JP2003316587A - Load distribution system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To execute load distribution for assigning a resident process in accordance with a daily time zone based on process information by a multi-process computer connected to a plurality of client computers connected via a communication network. <P>SOLUTION: A resident process in accordance with an optimal daily time zone is preliminarily defined from daily execution process information to execute accurate load distribution. Thus, it is possible to improve the use efficiency of a computer. <P>COPYRIGHT: (C)2004,JPO

Description

Description: BACKGROUND OF THE INVENTION [0001] 1. Field of the Invention [0002] The present invention relates to an apparatus for executing a process in a computer system via a communication network. 2. Description of the Related Art Conventionally, load distribution is performed by monitoring and managing the number of jobs using a CPU usage rate as disclosed in, for example, Japanese Patent Application Laid-Open No. 11-3323. There is something. [0003] The above-mentioned prior art controls the number of jobs (multiplicity) using the CPU utilization to perform load distribution. It is an object of the present invention to provide a system in which operation statistical information is aggregated for each business and the number of processes (multiplicity) is controlled for each business in accordance with the load for each date and time using the totaled result, thereby distributing the load. . [0004] The load distribution arithmetic unit 101 includes an operation statistics acquisition unit 102, a use condition information acquisition unit 105, an operation statistics aggregation unit 106, a process number determination unit 107, and a process activation unit 108.
And the schedule table management unit 109, which aggregates the execution program operation statistical information for each day-to-day operation, and determines the number of parallel processes (multiplicity) of the process for each operation in accordance with the optimum daily time from the information. Things. As a result, accurate load distribution can be performed, and efficiency of computer use can be improved. An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 shows an embodiment of the present invention. In FIG. 1, a load balancing arithmetic device 101 includes a business program 111 and a DB.
It has a connection 113 and is connected to one or more client computers 112 via a network 110. [0007] The load balancing arithmetic unit 101 compiles the operation program operation statistical information for each day's work, and determines the multiplicity of the process according to the time period of each day from the information. And usage condition information acquisition unit 1
05, an operation statistics totalizing unit 106, a process number determining unit 107, a process starting unit 108, and a schedule table managing unit 109. [0008] Here, a process is an instance of a program executed by an operating system. The operation statistics acquisition unit 102 acquires a program name, a processing request occurrence time, a start time, and an end time of an execution process for each currently used task and stores them in the DB 113. The process information acquisition unit 103 and the time measurement unit Consists of 104. This flow is shown in FIG. A use condition information acquisition unit 105 acquires a target average response time for each task and each date and time defined and set in advance by a user of the present system.
To pass that information on. The operation statistics totalizing unit 106 obtains the operation statistical information from the DB 113, totals the average processing time for each task and for each date and time, and passes the result to the schedule table management unit 109. In the method of counting the average processing time, the difference between the end time and the start time is used as the processing time, the processing time for each task is integrated for each date and time, and the result is divided by the number of members to obtain the average processing time. In addition, the method of counting the average processing request generation intervals is to sort the processing request generation time for each task, count the number of data for each date and time, and use it as the average processing request generation number. The number-of-processes determining unit 107 determines the multiplicity of a process that satisfies the average response time target value condition set in advance by the user based on the average occurrence time and the average processing time for each task and date and time acquired from the schedule table management unit 109. And passes the information to the schedule table management unit 109. This flow is shown in FIG. The process starting unit 108 starts instances of the business program 111 from the schedule table management unit 109 for the number of multiplicity of the process determined for each business and for each date and time. Schedule table management unit 109
Is created by the use condition information acquisition unit 105 and the number-of-processes determination unit 107 for each task and each date. Details of the schedule table management unit 109 are shown in FIG. The DB 113 stores the information of the business program name, the processing request generation time, the start time, and the end time obtained by the operation statistics acquisition unit 102. The details are shown in FIG. FIG. 2 is a flowchart of the operation statistics acquisition unit 102. The operation statistics acquisition unit 102 acquires a processing request generation time (201) upon arrival of a processing request from a client, and acquires a business program name (202). Thereafter, the start time of the business program is obtained (203), the program is executed, and the end time of the business program is obtained (204) and stored in the DB 113 (205). FIG. 3 is a flowchart of the process number determination unit 107. Obtain the average processing time and the average number of processing requests generated for each task and date from the schedule table management unit 109 (301)
Then, a process number determination calculation (302) is performed. In the calculation for determining the number of processes, a generally known process is used. For example, there is queuing theory. (New Version Information Processing Handbook P46) (Priority Measures for Afternoon Problems of Software Development Engineer 2002 P249) The resulting multiplicity of the process for each task is passed to the schedule table management unit 109 (30
3). As an example, the schedule table management unit 109
From the average number of processing request occurrences per business and date and time (404),
Obtain the average processing time (405) and calculate the processing rate.Based on the simplified calculation chart (M / M / m type), the number of service counters within the target Determined as severe. FIG. 4 shows a schedule table management table. Usage condition information acquisition unit 105, operation statistics aggregation unit 10
6.Aggregation unit (401), business program name (402), average response time (403), average processing request occurrence count (404), average processing time (401) Four
05) and the number of processes (406) are stored. FIG. 5 shows a business program information table. Business program name (501), processing request generation time (502), start time for each business and date and time obtained from the operation statistics acquisition unit
(503) and the end time (504). In this way, the number of accepted processes is obtained from the past totaling result, child processes for the number of accepted processes are generated in advance, and the space on the memory is secured, so that the load distribution can be efficiently performed. It is possible to do. According to the present invention, load distribution is performed by totalizing operation statistical information for each business and controlling the number of processes (multiplicity) according to the load for each date and time for each business using the totaled result. As a result, it is possible to realize an efficient load sharing system that realizes efficiency between computers and is flexible.

[Brief description of the drawings] FIG. 1 shows an embodiment of a load distribution system 1 FIG. 2 is a flowchart of an operation statistics acquisition unit 102; FIG. 3 is a flowchart of a process number determination unit 107; FIG. 4 is a schedule table. FIG. 5 is a business program information table. [Explanation of symbols] 101 load balancing arithmetic unit 102 Operation statistics acquisition unit 103 Process information acquisition unit 104 Time measurement unit 105 Usage condition information acquisition unit 106 Operation statistics aggregation section 107 Process number determination unit 108 Process starter 109 Schedule table management unit 110 network 111 Business Program 112 client computer 113 Database

Claims (1)

Claims: 1. A process operation device of a multi-process computer connected to one or a plurality of client computers via a network, and collects operation statistical information for each day's operation of the system and tabulates it. Based on the result and the usage condition information defined and set by the user in advance, the load distribution calculation device performs load distribution in accordance with a daily time zone.