JP5743334B2 - Congestion control device - Google Patents

Description

  The present invention relates to a technique for recovering / recovering from a congestion state.

  In many information processing systems including communication systems, if an excessive input than expected is added for some reason, the processing will not be completed, and additional processing will be added. The event is known. In general, such an event is called congestion.

  In any system, if there is an input that exceeds the processing capacity of the system, it is obvious that the process will not be completed and it will be in a congested state, but temporarily an excessive input occurs, When the processing capacity of the system is lowered, there is a problem that the congestion state where the processing capacity of the system is lowered continues for a long time even when the input from the outside returns to a steady state.

  For this reason, when such a congestion state occurs or is likely to occur, the input amount to be processed is limited, and the congestion state is quickly restored to the steady state so that a serious problem does not occur. Control (congestion control) is required to prevent the congestion state from occurring.

  In general, congestion is not only a problem that the processing capacity of the system is lowered, but also various unexpected events are complicatedly intertwined, and it can sometimes be a fatal obstacle. For this reason, it is expected that such a congestion does not occur, and that even if it occurs, it returns to the steady state as soon as possible.

  In particular, in recent years, there has been an increasing number of cases where a single system provides multiple services at the same time. Congestion caused by any of the services may affect the entire system. Is required.

  In the conventional congestion control, in order to prevent the processing request to the system from exceeding the processing capacity of the system, the generation status of the processing request to the system or the status of the request processing in the system is monitored, and the system overloads the system. When load is likely to be applied, the degree is divided into several levels, and the processing capacity related to the processing is restricted (regulated) according to the degree of congestion, thereby securing processing capacity and congestion. The situation has been prevented and recovery from the congested state has been carried out.

  For example, in Patent Document 1, regarding the restriction (regulation) of a processing request for recovery from congestion, it is assumed that the processing request that may cause congestion should be regulated, and the processing request is transmitted as much as possible. It shows how to specify the range so that it can be regulated at a location close to the original.

  In addition, various methods have been proposed for monitoring processing request inputs for congestion control, such as direct monitoring of the amount of processing request inputs per hour or indirect monitoring based on CPU usage. By carrying out appropriate regulation according to the degree of congestion, early recovery from congestion has been attempted.

  Such monitoring / congestion control may be performed collectively for the entire system, or may be performed for a part of the system, for example, for each processing module corresponding to a specific service.

  In addition, there is an expectation that more precise congestion control can be performed if more detailed determination is performed for the congestion state determination level, which is a premise for determining regulation control to be activated as congestion control. At the same time, it also increases the cost required for this congestion control, and it has been known that the effect is limited even if the number of steps is increased more than necessary.

JP 2010-200029 A

  The above-described conventional congestion control method is implemented by monitoring the processing status that directly causes congestion and regulating the input of processing requests to the system according to the processing status.

  However, once a congestion state occurs, the scope of its influence spreads in many ways, and the indirect factors that result from this influence hinder recovery from congestion. I did not.

  The present invention has been made in view of the above, and an object of the present invention is to efficiently realize return / recovery from a congestion state to a steady state.

The congestion control apparatus according to claim 1, an accepting unit that accumulates processing requests for services accepted by a computer for each different service, an execution unit that accepts the processing requests for each different service, and executes service processing; measuring means for measuring the duration of the congestion state should regulate generated within the computer reads out the specified time for each of the service from the storing means, the congestion state monitoring for each of the service, the congestion state is the to continue prescribed time or more, to regulate the reinforcing means to enhance the control level for the processing of the service according to the congestion state, the processing of the service according to the congestion state by using the control information of the enhanced control level has a further, and regulating means for regulating the process in the same computer other than the processing of the service according to the congestion state And wherein the door.

According to the present invention, in order to reinforce the restriction level for the processing related to the congestion state when the duration of the congestion state to be restricted occurring in the computer is measured and the congestion state continues for a predetermined time or more, It is possible to efficiently realize the return / recovery from the state to the steady state. In addition, since the processing in the same computer other than the processing related to the congestion state is further restricted, the return / recovery from the congestion state to the steady state can be more efficiently realized.

  According to the present invention, recovery / recovery from a congestion state to a steady state can be efficiently realized.

It is a figure which shows the functional block structure of the conventional congestion control apparatus. It is a figure which shows the functional block structure of the congestion control apparatus which concerns on 1st Embodiment. It is a figure which shows operation | movement of the congestion control apparatus which concerns on 1st Embodiment. It is a figure which shows the functional block structure of the conventional congestion control apparatus. It is a figure which shows the functional block structure of the congestion control apparatus which concerns on 2nd Embodiment. It is a figure which shows operation | movement of the congestion control apparatus which concerns on 2nd Embodiment.

  In the case of a computer such as an information processing system in which congestion continues even if processing of a service processing request or the like that directly causes congestion is restricted, the congestion may be caused by a direct or indirect cause caused by the congestion. By determining that it is continuing and by strengthening restrictions on the same process for its direct cause, and / or by regulating another process in the same computer for its indirect cause The object is to efficiently realize the return / recovery from the congestion state to the steady state.

[First Embodiment]
First, a general congestion control method according to the conventional method will be described. A conventional congestion control device 100 operates on an information processing system as shown in FIG. 1, for example, and includes a service processing request receiving unit 11, a service processing execution unit 12, a system monitoring unit 13, and a restriction threshold table 14. The congestion control unit 15 is mainly configured.

  The service process request accepting unit 11 accepts a service process request transmitted from a service request source such as an information processing terminal, and the service process executing unit 12 executes a service process based on the service process request.

  Thereafter, the system monitoring unit 13 measures the CPU usage rate related to the service processing being executed and reports the CPU usage rate to the congestion control unit 15, and the congestion control unit 15 reports the CPU usage rate reported for the service processing. The measured value is compared with the regulation threshold value in the regulation threshold value table 14, and when the regulation threshold value is exceeded, the regulation level to be implemented is determined by recognizing the occurrence of congestion to be regulated. Based on this, reception of service processing requests is restricted.

  On the other hand, the functional block configuration of the congestion control apparatus 100 according to the present embodiment is shown in FIG. In addition to the conventional congestion control device 100 shown in FIG. 1, the congestion control device 100 sets the duration of the congestion state to be regulated to the regulation duration (specified time) in order to further strengthen the regulation for acceptance of service processing requests. The regulation duration table 16 stored as is further provided.

  Hereinafter, the operation of the congestion control apparatus 100 according to the present embodiment will be described with reference to FIG. The restriction threshold value table 14 stores restriction threshold values for starting restriction on the CPU usage rate related to the service process being executed. In addition, the congestion control unit 15 holds in advance a correspondence relationship table (storage means) that defines the correspondence relationship between the multi-level restriction levels and the restriction information executed at each restriction level.

  First, the service processing request reception unit 11 receives a service processing request transmitted from a service request source via a communication network, and accumulates request data related to the request in a reception queue (step S101). For example, a plurality of web search requests for acquiring URLs of web pages corresponding to the search query are sequentially received, and the plurality of request data are sequentially stored in the received order.

  Next, the service process execution unit 12 acquires the request data from the reception queue, and executes the service process based on the service process request in the request data (step S102). In the case of the above example, request data related to the web search request is sequentially acquired, and a process of searching the web page hitting the search query from the Internet and acquiring the URL is repeatedly executed.

  Next, the system monitoring unit 13 continuously measures the CPU usage rate in the information processing system related to the service processing executed in step S102 in time, and notifies the congestion control unit 15 whenever necessary. (Step S103). In the case of the above example, the CPU usage rate used at the time of web page search is measured.

  Next, the congestion control unit 15 reads out the regulation threshold determined for the service process from the regulation threshold table 14, and the measured value of the CPU usage rate related to the service process notified from the system monitoring unit 13 is It is determined whether the regulation threshold is exceeded (Yes) or not (No) (step S104).

  If the determination result in step S104 is Yes, it recognizes that congestion that should be regulated occurs in the information processing system, and executes a regulation process described later. On the other hand, if the determination result is No, it is recognized that congestion has not occurred, and the process ends without executing any restriction process.

  Subsequently, since the measurement value of the CPU usage rate related to the service processing exceeds the regulation threshold value, the congestion control unit 15 determines, for example, based on the magnitude of the difference between the measurement value of the CPU usage rate and the regulation threshold value. A restriction level to be executed is determined (step S105), and reception of a service process request is restricted based on restriction information of the determined restriction level (step S106). For example, the acceptance of a new service processing request is rejected for a certain period of time.

  Thereafter, the congestion control unit 15 starts measuring the duration after the measured value of the CPU usage rate exceeds the regulation threshold (the duration in a state where the CPU usage rate does not fall below the regulation threshold halfway) (Step S107). A timer is set (step S108).

  Then, after the measurement timer expires (step S109), similarly to step S104, it is determined again whether the measured value of the CPU usage rate related to the service process exceeds the restriction threshold (Yes) or not (No). (Step S110). For example, when the measurement time timer is set to 1 second, the re-determination is executed after 1 second has elapsed from step S108.

  Here, if the result of the redetermination in step S110 is No, the congestion state has been eliminated, so the measurement of the duration started in step S107 is terminated (step S111), and the process is terminated. A process for lowering the current restriction level may be executed after the end of the measurement of the duration time and before the end of the process.

  On the other hand, if the result of the redetermination in step S110 is Yes, the congestion state continues, so the regulation duration set for the service process is read from the regulation duration table 16 and in step S107. It is determined whether or not the measured value of the continuation time when the measurement is started exceeds the regulation continuation time (Yes) or not (No) (step S112).

  If the determination result in step S112 is No, the process returns to step S108 and steps S108 to S110 are repeated. For example, when the regulation duration is set to 10 seconds, if the measured value of the duration does not exceed 10 seconds, the measurement time timer is set again, and the redetermination in step S110 is executed again.

  On the other hand, if the determination result in step S112 is Yes, it is determined that congestion is continuing due to the service processing request, the measurement of the duration time started in step S107 is terminated (step S113), and the current regulation is The level is increased by n stages (n: natural number) (step S114), and the restriction on acceptance of service processing requests is strengthened based on the restriction level regulation information increased here (step S115). For example, the acceptance of a new service processing request is rejected for a time longer than the predetermined time. It should be noted that after executing the process of step S115, the process may return to step S107.

  As described above, according to the present embodiment, the congestion control unit 15 of the congestion control device 100 operating in the information processing system measures the duration of the congestion state to be regulated that has occurred in the information processing system, and the congestion When the state continues for more than the restriction duration, the restriction level for accepting service processing requests related to the congestion state is strengthened, so the time required for returning from the congestion state to the steady state and recovery is shortened. Recovery and recovery can be realized quickly and efficiently.

[Second Embodiment]
First, a general congestion control method according to the conventional method will be described. As shown in FIG. 4, the conventional congestion control device 100 includes the same functional units 11 to 15 as in FIG. 1. However, in order to execute service processes corresponding to the two services A and B, the service process request receiving unit 11 includes a service A process request receiving unit 11a and a service B process request receiving unit 11b. Consists of a service A process execution unit 12a and a service B process execution unit 12b.

  The service A process request receiving unit 11a receives a service A process request transmitted from a service request source, and the service A process executing unit 12a executes a service A process based on the service A process request. On the other hand, the service B process request receiving unit 11b receives a service B process request transmitted from another service request source, and the service B process executing unit 12b executes a service B process based on the service B process request.

  Here, the system monitoring unit 13 sets only the service A process by the service A process execution unit 12a as a monitoring target, and if the CPU usage rate related to the service A process exceeds the regulation threshold, the system monitoring unit 13 is the monitoring target. In order to regulate the processing related to the service A processing, only acceptance of the service A processing request accepted by the service A processing request acceptance unit 11a is restricted.

  That is, when congestion occurs only in the service A process, conventionally, only reception of the service A process request is subject to restriction, and reception of other service B process requests is not subject to restriction.

  On the other hand, in the present embodiment, when a plurality of services coexist in the information processing system, the restriction is performed on another service that is not directly regulated. In the first embodiment, in order to quickly recover the congestion that has occurred in the information processing system, the regulation on the reception of the service A processing request that is the direct cause of the congestion has been strengthened. In the second embodiment, Restriction is also applied to reception of service B processing requests that are not directly related to the congestion.

  FIG. 5 shows a functional block configuration of the congestion control apparatus 100 according to the present embodiment. This congestion control device 100 is composed of the same functional units 11 to 16 as in FIG. However, the service process request accepting unit 11 includes a service A process request accepting unit 11a and a service B process request accepting unit 11b, and the service process executing unit 12 includes a service A process executing unit 12a and a service B process executing unit 12b. ing.

  Hereinafter, the operation of the congestion control apparatus 100 according to the present embodiment will be described with reference to FIG.

  First, the service A process request accepting unit 11a accepts a service process request transmitted from a service request source via a communication network, accumulates request data related to the request in its own accept queue, and concurrently executes it. The service B process request accepting unit 11b accepts a service B process request transmitted from another service request source, and accumulates request data related to the request in its own accept queue (step S201).

  Next, the service A process execution unit 12a acquires the request data from its own reception queue, executes the service A process based on the service A process request in the request data, and concurrently executes the service A process. The execution unit 12b acquires the request data from its own reception queue, and executes the service B process based on the service B process request in the request data (step S202).

  Next, the system monitoring unit 13 continuously measures the CPU usage rate in the information processing system related to the service A process executed in step S202, and notifies the congestion control unit 15 whenever necessary. (Step S203).

  Next, the congestion control unit 15 reads out the regulation threshold determined for the service A process from the regulation threshold table 14, and the measured value of the CPU usage rate related to the service A process notified from the system monitoring unit 13 (Yes) or not (No) is determined (step S204).

  If the determination result in step S204 is Yes, it recognizes that congestion that should be regulated occurs in the information processing system, and executes a regulation process described later. On the other hand, if the determination result is No, it is recognized that congestion has not occurred, and the process ends without executing any restriction process.

  Next, since the measured value of the CPU usage rate related to the service A process exceeds the regulation threshold value, the congestion control unit 15 is based on the magnitude of the difference between the measured value of the CPU usage rate and the regulation threshold value, for example. Then, a restriction level to be executed for reception of the service A processing request is determined (step S205), and reception of the service A processing request is restricted based on the restriction information of the determined restriction level (step S206). For example, acceptance of a new service processing request is rejected.

  Thereafter, the congestion control unit 15 starts measuring the duration after the measured value of the CPU usage rate exceeds the regulation threshold (the duration in a state where the CPU usage rate does not fall below the regulation threshold halfway) (step S207). A timer is set (step S208).

  Then, after the measurement timer expires (step S209), similarly to step S204, it is determined again whether the measured value of the CPU usage rate related to the service A process exceeds the restriction threshold (Yes) or not (No). (Step S210). For example, when the measurement time timer is set to 1 second, the re-determination is executed after 1 second has elapsed from step S208.

  Here, if the result of the redetermination in step S210 is No, the congestion state has been eliminated, so the measurement of the duration time started in step S207 is terminated (step S211), and the process is terminated. After the measurement of the duration time and before the end of the process, a process for lowering the restriction level for accepting the current service A process request may be executed.

  On the other hand, if the result of the redetermination in step S210 is Yes, the congestion state continues, so the regulation continuation time determined for the service A process is read from the regulation duration table 16 and step S207 is performed. In step S212, it is determined whether or not the measured value of the duration time at which the measurement is started exceeds the regulation duration time (Yes) or not (No). You may make it determine whether the congestion deterioration by the service A process is detected.

  If the determination result in step S212 is No, the process returns to step S208 and steps S208 to S210 are repeated. For example, when the regulation duration is set to 10 seconds, if the measured value of the duration does not exceed 10 seconds, the measurement time timer is set again, and the redetermination in step S210 is executed again.

  On the other hand, if the determination result in step S212 is Yes, recovery from the congestion state can be sent as it is, and the congestion state may not converge. Therefore, the measurement of the duration started in step S207 ends ( In step S213), the restriction process is also executed for processes other than the reception of the service A process request.

  Here, the restriction level to be executed for the reception of the service B processing request is determined (it can be the same level as or different from the level determined in step S205) (step S214), and the restriction information of the determined restriction level is added. Based on this, reception of the service B processing request is also restricted (step S215). For example, all the request data accumulated in the reception queue of the service B processing request reception unit 11b is discarded. It should be noted that after executing the process of step S215, the process may return to step S207.

  As described above, according to the present embodiment, since the reception of service B processing requests other than the reception of service A processing requests related to the congestion state is further regulated, the return / recovery from the congestion state to the steady state can be performed more efficiently and It can be realized more quickly.

  Finally, in each embodiment, the CPU usage rate has been described as an example of the criteria for determining whether regulation is necessary. However, other criteria, such as the memory usage and the memory usage rate necessary for executing the service processing request, are used. Also good.

  Further, in each embodiment, the reception of a service processing request has been described as an example of a restriction target. However, other processes such as whether or not to allocate memory necessary for executing a service processing request and a reduction in the allocation rate are also subject to restriction. Good.

  Further, the processes described in the embodiments may be combined. For example, when the determination result in step S212 is Yes, the restriction level for accepting the service A processing request may be strengthened.

  In addition, the services described in each embodiment can be divided into meaningful units regardless of physical units (for example, CPU, LAN port) and logical units (for example, applications and processes). is there.

  Further, in each embodiment, the information processing system has been described as an example where congestion occurs. However, a single information processing apparatus may be used, and a computer including the information processing system and the information processing apparatus is widely used. Is targeted.

  Further, the congestion control device 100 described in each embodiment can be realized by a computer having a storage unit such as a memory and a calculation / control unit such as a CPU. The operation of the congestion control device 100 can be executed by a program.

DESCRIPTION OF SYMBOLS 100 ... Congestion control apparatus 11 ... Service process request reception part 11a ... Service A process request reception part 11b ... Service B process request reception part 12 ... Service process execution part 12a ... Service A process execution part 12b ... Service B process execution part 13 ... System monitoring unit 14 ... regulation threshold table 15 ... congestion control unit (continuation means, strengthening means, regulation means)
16 ... Regulation duration table (storage means)
S101-S115, S201-S215 ... step

Claims (1)

Receiving means for storing processing requests of services received by the computer for each different service;
Execution means for accepting the processing request for each different service and executing processing of the service;
Measuring means for measuring the duration of the congestion state should regulate generated within the computer,
The specified time for each service is read from the storage means, the congestion state is monitored for each service , and if the congestion state continues for the specified time or longer, the restriction level for processing of the service related to the congestion state is strengthened Strengthening means to
Restriction means for restricting processing of the service related to the congestion state using restriction information of the enhanced restriction level, and further restricting processing in the same computer other than processing of the service related to the congestion state;
A congestion control apparatus comprising:

Images