JP2019054496A - Communication relay device - Google Patents

Abstract

To provide a communication device saving a communication quality while housing a flow of the number of ques or more.SOLUTION: A communication relay device that relays a packet, comprises: a reception part that receives the packet and outputs the packet to be received; a distribution part that converts the packet output from the reception part to a hush value, and distributes the packet output by the reception part to a plurality of queues on the basis of a correspondence relationship of the hush value and each queue and the converted hush value; a buffer that stores the packet distributed by the distribution part into the plurality of queues; a learning part that accumulates information on a state of the plurality of queues of the buffer; a learning result processing part that generates information of a relationship where the plurality of hush values correspond to the queues on the basis of the information on the state in regards to the queues accumulated by the learning part, and transmits it to the distribution part; and a transmission part that transmits the packet stored to the buffer.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a communication relay device.

  In recent years, so-called streaming technology that distributes electronic content such as video and audio in real time via a network has become widespread. When such streaming is realized via the Internet, the same line is shared among many users. For this reason, the cost per band can be kept low.

  On the other hand, communication quality (QoS: Quality of Service) is ensured by suppressing the amount of discarded communication packets. For this reason, a technology is known in which a relay device connected to the Internet is equipped with a shaper, and data that is transmitted in large quantities at a time is leveled and transferred at a constant rate.

  In Patent Document 1, flow conditions corresponding to users and services are set in advance as configuration definition information, and received communication packets are temporarily stored in a queue prepared for each flow, and predetermined shaping conditions (transmission) A technique for transferring a communication packet based on a bandwidth value and a priority) is disclosed.

JP-A-11-346246

  However, since the technique of Patent Document 1 needs to set a flow for each user or service, it is difficult to deal with a large number of users and services. For example, when a flow is set for each millions of users, it is difficult to secure the same number of queues as the number of users. Even if the same number of queues as the number of users can be secured, it is inefficient to secure queues for users with a low active rate.

  An object of the present invention is to protect communication quality while accommodating a flow exceeding the number of queues.

  In order to solve the above-described problems, a typical communication device according to the present invention is a communication relay device that relays a packet, and receives a packet and outputs a received packet, and the reception unit outputs A distribution unit that converts the packet output from the reception unit into a plurality of queues based on the converted hash value according to a correspondence relationship between the hash value and the queue, A buffer for storing packets distributed by the plurality of queues in a plurality of queues, a learning unit for storing state information regarding each of the plurality of queues of the buffer, and a state information regarding each of the queues stored by the learning unit, A learning result processing unit that creates relation information corresponding to a plurality of hash values in a queue and sends the information to the distribution unit, and a buffer stored in the buffer. And a sending unit for transmitting Tsu bets.

  According to the present invention, it is possible to protect the communication quality while accommodating flows exceeding the number of queues. Problems, configurations, and effects other than those described above will become apparent from the description of the following examples.

It is a figure which shows an example of a communication system. It is a block diagram which shows the structural example of a router. It is a figure which shows the structural example of a distribution part. It is a figure which shows the structural example of a learning part. It is a figure which shows the structural example of a learning result process part. It is a flowchart which shows the example of the creation process procedure by a learning result process part. It is a flowchart which shows the example of the update process procedure by a learning result process part.

<System configuration example>
FIG. 1 is a diagram illustrating an example of a communication system. The communication system 100 is a network system that provides content such as moving images to users by streaming, for example. As illustrated, the communication system 100 includes servers 111 and 112, routers 120, 140, and 161 to 163 that are relay devices, and personal computers 171 to 173 that are user terminals.

  The servers 111 and 112 are servers that provide content, and store various types of content. The servers 111 and 112 are directly connected to the router 120. Note that the number of servers connected to the router 120 is shown as two for simplicity of explanation, but may be three or more.

  The router 120 is connected to the network 130. The network 130 is, for example, the Internet. However, the type of the network 130 is not particularly limited, and may be a WAN (Wide Area Network) such as a dedicated line or a LAN (Local Area Network).

  The router 140 is also connected to the network 130. The servers 111 and 112 and the routers 120 and 140 are installed by, for example, an ISP (Internet Service Provider) that provides a content distribution service.

  The router 140 is connected to personal computers 171 to 173 via routers 161 to 163 installed in the homes 151 to 153, respectively. Another relay device or a concentrator may be interposed between the router 140 and the routers 161 to 163.

  The personal computers 171 to 173 are user terminals and are general-purpose personal computers. However, the personal computers 171 to 173 are not limited to this, and the terminals connected to the routers 161 to 163 are contents distributed by the servers 111 and 112. As long as the computer can be used, for example, a network-compatible television may be used.

  A plurality of terminals (not shown) may be connected to each of the routers 161 to 163. In addition, the router 140 is connected to the networks of the homes 151 to 153 for the sake of simplicity of explanation, but is actually connected to a network of thousands or tens of thousands of homes.

  In the communication system 100, when the server 111 or 112 receives a content usage request from the personal computers 171 to 173, it distributes the requested content to the usage requesting personal computers 171 to 173 in packets. .

  When receiving the packets distributed by the servers 111 and 112, the routers 120 and 140 temporarily accumulate the respective packets in a buffer, level the packets, and transfer the packets to the personal computers 171 to 173. The personal computers 171 to 173 receive the packets transferred via the routers 161 to 163 and use them in applications installed in advance.

  Packets distributed from the servers 111 and 112 to the personal computers 171 to 173 are, for example, three communication flows corresponding to the three personal computers 171 to 173 as destinations. However, the flow is not limited to this, and may be a flow identified based on other information.

<Example of router configuration>
FIG. 2 is a block diagram illustrating a configuration example of a router. Since the routers 120 and 140 shown in FIG. 1 have the same configuration in this embodiment, the configuration of the router 120 will be described below unless otherwise specified.

  The router 120 includes a reception unit 201, a distribution unit 202, a learning unit 203, a buffer 204, a bandwidth control unit 205, a transmission unit 206, and a learning result processing unit 207.

  The receiving unit 201 is an interface that receives packets 250r-1 to 250r-f (f is an arbitrary integer equal to or greater than 2) from the outside. In the present embodiment, the receiving unit 201 includes two physical ports (not shown). Servers 111 and 112 are connected to these two physical ports.

  The receiving unit 201 receives the packets 250r-1 to 250r-f from the servers 111 and 112, and outputs them to the allocating unit 202. Note that the receiving unit 201 may include three or more physical ports. In this case, the receiving unit 201 may be connected to three or more servers.

  Each of the packets 250r-1 to 250r-f is a packet belonging to the flows 260r-1 to 260r-f. For this reason, f is the number of flows, and may be the number of destination personal computers to which the packets 250r-1 to 250r-f are distributed. Each of the packets 250r-1 to 250r-f includes headers 251r-1 to 251r-f.

  Since the packet output from the receiving unit 201 to the allocating unit 202 has been received and is information inside the router 120, it is expressed as a packet 250 to indicate that it is internal information. The contents are the same as any one of the packets 250r-1 to 250r-f. Correspondingly, in the router 120, the flows 260r-1 to 260r-f are expressed as a flow 260, and the headers 251r-1 to 251r-f are expressed as a header 251.

  The distribution unit 202 distinguishes the packets 250 input from the reception unit 201 from the queues 214-1 to 214-n (n is an arbitrary integer equal to or greater than 2 and queues 214-1 to 214-n) secured in the buffer 204. If not, it is distributed to one of the queues 214).

  Here, each of the queues 214-1 to 214-n is pointed by Qp1 to Qpn which are queue pointers. However, Qp1 to Qpn may be the names and identifiers of the queues 214-1 to 214-n as long as they are information specifying that the queues 214-1 to 214-n can be accessed.

  As a specific distribution process, for example, the distribution unit 202 converts the header information included in the header 251 of the packet 250 into a hash value using a predetermined hash function, and the packet 250 is a queue corresponding to the hash value. 214.

  For this purpose, the distribution unit 202 includes a conversion unit 221 and a hash value table 222. The conversion unit 221 converts the header information included in the header 251 of the packet 250 to a hash value H by giving the hash function.

  The information given to the hash function is header information included in the header 251 of the packet 250, and includes, for example, a layer 2 header such as a MAC header, a layer 3 header such as an IPv4 header and an IPv6 header, and a TCP header and a UDP header. Any one or more fields of the layer 4 header may be used.

  The selection of the field to be used may be fixed to the device (preset by the router 120 and fixed), or may be specified by the device operator. However, it is preferable that a field including information that can identify the communication flow 260 distributed from the servers 111 and 112 to the personal computers 171 to 173 is selected, and a field including the destination address of the personal computers 171 to 173 is selected. It is preferred that

  The hash value table 222 is a table for associating the hash value H obtained from the conversion unit 221 with the queue 214 of the buffer 204. For example, when the hash value H is H1, the corresponding queue 214 stores queue pointer information such as Qp1.

  Information on the queue pointer in the hash value table 222 is rewritable and can be changed by an instruction from the learning result processing unit 207. The allocating unit 202 is realized by an integrated circuit, for example. In this case, the hash value table 222 is stored in a memory in the integrated circuit.

  The learning unit 203 monitors the packet 250 distributed to the plurality of queues 214 corresponding to the hash value by the distribution unit 202, and distributes the packet 250 with which hash value and the flow rate to which queue 214. To learn.

  Specifically, for example, the learning unit 203 receives the hash value H and the queue pointer Qp together with the packet 250, and monitors the current queue length and the flow rate of the packet 250 using the queue pointer Qp as a target queue for each hash value H. And keep it as learning information.

  The learning unit 203 sends the information of the packet 250 to the queue 214 in accordance with the distribution by the distribution unit 202, and outputs the learning result to the learning result processing unit 207. Note that the learning unit 203 is realized, for example, by an integrated circuit or by causing a processor to execute a program stored in a memory.

  The buffer 204 is a storage area for temporarily storing the packet 250 received by the receiving unit 201, and includes a plurality of queues 214-1 to 214-n. In the present embodiment, for example, the same storage capacity is allocated to each of the queues 214-1 to 214-n.

  Each of the queues 214-1 to 214-n stores each packet 250 according to the distribution result determined by the distribution unit 202 as described above. Note that, at the time of congestion, the packet 250 exceeding the capacity of any one of the queues 214-1 to 214-n is discarded.

  For example, when packets 250 corresponding to the capacity of the queue 214-i have already been accumulated in the queue 214-i (i is an arbitrary integer from 1 to n), the packet 250 newly input to the buffer 204. Are discarded without being stored in the queue 214-i.

  The bandwidth control unit 205 controls the bandwidth for each of the queues 214-1 to 214-n, and outputs the packets 250 accumulated in the queues 214-1 to 214-n to the transmission unit 206. Specifically, for example, the bandwidth control unit 205 performs scheduling based on preset bandwidth control rules, sequentially reads the packets 250 from the queues 214-1 to 214-n, and outputs the packets 250 to the transmission unit 206. . Band controller 205 is realized by, for example, an integrated circuit.

  The transmission unit 206 is an interface that transmits the packet 250 to the outside of the router 120. The contents of the packets 250s-1 to 250s-f transmitted to the outside are the same as the contents of the packets 250r-1 to 250r-f, and the flows 260s-1 to 260s-f are flows 260r-1 to 260r, respectively. Corresponds to -f.

  The headers 251s-1 to 251s-f correspond to the headers 251r-1 to 251r-f, but some information may be rewritten by the operation of the router 120 as a router.

  The learning result processing unit 207 receives the learning result from the learning unit 203 and executes various processes. For example, the learning result processing unit 207 calculates based on the learning result so as to bundle flows having a flow rate equal to or lower than a specific threshold, and reflects the calculated result in the hash value table 222 of the allocating unit 202.

  In addition, the learning result processing unit 207 uses a message such as a trap or a system log to transmit the learning result from the transmission unit 206 to the operator's terminal or display it on a display unit (not shown). Notify operators of learning results.

  Furthermore, when the learning result is registered or deleted, the learning result processing unit 207 may notify the operator of registration or deletion information. Note that the learning result processing unit 207 is specifically realized by causing a processor to execute a program for executing the processing of the learning result processing unit 207, for example.

<Configuration Example of Distribution Unit 202>
FIG. 3 is a diagram illustrating a configuration example of the distribution unit 202. As shown in FIG. 2, the distribution unit 202 includes a conversion unit 221 and a hash value table 222.

  The conversion unit 221 converts the packet 250 into a hash value H based on a hash function. The range of the hash value H converted by the conversion unit 221 can be converted in a range that is equal to or greater than the number of queues 214 included in the apparatus. In this example, conversion is possible within a range of about twice the number of queues 214.

  The hash value table 222 has a hash value 301 and a queue pointer 302. The hash value 301 is a storage area for storing the hash values H1 to Hm (m is an arbitrary integer equal to or greater than 2, and is expressed as the hash value H when the hash values H1 to Hm are not distinguished).

  The queue pointer 302 is a storage area for storing the queue pointers Qp1 to Qpn (n is the number of the queues 214, and is expressed as the queue pointer Qp when the queue pointers Qp1 to Qpn are not distinguished). The queue pointer Qp serves as a pointer that links the hash value H and the queue 214.

  Specifically, in the example of FIG. 3, when the queue 214-1 corresponds to the hash value H1 of the hash value 301, the queue pointer Qp1 pointing to the queue 214-1 is stored in the queue pointer 302.

  In the example of FIG. 3, since the number of hash values 301 is about twice the number of queues 214, m is about twice the value of n. For example, the queue pointer Qp1 of the queue pointer 302 is This corresponds to the two hash values H of the hash value 301. That is, there are two queue pointers Qp1 of the queue pointer 302 that links the queue 214-1 to the hash value H.

  In the hash value table 222, the queue 214 may not be associated with the hash value H of the hash value 301. In this way, NULL, which is information that is not associated with any of the queues 214 with the hash value H, may be stored in the queue pointer 302. Then, when the hash value H is used, a queue pointer Qp for associating the queue 214 with the hash value H may be stored instead of NULL.

  When the number of hash values H1 to Hm is sufficient with respect to the number of flows 260r-1 to 260r-f that are actually communicated, and the hash function is a function in which no collision occurs in that state, Since one packet 260 of one flow 260 is converted into one hash value H, one hash value H described below may be regarded as one flow 260.

<Detailed Configuration Example of Learning Unit 203>
FIG. 4 is a diagram illustrating a configuration example of the learning unit 203. The learning unit 203 monitors and learns the target queue Qp, the current queue length QL, and the flow rate R for each hash value H with respect to the flow 260 of the packet 250 distributed to each queue 214 by the distribution unit 202. For this purpose, the learning unit 203 inputs the hash value H and the queue pointer Qp from the distribution unit 202 together with the packet 250.

  In the learning information 400, each of the hash value 401 and the target queue 402 stores the hash value H and the queue pointer Qp input from the distribution unit 202 in the input correspondence relationship, and the correspondence relationship is stored in the learning unit. 203 learns.

  The current queue length 403 stores the number of packets 250 in the queue 214 pointed to by the queue pointer Qp of the target queue 402 and converted into the hash value H of the hash value 401. Instead of the number of packets 250, the number of bytes of one or more packets 250 may be used.

  The flow rate 404 stores the number of packets 250 output to the queue 214 pointed to by the queue pointer Qp of the target queue 402 and converted into the hash value H of the hash value 401 per second. Instead of the number of packets 250, it may be the number of bytes of one or more packets 250, or per other period instead of per second.

  The flow rate 404 is a packet 250 output from the queue 214 pointed to by the queue pointer Qp of the target queue 402, and stores the number of packets 250 converted to the hash value H of the hash value 401 per second. Also good. For this purpose, the learning unit 203 may acquire information for the flow rate R from the buffer 204 or the band control unit 205.

  In the example of FIG. 4, the number of packets 250 converted to the hash value H1 in the queue 214-1 pointed to by the target queue Qp1 is the current queue length QL1, and the hash value H1 output to the queue 214-1 The number of packets 250 converted into 1 per second indicates that the value is the flow rate R1.

  For the hash value H7, the queue 214 is not linked, and the current queue length 403 and flow rate 404 are constant C. Here, the constant C is preferably a number that does not correspond to the rule described with reference to FIG.

  The learning unit 203 may perform learning by monitoring and updating the learning information 400 for each packet 250 (hash value H and queue pointer Qp) input from the allocating unit 202, or may be set in advance. Learning may be performed by monitoring the packet 250 input from the allocating unit 202 and updating the learning information 400 every period.

<Configuration Example of Learning Result Processing Unit 207>
FIG. 5 is a diagram illustrating a configuration example of the learning result processing unit 207. The learning result processing unit 207 receives the learning result 271 from the learning unit 203, reflects it in the redistribution information 272, and sends it to the distribution unit 202. The learning result 271 is a learning result of the learning information 400 illustrated in FIG.

  Upon receiving the learning result 271 from the learning unit 203, the learning result processing unit 207 associates the hash value H with the queue pointer Qp of the target queue, that is, links the hash value H and the queue 214 based on a predetermined rule. Redistribute.

  For example, in the case of a rule that aggregates packets in units of hash values so that the current queue length is 0 and the flow rate is 30 or less in one queue 214, the flow rate already exceeds 30. The hash value H2 and the hash value H3 for which the current queue is already 1 or more are excluded, and the learning result processing unit 207 determines that the learning result 2711, the learning result 2713, and the learning result 2714 can be aggregated.

  As a result of the determination, the learning result processing unit 207 associates the hash value H4 and the hash value H5 of the learning result 2713 and the learning result 2714 with the queue 214-1 pointed to by the queue pointer Qp1 that is the target queue of the hash value H1. In the redistribution information 272, the queue pointer Qp1 is set in the target queue as the redistribution information 2721 corresponding to the hash value H4, and the queue pointer Qp1 is set in the target queue as the redistribution information 2722 corresponding to the hash value H5. .

  In the redistribution information 272, the hash values H1 to H3 that are not reconnected are the same as the learning results 271 at the time of reception from the learning unit 203, and the queue pointers Qp1 to Qp3 remain unchanged. These are the same as the relationship between the hash values H1 to H3 and the queue pointers Qp1 to Qp3 in the learning information 400 and the hash value table 222, and it can be said that the initial information remains as it is.

  When the learning result processing unit 207 performs redistribution based on a certain rule, a hash value that is not subject to redistribution is specified by an instruction from the device operator, for example, configuration, and the specified hash value is It may be excluded from certain rules.

  The distribution unit 202 receives the redistribution information 272 from the learning result processing unit 207 and reflects the redistribution information 272 to the queue pointer 302 of the hash value table 222. In particular, the redistribution information 2721 and 2722 is set in the queue pointer 302 of the hash value table 222.

<Example of processing procedure for creating redistribution information 272 by learning result processing unit 207>
FIG. 6 is a flowchart illustrating an example of a processing procedure performed by the learning result processing unit 207. As described with reference to FIG. 5, the learning result processing unit 207 creates the redistribution information 272 from the learning result 271, and thus becomes a flowchart illustrating an example of the creation processing procedure. Here, certain rules are as already described.

  The learning result processing unit 207 receives the learning result 271 from the learning unit 203 (step 601), and sets 0 to the variable i in order to determine the current queue length and flow rate from the young number of the hash value H of the received learning result 271. Is set (step 602), and 1 is added to the variable i (step 603).

  Steps 603 to 612 form a loop, and the variable i is processed from 1 to m, which is the number of hash values H, so that the learning result processing unit 207 determines whether the variable i is less than m + 1. (Step 604), and if it is determined that the variable i is not less than m + 1, since the hash value H1 to the hash value Hm have already been processed, the process proceeds to Step 613 to end the loop.

  Note that steps 602 to 604 are examples of processing for the loop, and the processing is not limited to these examples as long as the hash value H1 to the hash value Hm can be processed.

  If it is determined in step 604 that the variable i is less than m + 1, the learning result processing unit 207 determines whether or not the current queue length of the hash value Hi column in the learning result 271 is 0 (step 605). When it is determined that the length is 0, it is determined whether or not the flow rate of the hash value Hi column in the learning result 271 is equal to or less than a threshold set in a certain rule (step 606).

  If it is determined in step 605 that the current queue is not 0, or if it is determined in step 606 that the flow rate is not less than or equal to the threshold value, the process returns to step 603, 1 is added to the variable i, and the next hash value Hi is processed. It becomes a target.

  Here, the threshold is, for example, 30 as already described. Even if step 606 is not provided, the same processing result can be obtained by step 611. However, in the case of the hash value H2 in the learning result 271 shown in FIG. 5, execution of a loop from step 608 to step 611 described below is performed. Can be omitted in advance.

  When it is determined in step 606 that the flow rate is equal to or less than the threshold value, the hash value Hi is aggregated in the queue pointer Qp that is the target queue of the column of the hash value Hj (j is an integer less than i) that is younger than the hash value Hi. Therefore, the learning result processing unit 207 sets 0 to the variable j (step 608), and adds 1 to the variable j (step 608).

  Steps 608 to 611 form a loop, and the variable j is processed from 1 to the variable i−1 and is terminated. Therefore, the learning result processing unit 207 determines whether the variable j is less than the variable i. (Step 609) When it is determined that the variable j is not less than the variable i, the hash value H1 to the hash value Hi-1 have already been processed, and the hash value Hj to be aggregated is not found, so the process returns to Step 603. .

  Steps 607 to 609 are examples of processing for the loop, and the processing is not limited to these examples as long as processing can be performed from the hash value H1 to the hash value Hi-1.

  When it is determined in step 609 that the variable j is less than the variable i, the learning result processing unit 207 determines whether or not the current queue length of the hash value Hj column in the learning result 271 is 0 (step 610). When it is determined that the queue length is 0, it is determined in the learning result 271 whether the total value of the flow rate of the hash value Hj column and the flow rate of the hash value Hi column is equal to or less than a threshold set in a certain rule. (Step 611).

  If it is determined in step 610 that the current queue is not 0, or if it is determined in step 611 that the total flow rate is not less than or equal to the threshold value, the process returns to step 608, 1 is added to the variable j, and the next hash value Hj Is the target of processing.

  If it is determined in step 611 that the total value of the flow rates is equal to or less than the threshold value, the hash value Hi can be aggregated into the queue pointer Qp that is the target queue in the column of the hash value Hj. The sequence of hash values Hj is copied to the sequence of hash values Hi (step 612).

  Since this copy includes the queue pointer Qp of the target queue, the queue pointer Qp corresponding to the hash value Hj also corresponds to the hash value Hi. In addition, the learning result processing unit 207 sets the total flow rate to the flow rate in the column of the hash value Hj in the learning result 271 in step 612, and includes this total value in the copy.

  Thereby, when the flow rate of the hash value Hj column in step 611 is already the total value (5 + 2 = 7 of the hash value H1 and the hash value H4 shown in FIG. 5), the flow rate of the hash value Hi column is further increased. The added total value (7 + 3 = 10 plus the hash value H5 shown in FIG. 5) can be obtained.

  Further, since the total value is also copied, the hash value Hi is calculated based on the flow rate of the hash value Hj column, even though the hash value Hi is aggregated in the queue pointer Qp that is the target queue of the hash value Hj column. A phenomenon that the flow rate of the column is reduced does not occur, and the hash value Hj cannot be aggregated and the hash value Hi is not aggregated.

  In steps 611 to 612, the learning result 271 is changed. However, the learning result 271 is not limited to this example. The queue pointer Qp to be aggregated and the total flow rate are stored separately from the learning result 271. However, a processing procedure corresponding to that may be used.

  When the learning result processing unit 207 executes step 612, the aggregation related to the hash value Hi is completed. Therefore, the learning result processing unit 207 returns to step 602 and adds 1 to the variable i to set the next hash value Hi as a processing target.

  When it is determined in step 604 that the variable i is not less than m + 1, the learning result processing unit 207 uses the hash value and the target queue of the learning result 271 including the changed target queue as the hash value and the target queue of the redistribution information 272. And finish the process.

<Example of Update Processing Procedure of Redistribution Information 272 by Learning Result Processing Unit 207>
FIG. 7 is a flowchart illustrating an example of a processing procedure performed by the learning result processing unit 207. The redistribution information 272 created by the learning result processing unit 207 described with reference to FIGS. 5 and 6 does not use the queue pointer Qp of the aggregated hash value H, that is, the queue 214.

  Therefore, the flowchart shows a processing procedure for associating another hash value H with the queue 214 that is no longer used and allocating it to the buffer of the flow 260 of the packet 250 converted to the other hash value H. Here, although described as processing of the learning result processing unit 207, the allocating unit 202 may perform processing instead of the learning result processing unit 207.

  When associating another hash value H with the queue 214 that is no longer used, the learning result processing unit 207 sends the redistribution information 272 to the distribution unit 202 after executing the processing procedure of the flowchart shown in FIG. Before, the processing procedure described below is executed. If no other hash value H is associated with the queue 214 that is no longer used, the redistribution information 272 is sent as described with reference to FIG. 6 without executing the processing procedure described below.

  The learning result processing unit 207 acquires a newly assigned hash value Hk (k is an arbitrary integer from 1 to m) (step 701). The hash value Hk is, for example, the hash value H converted from the packet 250 of the newly added flow 260, and since there is no communication and the queue 214 is not linked, the queue pointer (target) shown in FIGS. Queue) is a hash value H7 that is NULL.

  In the example of FIG. 5, the hash value H7 is not shown, but the learning result 271 is the same information as the learning information 400 shown in FIG. 4, and the redistribution information 272 is a copy of the learning result 271. The target queue corresponding to the hash value H7 of the redistribution information 272 is NULL.

  The learning result processing unit 207 sets 0 to each bit of the n-bit bitmap corresponding to the number of queue pointers Qp, that is, the number of queues 214 (step 702). Each bit in this n-bit bitmap represents an unused queue 214 as 0, and step 702 is this initialization.

  Note that hash values H1, H4, and H5 may be associated with the queue 214-1 indicated by the queue pointer Qp1 as in the redistribution information 272 illustrated in FIG. 5, which includes the hash value table 222, the learning information 400, Also, the hash values H1 and H5 remain even if some hash values H4 are aggregated in other queues 214, for example, based on the learning after the reflection and also reflected in the learning result 271.

  For this reason, in step 612 shown in FIG. 6, there is a possibility that the target queue of the hash value Hi is associated with another hash value H, and the queue 214 that is no longer used cannot be specified. The information 272 is scanned to reflect the usage status of the queue 214 in the n-bit bitmap.

  The learning result processing unit 207 sets 0 to the variable i (step 703), adds 1 to the variable i (step 704), and determines whether the variable i is less than m + 1 (step 705). These steps 703 to 705 are the same as steps 602 to 604 shown in FIG.

  If the learning result processing unit 207 determines in step 705 that the variable i is less than m + 1, the learning result processing unit 207 acquires the queue pointer Qpj of the target queue in the column of the hash value Hi in the redistribution information 272, and stores the n-bit bitmap. The j-th bit is set to 1 (step 706), and the process returns to step 704.

  If it is determined in step 705 that the variable i is not less than m + 1, the learning result processing unit 207 reflects the information of the queue pointer Qp of the target queue of the column up to the hash function Hm in the n-bit bitmap. In order to examine the bitmap information, 0 is set to the variable i (step 707), 1 is added to the variable i (step 708), and it is determined whether the variable i is less than n + 1 (step 709).

  If the learning result processing unit 207 determines in step 709 that the variable i is less than n + 1, the learning result processing unit 207 determines whether the i-th bit of the n-bit bitmap is 0 (step 710), and the n-bit bit When it is determined that the i-th bit of the map is 0, the queue pointer Qpi is set in the target queue of the column of the hash value Hk in the redistribution information 272, and the redistribution information 272 is updated (step 711).

  If the learning result processing unit 207 determines in step 710 that the i-th bit of the n-bit bitmap is not 0, the queue pointer Qpi is used, so the process returns to step 708, and in step 709, the variable i is n + 1. If it is determined that it is not less than the queue pointers Qp1 to Qpn, there are no unused queue pointers, so the processing ends.

  When the processing of the flowchart shown in FIG. 7 is completed, as described above, the learning result processing unit 207 sends the updated redistribution information 272 to the distribution unit 202. Here, when it is determined in step 709 that the variable i is not less than n + 1, the queue 214 having a small current queue length and / or flow rate may be linked to the hash value Hk, or the operator may be notified.

  As described above, even if hash values H that are more than the number of queues 214, that is, flows 260 are accommodated, a plurality of hash values H having a small current queue length and flow rate can be aggregated. Since the packet 250 is not discarded, the communication quality can be protected.

  Further, since a new hash value H, that is, a new flow 260 can be accommodated in the queue 214 that is no longer used due to aggregation, it is possible to increase the number of flows 260 that can be accommodated while maintaining communication quality.

  In addition, this invention is not limited to the Example demonstrated above, The various modification within the meaning of a claim and an equivalent structure are included. For example, the above embodiments have been described in detail for easy understanding of the present invention, and the present invention is not necessarily limited to those having all the configurations described.

  In addition, each of the configurations, functions, processing units, processing means, and the like described above may be realized in hardware by designing a part or all of them, for example, with an integrated circuit. It may be realized by software by interpreting and executing a program for realizing the above.

  Information such as programs, tables, and files for realizing each function is recorded on a memory, a hard disk, a storage device such as an SSD (Solid State Drive), an IC (Integrated Circuit) card, an SD card, or a DVD (Digital Versatile Disc). It may be stored on a medium.

  Further, the control lines and the information lines are those that are considered necessary for the explanation, and not all the control lines and the information lines that are necessary for the mounting are shown. In practice, it can be considered that almost all the components are connected to each other.

120 router 201 reception unit 202 distribution unit 203 learning unit 204 buffer 205 bandwidth control unit 206 transmission unit 207 learning result processing unit 214 queue 221 conversion unit 222 hash value table

Claims (6)

A communication relay device that relays packets,
A receiver that receives the packet and outputs the received packet;
A distribution unit that converts packets output from the reception unit into hash values, and distributes the packets output from the reception unit to a plurality of queues based on the correspondence between the hash values and the queues, based on the converted hash values. When,
A buffer for storing packets distributed by the distribution unit in a plurality of queues;
A learning unit that accumulates state information about each of the plurality of queues of the buffer;
Based on the state information regarding each queue accumulated by the learning unit, a learning result processing unit that creates a relationship information corresponding to a plurality of hash values in the queue and sends the information to the allocating unit;
A communication relay device, comprising: a transmission unit that transmits a packet stored in the buffer. The communication relay device according to claim 1,
The distribution unit is
A hash value table for storing the correspondence between hash values and queues;
The packet header output by the receiver is converted into a hash value,
Obtain queue information corresponding to the converted hash value from the hash value table,
A communication relay device, wherein the queue information acquired from the hash value table is sent to the buffer to distribute the packets output by the receiving unit to a plurality of queues. The communication relay device according to claim 2,
The packet received by the receiving unit includes a destination address in a header,
The distribution unit is
A communication relay device that converts the address into a hash value including the destination address. The communication relay device according to claim 3,
The learning unit
A communication relay device that accumulates the queue length of the queue of the buffer and the flow rate of packets distributed by the distribution unit as state information for each of the plurality of queues of the buffer for each hash value . The communication relay device according to claim 4,
The learning result processing unit
Multiple hash values correspond to queues by aggregating queues into hash values converted from packet headers where the queue length is smaller than the preset value and the bucket flow rate is less than the preset value A communication relay device that creates information on a relationship to be sent to the distribution unit. The communication relay device according to claim 5,
The learning result processing unit
A communication relay device, characterized in that information on a relationship in which a queue before aggregation of hash values into which queues are aggregated is associated with other hash values is created and sent to the distribution unit.

Images