO), the bandwidth control unit 25 instructs the output controller 15 to further send the packet from the specified session queue 14. In response to this instruction, further packets are sent from the session queue 14 to the WAN 1 (S72). On the other hand, when there are no more packets in the specified session queue 14 (S76; YES), the bandwidth control unit 25 newly obtains a session ID indicating a session with a higher priority from the set of sessions (S). Extract (S71). Thereafter, the bandwidth control unit 25 performs the packet transmission process as described above for the newly extracted session (S72).

  The bandwidth control unit 25 sequentially accumulates the packet lengths of the transmitted packets (S73), and repeats the packet transmission process until the accumulated packet length becomes equal to or less than the target transmission data amount. If there is no remaining session in the set of sessions (S) before the accumulated packet length becomes equal to or less than the target transmission data amount (S77; YES), the bandwidth control unit 25 determines the packets of other sessions. Are sequentially sent to the output controller 15 (S78). Thereafter, as usual, packets for each session are sequentially transmitted to WAN1.

<Operation and Effect of Example 1>
As described above, in the bandwidth control device 10 according to the first embodiment, each session relayed by the bandwidth control device 10 uses the session management table 13 by analyzing a packet to be transmitted from the base network 2 in the direction of WAN1. Are managed individually. Packets communicated in each session are respectively stored in a session queue 14 provided for each session, are sequentially taken out from the session queue 14 by the output controller 15 according to the priority order of each session, and sent to the WAN 1. .

  The response confirmation processing unit 24 monitors the ACK packet, and for each session, the round trip delay time, the weighted average value of the round trip delay time (average RTT), and the average deviation of the weighted average response delay time ( RTT average deviation) is obtained and, as a result, an increase in RTT is detected.

  Since the range of values that RTT can take is large, it is difficult to determine the increase in RTT. However, according to the bandwidth control device 10 in the first embodiment, accurate congestion detection based on the increase in RTT is performed by using the average deviation. Can do.

  Further, the response confirmation processing unit 24 monitors the non-delivery of the ACK packet to obtain the packet loss rate, and detects an increase in the packet loss rate when the packet loss rate is greater than a predetermined threshold. Eventually, session congestion is detected by an increase in RTT and an increase in packet loss rate.

  When session congestion is detected, route information to the destination IP address of this session is collected using an ICMP time exceeded message. As a result, based on the router information included in the collected route information and the RTT of each router, a link causing congestion on the route of the session is detected as a bottleneck link.

As described above, according to the first embodiment, since the route information of each session is acquired by the ICMP time exceeded message, even if WAN1 is an unmanaged black box network, the bottleneck link is accurately set. Can be detected.

  In the first embodiment, the route information table 26 manages router information passed to the destination IP address and RTT information (average RTT and RTT average deviation) to each router. Further, based on the route information table 26, information of each link and session information using each link are managed by the link set table 27 by a combination of adjacent routers.

  Thus, according to the first embodiment, when a bottleneck link is detected, a set of sessions including the bottleneck link on the route can be acquired immediately. Since the extracted session set is a session group sharing the bottleneck link, each communication is a cause of congestion. Therefore, transmission of the extracted session group is limited in transmission bandwidth.

  Therefore, according to the first embodiment, when congestion of a certain session is detected, it is possible to quickly specify the communication that is the cause, and to perform bandwidth control that eliminates the congestion only for the cause communication. It can be carried out. According to the first embodiment, even if the WAN 1 is a black box network that is not a management target such as a public network, congestion in the black box network can be avoided.

  Hereinafter, the bandwidth control device (edge node) in the second embodiment will be described. In the first embodiment described above, the route search of each session is performed using the ICMP time exceeded message, and the bottleneck link is detected. The edge node in the second embodiment avoids congestion in the WAN by controlling the bandwidth of the session without using the ICMP time exceeded message. The system configuration of the WAN communication system including the edge node in the second embodiment is the same as that of the first embodiment shown in FIG. The configuration of Example 2 described below is an exemplification, and the present embodiment is not limited to the following configuration.

〔Device configuration〕
FIG. 7 is a block diagram illustrating a schematic configuration of the edge node 10 according to the second embodiment. As illustrated in FIG. 7, the edge node 10 in the second embodiment is replaced with the route search unit 23, the route information table 26, and the link set table 27 in the first embodiment, instead of the shared inspection unit 71, the inspection communication unit 72, and A shared session table 73 is included. Hereinafter, each processing unit will be described with a focus on differences from the first embodiment.

  When the session analysis unit 12 detects a new session, the session analysis unit 12 adds a record indicating the new session to the session management table 13 and notifies the shared inspection unit 71 of information about the new session. In response to this notification, the session analysis unit 12 receives a pointer to the shared session table 73 from the shared inspection unit 71 and stores this pointer in the shared information field of the corresponding record of the session management table 13.

  FIG. 8 is a diagram illustrating an example of a session management table in the second embodiment. As shown in FIG. 8, the session management table 13 in the second embodiment has a shared information field instead of the route information field in the first embodiment. In the shared information field, a pointer to a record indicating a shared session group including the session in the shared session table 73 described later is stored.

In accordance with an instruction from the session analysis unit 12, the shared inspection unit 71 inspects a newly detected session (hereinafter referred to as a new session) and another session that shares a link that can become a bottleneck link in the WAN 1. A set of sessions sharing a link that can be a bottleneck link in WAN 1 is referred to as a shared session group.

  FIG. 9 is an image diagram showing a shared session group. Within the WAN 1, there are a plurality of relay apparatuses 92, 93, 94 and 95 such as routers. Each relay apparatus is connected by links 97, 98 and 99 having a predetermined physical bandwidth. Each link is used for communication for realizing each session. For link 97, sessions 91 (# 2), 91 (# 3), and 91 (# 4) are shared sessions. For link 98, sessions 91 (# 2) and 91 (# 3) are shared sessions.

  By the way, in FIG. 9, each block showing each link 97, 98 and 99 is shown in a size corresponding to its physical bandwidth. In the example of FIG. 9, the link 98 has a smaller physical bandwidth than the other links, and passes through the sessions 91 (# 2) and 91 (# 3) of the terminals 90 (# 2) and 90 (# 3). In this state, if another session is assigned to the link 98, the link 98 becomes a bottleneck link, and congestion is likely to occur. However, although the link 97 is shared by the three sessions 91 (# 2), 91 (# 3), and 91 (# 4), the physical bandwidth is larger than the other links, so other sessions are added. Even if it is done, there is a possibility that congestion will not occur. According to the example of FIG. 9, the shared inspection unit 71 determines sessions 91 (# 2) and 91 (# 3) that share a link that can be a bottleneck link such as the link 98 as a shared session group.

  Information on the shared session group determined by the shared inspection unit 71 is stored in the shared session table 73. FIG. 10 is a diagram illustrating an example of the shared session table 73. Each record of the shared session table 73 stores, as information about each shared session group, a list of session IDs of sessions sharing a link that can be a bottleneck link. Furthermore, in each record, each session ID is classified and stored for each priority. As the priority order of each session in the shared session table 73, the priority order stored in the session management table 13 is used. In each record, one of the sessions having the highest priority among all sessions included in the group is set as the representative session (the session underlined in FIG. 10). In each record, the highest priority of the group is set.

  When the shared inspection unit 71 obtains information on a new session, it checks which shared session group the new session belongs to is stored in the shared session table 73. The concept of this inspection method will be described below, and the details will be described in the operation example section. The shared inspection unit 71 instructs the bandwidth control unit 25 to perform transmission bandwidth control that reduces the transmission bandwidth of the representative session of the shared session group by a predetermined bandwidth and increases the transmission bandwidth of the new session by the predetermined bandwidth. .

  When a predetermined time elapses from the start of the transmission bandwidth control and the RTT increases and the packet loss rate increases for the new session, the shared inspection unit 71 sets a link that can be a bottleneck link with the representative session. Determine not to share. On the other hand, if the RTT does not increase for a new session even after a predetermined time has elapsed from the start of transmission bandwidth control, the shared inspection unit 71 shares a link that can be a bottleneck link with the representative session. That is, it is determined that it belongs to the same shared session group as the representative session. As the predetermined time, for example, a time twice as long as RTT is used. In accordance with this determination result, the shared inspection unit 71 includes the session ID of the new session in a record indicating the same shared session group in the shared session table 73.

  Hereinafter, the basis of the above determination method for determining whether or not the session belongs to the shared session group will be described. In the example of FIG. 9, it is assumed that session 91 (# 2) is a new session and session 91 (# 3) is a representative session. In this case, for example, the shared inspection unit 71 performs transmission bandwidth control such that the transmission bandwidth of the session 91 (# 2) is increased by 100 (Mbps) and the transmission bandwidth of the session 91 (# 3) is reduced by 100 (Mbps). The bandwidth control unit 25 is instructed. In this example, since the total transmission bandwidth of the sessions 91 (# 2) and 91 (# 3) does not change before and after the control, an increase in RTT does not occur. That is, the sessions 91 (# 2) and 91 (# 3) can be determined as the same shared session group sharing the bottleneck link 98.

  On the other hand, when session 91 (# 2) is a new session and session 91 (# 4) is a representative session, the total transmission bandwidth of sessions 91 (# 2) and 91 (# 3) after transmission bandwidth control The width will increase by 100 (Mbps) from before the control. In this example, congestion occurs in the link 98, and an increase in RTT occurs. That is, it can be determined that the sessions 91 (# 2) and 91 (# 4) do not share the bottleneck link 98 and do not belong to the same shared session group.

  That is, the shared inspection unit 71 detects whether the total transmission bandwidth does not change before transmission bandwidth control and after transmission bandwidth control or increases by a predetermined bandwidth by increasing the RTT, etc. inspect. Here, the predetermined bandwidth may be set to an optimum value by repeating simulation or the like. For example, the predetermined bandwidth is set to 0.5 times the actual transmission bandwidth of the new session.

  Regarding the transmission bandwidth control as described above, the shared inspection unit 71 specifically designates the transmission bandwidths of the new session and the representative session as the inspection bandwidths, thereby allowing the bandwidth control unit 25 to transmit the transmission bandwidths. Request control. Based on this designation, the bandwidth control unit 25 controls the output controller 15 so that the transmission bandwidths of the new session and the representative session become the inspection bandwidth. The bandwidth control unit 25 monitors the actual transmission bandwidth field of the session management table 13, and when the actual transmission bandwidth of the representative session exceeds the designated inspection bandwidth, the bandwidth control unit 25 starts from the session queue 14 in which the representative session is stored. The output controller 15 is instructed to stop taking out. On the other hand, when the actual transmission bandwidth of the new session is smaller than the designated inspection bandwidth, the bandwidth control unit 25 causes the inspection communication unit 72 to include an inspection dummy packet that is less than the designated inspection bandwidth. Instruct to send minutes.

  Further, the bandwidth control unit 25 is different from the first embodiment in the processing when receiving a notification of occurrence of congestion from the response confirmation processing unit 24. When the bandwidth control unit 25 receives the notification of occurrence of congestion, the bandwidth control unit 25 extracts information on the shared session group including the session ID to be notified from the shared session table 73. The bandwidth control unit 25 limits the transmission bandwidth of a session included in the shared session group using the extracted shared session group as a bandwidth limitation target by the same method as in the first embodiment.

  During the inspection of the shared session group as described above, the session management table 13 stores a record indicating the new session to be inspected, but the shared session table 73 has a session ID indicating the new session. Not stored. If the session ID to be notified of congestion occurrence is not included in the shared session table 73, the bandwidth control unit 25 determines that the shared session group is being inspected, and the occurrence of congestion together with this session ID is shared. Notification to the unit 71.

The inspection communication unit 72 generates an inspection dummy packet for the designated bandwidth in accordance with an instruction from the bandwidth control unit 25 and sends the inspection dummy packet to the output controller 15. Hereinafter, this inspection dummy packet is simply referred to as an inspection packet. The inspection packet uses, for example, a UDP discard packet in which a discard (DISCARD) port (port number = 9) is set in the destination port number field and the destination IP address of the new session is set in the destination IP address field. Is done.

[Operation example]
Hereinafter, an operation example of the bandwidth control apparatus according to the second embodiment will be described with reference to FIGS. 11 and 12.

  FIG. 11 is a flowchart illustrating an operation example of the shared session group inspection according to the second embodiment. The session analysis unit 12 determines whether or not the session identified according to the received packet is already managed in the session management table 13 based on the analysis result of the packet received by the base side reception unit 11. When the session is not yet stored in the session management table 13, the session analysis unit 12 determines that the session is a new session (S101). In FIG. 11, the new session detected by the session analysis unit 12 is represented as a new session (i). The session analysis unit 12 adds a record indicating the new session to the session management table 13 and notifies the shared inspection unit 71 of information about the new session.

  When the shared inspection unit 71 obtains information on the new session, the shared inspection unit 71 selects the records stored in the shared session table 73 in order from the record indicating the shared session group having the highest priority (S102). The shared inspection unit 71 extracts the representative session of the selected shared session group from the shared session table 73 (S103). In FIG. 11, this extracted representative session is represented as a representative session (j).

  The sharing checking unit 71 checks whether the new session (i) and the representative session (j) share a link that can be a bottleneck link as follows. The shared inspection unit 71 determines the inspection bandwidth (J) of the representative session (j) and the inspection bandwidth (I) of the new session (i). Specifically, the shared inspection unit 71 subtracts a value obtained by subtracting a value obtained by multiplying the actual transmission bandwidth of the new session (i) by the inspection coefficient from the actual transmission bandwidth of the representative session (j). The inspection bandwidth (J) is determined (S104). In the example of FIG. 11, the inspection coefficient is set to 0.5. The shared inspection unit 71 adds a value obtained by adding a value obtained by multiplying the actual transmission bandwidth of the new session (i) to the actual transmission bandwidth of the new session (i) to the inspection bandwidth ( I) is determined (S105). The actual transmission bandwidth of each session is extracted from the actual transmission bandwidth field of the corresponding record in the session management table 13.

  The shared inspection unit 71 sends a transmission bandwidth control request to the bandwidth control unit 25 together with the inspection bandwidths (I) and (J) of the new session (i) and the representative session (j) (S106). The bandwidth control unit 25 controls the output controller 15 so that the new session (i) and the representative session (j) are transmitted with the inspection bandwidths (I) and (J), respectively. Specifically, the bandwidth control unit 25 stores the representative session (j) when the actual transmission bandwidth of the representative session (j) stored in the session management table 13 exceeds the inspection bandwidth (J). The output controller 15 is instructed to stop the removal from the session queue 14 to be executed. Thereby, the packet of the representative session (j) whose bandwidth is limited is taken out from the section queue 14 and transmitted so as not to exceed the inspection bandwidth (J).

  On the other hand, when the actual transmission bandwidth of the new session (i) is smaller than the inspection bandwidth (H) (S107; YES), the bandwidth control unit 25 calculates the actual transmission bandwidth from the inspection bandwidth (H). The inspection communication unit 72 is requested to transmit the inspection packet together with the bandwidth information obtained by subtracting.

  In response to a request from the bandwidth control unit 25, the inspection communication unit 72 generates a UDP discard packet having the same destination as that of the new session (i), and the output controller 15 outputs an amount of the discarded packet that does not satisfy the inspection bandwidth. (S108).

  The response confirmation processing unit 24 determines whether congestion has occurred in each session by monitoring the RTT increase and the packet loss rate based on the session management table 13 as in the first embodiment. When the bandwidth control unit 25 receives a notification of the occurrence of congestion from the response confirmation processing unit 24, the bandwidth control unit 25 determines whether or not the shared session table 73 stores a session ID that is a target of the notification. When this session ID is the session ID of the new session (i), this session ID is not stored in the shared session table 73. In this case, the bandwidth control unit 25 determines that the shared session group check is in progress, and sends a congestion occurrence notification to the shared check unit 71 together with the session ID of the new session (i).

  When receiving the notification of occurrence of congestion and the session ID, the shared inspection unit 71 determines whether or not a predetermined time (2 × RTT time) has elapsed since the start of the shared session group inspection for the session ID. If a predetermined time has elapsed from the start of the inspection, the shared inspection unit 71 determines that congestion has occurred in the new session (i) (S110; YES), and the new session (i) and the representative session (j) are It is determined that they do not belong to the same shared session group (S112).

  On the other hand, if the shared inspection unit 71 does not receive a notification of occurrence of congestion even after a predetermined time has elapsed from the start of the inspection, it determines that congestion does not occur in the new session (i) (S110). When the shared inspection unit 71 determines that congestion does not occur in the new session (i) (S110; NO), it determines that the new session (i) and the representative session (j) belong to the same shared session group (S111).

  When the shared inspection unit 71 determines that the new session (i) and the representative session (j) belong to the same shared session group, the shared inspection unit 71 adds a new record to the shared session group record to which the representative session (j) belongs. Add a session ID for session (i). The shared inspection unit 71 confirms the presence of another shared session group in the shared session table 73 (S113). If there is no other shared session group (S113; NO), the shared inspection unit 71 ends the process. On the other hand, when there is another shared session group (S113; YES), the shared inspection unit 71 selects a record indicating the shared session group with the next highest priority from the shared session table 73 (S102). Thereafter, the same processing as described above is executed for the newly selected shared session group.

  FIG. 12 is a flowchart illustrating an operation example of transmission bandwidth control in the second embodiment. The transmission bandwidth control in the second embodiment is different from the first embodiment shown in FIGS. 6A and 6B in the way of acquiring a set of sessions (S), and other processes are the same as those in the first embodiment.

  As in the first embodiment, the response confirmation processing unit 24 detects the occurrence of session congestion (S61). When the bandwidth control unit 25 receives a notification of occurrence of session congestion together with the session ID, the bandwidth control unit 25 acquires a pointer to the shared session table stored in the shared information field of the session management table 13. The bandwidth control unit 25 identifies the corresponding record in the shared session table 73 using the acquired pointer. This specified record is a record indicating a shared session group including the target session.

The bandwidth control unit 25 mainly uses a set of sessions (S) included in the record specified in the shared session table 73 (S122). Thereafter, transmission band control is executed by the same method as in the first embodiment (S66 and after).

<Operation and Effect of Example 2>
In the bandwidth control apparatus 10 according to the second embodiment, when a new session is detected by analyzing a packet to be transmitted from the base network 2 in the WAN1 direction, the shared inspection unit 71 causes a bottleneck link in the new session and WAN1. Other sessions sharing possible links are examined. In this inspection, a set of sessions sharing a link that can be a bottleneck link in the WAN 1 is specified by controlling the bandwidth of the session without using an ICMP time exceeded message.

  Thus, according to the second embodiment, even if the WAN 1 includes a relay device that does not support the ICMP time over message, a set of sessions sharing a link that can be a bottleneck link is appropriately identified. can do.

  In the second embodiment, when a new session is detected, a set of sessions sharing a link that can be a bottleneck link with the new session is acquired and stored in the shared session table 73.

  Thus, according to the second embodiment, when a session in which congestion has occurred is detected, the communication that causes it quickly is accurately identified, and the bandwidth control that eliminates the congestion only for the cause communication It can be performed. According to the second embodiment, even if WAN1 is an unmanaged black box network such as a public network and the relay device in WAN1 does not support the ICMP time over message. , Congestion in WAN 1 can be avoided.

[Supplement]
In the above-described first and second embodiments, the response confirmation processing unit 24 monitors the ACK packet to identify a session in which congestion has occurred. That is, the session that can detect the occurrence of congestion in the response confirmation processing unit 24 is a TCP session. However, the sessions managed by the session management table 13, the route information table 26, the link set table 27, and the shared session table 73 are not limited to TCP sessions. Accordingly, a session other than a TCP session that shares a bottleneck link with a TCP session in which congestion has been detected (for example, a UDP session) can naturally be a target whose transmission bandwidth is limited. In the second embodiment, the shared session group is inspected by bandwidth control even when the new session is a session other than TCP.

[Others]
<About hardware components (Component) and software components (Component)>
A hardware component is a hardware circuit, for example, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a gate array, a combination of logic gates, a signal processing circuit, an analog circuit, etc. There is.

A software component is a component (fragment) that realizes the above processing as software, and is not a concept that limits a language, a development environment, and the like that realize the software. Examples of software components include tasks, processes, threads, drivers, firmware, databases, tables, functions, procedures, subroutines, predetermined portions of program code, data structures, arrays, variables, parameters, and the like. These software components are realized in one or a plurality of memories in a computer, or data in one or a plurality of memories is stored in one or a plurality of processors (for example, a CPU (Central Processes).
sing Unit), DSP (Digital Signal Processor), etc.).

  In addition, each above-mentioned embodiment does not limit the realization method of each said process part. Each processing unit may be configured as a hardware component, a software component, or a combination thereof by a method that can be realized by a normal engineer in this technical field.

<Appendix>
Regarding the above-described embodiment, the following additional notes are disclosed. The aspects disclosed in each section can be combined as much as possible as necessary.
(Appendix 1)
In a relay device that relays packets from the first network to the second network,
Congestion detection for detecting a communication unit in which congestion occurs among a plurality of communication units specified by at least a destination address and a transmission source address set in each packet relayed from the first network to the second network Means,
A specifying unit that specifies, among the plurality of communication units, another communication unit that shares at least one specific link in the second network with the communication unit in which congestion is detected detected by the congestion detection unit; ,
Band control means for limiting the transmission band to the second network of the other communication units specified by the specifying means and the communication units in which the congestion occurs among the plurality of communication units;
A relay device comprising:
(Appendix 2)
A collecting means for collecting route information of the communication unit in which the congestion occurs;
Detecting means for detecting a bottleneck link included in the path of the communication unit in which the congestion occurs, according to the path information collected by the collecting means;
Further comprising
The specifying unit specifies another communication unit that shares the bottleneck link detected by the detecting unit as the at least one specific link;
The relay device according to Supplementary Note 1, wherein:
(Appendix 3)
The collection means collects route information for each destination address of the plurality of communication units,
The specifying means includes storage means for storing, on the basis of the path information of the plurality of communication units collected by the collecting means, each set of communication units using the link for each link included in the path information. Including,
The relay device according to appendix 1 or 2, characterized in that.
(Appendix 4)
The congestion detection means determines whether or not the round trip delay time of each communication unit has increased using the average value of the round trip delay time and the average deviation of the round trip delay time, and identifies the session in which the congestion has occurred The relay apparatus according to any one of appendices 1 to 3, wherein the determination result of the increase in the round-trip delay time is used.
(Appendix 5)
Band control for increasing the transmission bandwidth of one communication unit by a predetermined bandwidth and reducing the transmission bandwidth of the other communication unit by the predetermined bandwidth in two communication units of the plurality of communication units And the one communication unit and the other communication unit are at least one specified in the second network according to a communication state of the one communication unit after the band control. Determining means for determining whether or not to share the link;
The relay device according to appendix 1, further comprising:
(Appendix 6)
Storage means for storing information on a set of communication units sharing at least one specific link in the second network based on a determination result by the determination means;
The relay device according to appendix 5, further comprising:
(Appendix 7)
In a bandwidth control method executed by a relay device that relays packets from the first network to the second network,
Detecting a communication unit in which congestion occurs among a plurality of communication units respectively specified by at least a destination address and a source address set in each packet relayed from the first network to the second network; ,
Identifying another communication unit sharing at least one specific link in the second network with the detected communication unit among the plurality of communication units;
Limiting the transmission bandwidth to the second network of the identified other communication unit and the communication unit in which the congestion has occurred among the plurality of communication units;
A bandwidth control method comprising:

1 WAN
2, 3, 4 Base network 10 Bandwidth control device (edge node)
DESCRIPTION OF SYMBOLS 11 Base side receiving part 12 Session analysis part 13 Session management table 14 Session queue 15 Output controller 16 WAN side transmission part 21 WAN side receiving part 22 Packet analysis part 23 Path search part 24 Response confirmation processing part 25 Band control part 26 Route information Table 27 Link set table 28 Base side transmission unit 71 Shared inspection unit 72 Inspection communication unit 73 Shared session table

Claims (4)

In a relay device that relays packets from the first network to the second network,
Congestion detection for detecting a communication unit in which congestion occurs among a plurality of communication units specified by at least a destination address and a transmission source address set in each packet relayed from the first network to the second network Means,
A specifying unit that specifies, among the plurality of communication units, another communication unit that shares at least one specific link in the second network with the communication unit in which congestion is detected detected by the congestion detection unit; ,
Band control means for limiting the transmission band to the second network of the other communication units specified by the specifying means and the communication units in which the congestion occurs among the plurality of communication units;
Band control for increasing the transmission bandwidth of one communication unit by a predetermined bandwidth and reducing the transmission bandwidth of the other communication unit by the predetermined bandwidth in two communication units of the plurality of communication units And the one communication unit and the other communication unit are at least one specified in the second network according to a communication state of the one communication unit after the band control. Determining means for determining whether or not to share the link;
A relay device comprising: A collecting means for collecting route information of the communication unit in which the congestion occurs;
Detecting means for detecting a bottleneck link included in the path of the communication unit in which the congestion occurs, according to the path information collected by the collecting means;
Further comprising
The specifying unit specifies another communication unit that shares the bottleneck link detected by the detecting unit as the at least one specific link;
The relay apparatus according to claim 1. The collection means collects route information for each destination address of the plurality of communication units,
The specifying means includes storage means for storing, on the basis of the path information of the plurality of communication units collected by the collecting means, each set of communication units using the link for each link included in the path information. Including,
The relay apparatus according to claim 2 . In a bandwidth control method executed by a relay device that relays packets from the first network to the second network,
Detecting a communication unit in which congestion occurs among a plurality of communication units respectively specified by at least a destination address and a source address set in each packet relayed from the first network to the second network; ,
Identifying another communication unit sharing at least one specific link in the second network with the detected communication unit among the plurality of communication units;
Limiting the transmission bandwidth to the second network of the identified other communication unit and the communication unit in which the congestion has occurred among the plurality of communication units;
Band control for increasing the transmission bandwidth of one communication unit by a predetermined bandwidth and reducing the transmission bandwidth of the other communication unit by the predetermined bandwidth in two communication units of the plurality of communication units Whether the one communication unit and the other communication unit share at least one specific link in the second network according to the communication state of the one communication unit after the bandwidth control. Determining whether or not,
A bandwidth control method comprising:

Images