JP2013046193A - Congestion detection method and congestion control method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform congestion detection and congestion control by taking process priority into consideration.SOLUTION: A process is classified into one of plural groups which are each assigned priority, and for each one group, an upper limit on the total CPU usage of processes belonging to groups having higher priority than that group is set as a threshold beforehand. Then the total CPU usage of processes belonging to each group is calculated for each group. For each one group, the cumulative value of the total CPU usage of processes belonging to groups having higher priority than that group is calculated. The calculated cumulative value is compared to a threshold set per group. If the cumulative value exceeds the threshold for a group, then the group is determined to be congested. Thereby, congestion detection can be performed taking unused CPU resources into consideration while a certain amount of CPU resources are distributed on the basis of priority of each group.

Description

  The present invention relates to a technique for detecting and controlling congestion in a computer system.

  In a communication system / session control system such as an exchange, when the system receives a call exceeding its processing capacity, it enters a congestion state in which the call connection completion rate decreases. In a congested state, the server's processing capacity declines or the service is degraded or stopped due to server down, so the communication system / session control system has a congestion detection function and a congestion control function that aims to protect the service and system when congestion is detected. It has.

  Since the CPU resources of the system are insufficient at the time of congestion, the allocation of the CPU resources to each process becomes important. If this is not done properly, processing delays and server downtime may occur. In the conventional exchange, CPU resources can be statically divided by time slices using a dedicated OS function, and CPU resources can be appropriately allocated for call processing processes, maintenance processes, and the like.

JP 2005-167769 A

  Conventional congestion control technology is provided as a function of a dedicated OS such as TRON. However, a general-purpose OS such as Linux is not provided. Therefore, when a system is constructed with a general-purpose OS, CPU resources must be allocated by an application during congestion. The general-purpose OS also has a process priority, and has a function of distributing CPU resources according to the priority. However, it is impossible to control CPU resource distribution in consideration of the characteristics and cooperation of each process of the application.

  On the other hand, in the dedicated OS, the CPU resource is statically divided and assigned to each process, so that even if there is room in one of the divided CPU resources, the CPU resource cannot be used by another process. was there.

  The present invention has been made in view of the above, and an object thereof is to perform congestion detection and congestion control in consideration of process priority.

  A congestion detection method according to a first aspect of the present invention is a congestion detection method for classifying processes into a plurality of prioritized groups and detecting congestion for each of the groups. A step of setting the upper limit of the total resource usage rate of processes belonging to a group having a priority level higher than the group as a threshold, and obtaining a cumulative value of the total resource usage rates of processes belonging to a priority level group or higher. And comparing the accumulated value with the threshold value set for the own group, and determining that the own group is congested when the accumulated value exceeds the threshold value.

  A congestion control method according to a second aspect of the present invention is a congestion control method for a process belonging to a group that is classified into a plurality of prioritized groups, detects congestion for each group, and is determined to be congested. For each of the plurality of groups, an upper limit of the total resource usage rate of processes belonging to a group having a priority higher than the self group is set as a threshold, and processes belonging to a group having a priority higher than the self group are set. A step of obtaining a cumulative value obtained by summing up the resource usage rates is compared with the cumulative value and the threshold value set for the self group, and when the cumulative value exceeds the threshold value, the self group is determined to be congested. And a process belonging to the group determined to be congested in the determining step, by a method determined in advance for each process And having the steps of Gosuru, the.

  In the congestion control method, the congestion control step is characterized in that congestion control is performed for a process belonging to a group having a priority lower than that of the group determined to be congestion in the determination step.

  According to the present invention, it is possible to perform congestion detection and congestion control in consideration of process priority.

It is a functional block diagram which shows the structure of the session control server in this Embodiment. It is a flowchart which shows the flow of the operation | movement which detects congestion. It is a figure explaining the detection of congestion concretely.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a functional block diagram showing the configuration of the session control server in the present embodiment. In FIG. 1, only the part regarding the congestion detection and congestion control of the session control server 1 is illustrated. As shown in FIG. 1, the session control server 1 in the present embodiment includes a CPU usage rate calculation unit 11, a congestion determination unit 12, and a congestion control unit 13. Each unit included in the session control server 1 may be configured by a computer including an arithmetic processing device, a storage device, and the like, and the processing of each unit may be executed by a program. This program is stored in a storage device included in the session control server 1, and can be recorded on a recording medium such as a magnetic disk, an optical disk, or a semiconductor memory, or provided through a network. Although the congestion detection method and the congestion control method for the session control server will be described here, the present invention is not limited to this, and can be similarly applied to other application servers.

  The session control server 1 classifies processes (or threads) into a plurality of groups based on priorities from the service viewpoint, not the process priorities managed by the OS, and assigns a priority higher than the own group to each group. The upper limit of the total CPU usage rate of the processes belonging to the group to be held is set as a threshold value. The correspondence between the process and the group and the threshold value of each group are stored in a storage device provided in the session control server 1.

  The CPU usage rate calculation unit 11 sums up the CPU usage rates of processes belonging to the same group, and calculates the CPU usage rate for each group.

  For each group, the congestion determination unit 12 determines the congestion of the group by comparing a cumulative value obtained by totaling the CPU usage rates of the groups having a priority higher than that group with a threshold value indicating the upper limit of the CPU usage rate. Specifically, for each group, a cumulative value obtained by totaling the CPU usage rates of processes belonging to a group having a priority level higher than that of the own group is obtained, the obtained cumulative value is compared with a threshold set for each group, and the cumulative value is obtained. When the value exceeds the threshold, the group is determined to be congested.

The congestion control unit 13 performs congestion control predetermined for each process on processes belonging to the group determined to be congested. Processes subject to congestion control are all processes belonging to a group determined to be congestion. When a plurality of groups are determined to be congested, congestion control is performed for all processes belonging to each group determined to be congested. The congestion control method is set and implemented for each process. For example, the backup process is forced to sleep for a fixed time, the command process is forcibly terminated, and new command input is restricted. Table 1 shows an example of the control method for each process. Note that the process-specific congestion control method shown in Table 1 is an example, and the congestion control method is not limited.

  Next, processing in which the session control server 1 detects congestion will be described.

  FIG. 2 is a flowchart showing a flow of processing in which the session control server 1 detects congestion.

  First, the CPU usage rate calculation unit 11 sums up the CPU usage rates of the processes belonging to each group to obtain the CPU usage rate of each group (step S11).

  Subsequently, for each group, the congestion determination unit 12 sums up the CPU usage rates of the groups having the priority higher than the own group, and obtains a cumulative value for each group (step S12).

  Then, for each group, the congestion determination unit 12 compares the accumulated value obtained in step S12 with a preset threshold value (step S13), and determines a group whose accumulated value exceeds the threshold value as congestion (step S14). . Thereafter, the congestion control unit 13 performs congestion control on processes belonging to the group determined to be congested by a method predetermined for each process.

  Next, detection of congestion will be described using a specific example.

  FIG. 3 is a diagram for specifically explaining the detection of congestion. In the example of FIG. 3, the OS / middle process, “highest” to which the inter-system synchronization process belongs, “high” to which the call processing system belongs, “medium” to which the maintenance command process belongs, The thresholds indicating the upper limit of the CPU usage rate of each group were set to 30%, 70%, 80%, and 85%.

  Assuming that the total CPU usage of processes for each group is 15% for “highest”, 40% for “high”, 26% for “medium”, and 3% for “low”, As shown in FIG. 3, the cumulative value of the CPU usage rate of the priority group higher than the own group is 15% for “high”, 55% for “high”, 81% for “medium”, and 84 for “low”. %. When the cumulative value and the threshold are compared for each group, “medium” has a cumulative value higher than the threshold and is determined to be congested. As a result, congestion control is performed for the maintenance command process belonging to “medium”.

  In the example illustrated in FIG. 3, “low” is not determined to be congestion because the cumulative value is lower than the threshold, but processes belonging to a group having a lower priority than the group determined to be congestion are also subject to congestion control. The regular backup process belonging to “” may be subjected to congestion control.

  As described above, according to the present embodiment, processes are classified into a plurality of prioritized groups, and for each group, the CPU usage rate of a process belonging to a group having a priority higher than the own group. The upper limit of the total is set as a threshold, the CPU usage rate of the processes belonging to the group is obtained for each group, and the CPU usage rates of the processes belonging to the group having the priority higher than the own group are totaled for each group. By comparing the calculated cumulative value with the threshold value set for each group and determining that the group is congested when the cumulative value exceeds the threshold value, the priority of each group is obtained. Based on this, it is possible to perform congestion detection in consideration of unused CPU resources while allocating certain CPU resources.

  According to the present embodiment, the congestion control is performed by a method set in advance for each process belonging to the group determined to be the congestion, so that the congestion control can be performed finely and with little waste.

1 Session control server 1
DESCRIPTION OF SYMBOLS 11 ... CPU usage rate calculation part 12 ... Congestion determination part 13 ... Congestion control part

Claims (3)

A congestion detection method for classifying processes into a plurality of prioritized groups and detecting congestion for each of the groups,
In each of the plurality of groups, an upper limit of the total resource usage rate of processes belonging to a group having a priority higher than the own group is set as a threshold,
A step of obtaining a cumulative value obtained by summing up resource usage rates of processes belonging to a group having a priority higher than the own group;
Comparing the cumulative value with the threshold value set for the self group, and determining the self group as congestion when the cumulative value exceeds the threshold value;
A congestion detection method characterized by comprising: A method for classifying processes into a plurality of prioritized groups, detecting congestion for each group, and a congestion control method for a process belonging to a group determined to be congestion,
In each of the plurality of groups, an upper limit of the total resource usage rate of processes belonging to a group having a priority higher than the own group is set as a threshold,
A step of obtaining a cumulative value obtained by summing up resource usage rates of processes belonging to a group having a priority higher than the own group;
Comparing the cumulative value with the threshold value set for the self group, and determining the self group as congestion when the cumulative value exceeds the threshold value;
Congestion control for a process belonging to the group determined to be congested in the determining step by a method predetermined for each process;
The congestion control method characterized by having.   3. The congestion control method according to claim 2, wherein the congestion control step performs congestion control on a process belonging to a group having a priority equal to or lower than the priority of the group determined as the congestion in the determination step.

Images