JPH05128077A - Method for distributed processing plural computer system - Google Patents

Abstract

PURPOSE:To obtain a response with a stable processing time even when the execution processing time of processing is short and to efficiently utilize plural computer systems by automatically changing a computer for executing requested processing in accordance with the execution quantity of the processing. CONSTITUTION:Processing schedulers 70 in respective computers 1 to 3 allocate requested processing to the computer with the highest execution priority order out of the plural computers 1 to 3 by referring to a processing executing computer priority order table 140 and a processing execution restricting time table 150 arranged in a shared memory 110 or a shared direct storage device 120 to be accessed from respective computers 1 to 3. When the processing time exceeds the execution restricting time of the computer concerned, the processing is temporarily interrupted and then succeeded by the computer with the succeeding execution order to successively level down the computers for executing the processing. Since the execution priority order is set up in accordance with the computing capacity of respective computers 1 to 3, a response with the shortest time can be obtained for processing with a short execution time.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a distributed processing method in a plurality of computer systems connected via a shared memory or a direct storage device, and particularly, the influence of a processing requiring a long execution time is minimized and other The present invention relates to a distributed processing method capable of stabilizing a processing response.

[0002]

2. Description of the Related Art Conventionally, as a distributed processing method of a plurality of computer systems in which independent computers are connected to each other, for example,
There is a method described in Japanese Patent Publication No. 2-30534.
In this method, the processing allocation device searches the issued processing request for a computer that satisfies the condition that the programs and files necessary for executing the processing can be accessed,
Among the computers that meet the conditions, the computer that minimizes the sum of the communication delay time and the expected value of the processing waiting time in the computer,
The method of allocating the process was used.

[0003]

As described above, in the conventional method of allocating processing in a multi-computer system, a computer having a minimum sum of an expected value of waiting time in a computer and an expected value of communication delay time is used. Process was assigned. However, in this method, when the processing that requires a large amount of calculation time is continuously executed, all the computers have a high load, and therefore even if the subsequent processing request has a small processing amount. However, there is a problem that the time until the end of processing becomes long. An object of the present invention is to solve such a conventional problem, and to realize efficient operation even when processing that requires a large amount of calculation time is continuously executed. To provide a method.

[0004]

In order to achieve the above object, a distributed processing method for a multiple computer system according to the present invention comprises a plurality of independent computers and a plurality of computers connected by a memory or a direct storage device shared by the computers. In the computer system, in order of the processing capacity of the computers, the execution priority of the processing is given to each computer, the execution time limit of the processing is set for each computer, and there is a processing request,
First, the requested process is executed by the computer with the highest priority, and the execution time of the process is monitored by the timer of the execution status monitoring device. If the result of the monitoring indicates that the execution time limit has been exceeded, the process on the computer is interrupted. Then, the processing is handed over to the computer with the next highest priority, and the processing is continued. Similarly, when the execution time limit for each computer is exceeded, the computer with the lower priority takes over. is there.

[0005]

According to the present invention, the processing execution computer priority table and the processing execution time limit table are stored in the shared memory or the direct storage device which can be accessed from each independent computer in the multiple computer system. Is set up, and the execution sequence of the processing and the time limit are described in these. As for the execution time limit in the computer, the process executed in the computer is divided for each processing class, and the execution time limit is set on the table for each processing class. Each computer is provided with a process execution state monitoring mechanism, and a timer in the computer is used to measure the execution time of each process. The process is initially assigned to the computer with the highest execution order and executed by that computer. However, if the execution time of the actual process exceeds the preset execution time limit of the process class, this computer is executed. Then, the continuation of the process in step 1 is interrupted and the process is continued by the computer having the next execution priority. As a result, a process having a short execution time is not hindered from being executed by a process having a long execution time, and thus a process having a short execution time can obtain a stable response of the processing time.

[0006]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a diagram showing a configuration example of a multi-computer system that executes the distributed processing method of the present invention. In FIG. 1, reference numerals 1, 2 and 3 are independent computers that form a plurality of computer systems, 110 is a shared memory shared by the computers 1, 2, and 3, and 120 is shared by the computers 1, 2, and 3. The shared direct storage devices 21, 22, 23 are direct storage devices directly connected to the computers 1, 2, 3 respectively, and 31, 32, 33 are terminal systems connected to the computers 1, 2, 3. The computers 1, 2 and 3 have the same architecture, and the computing capacity is highest in the computer 1 and lower in the order of the computers 2 and 3. The direct storage devices 21, 22, and 23 have programs and data executed by the computers 1, 2, and 3. Further, each computer 1, 2, 3 is provided with a main memory 40, a timer 50, and a buffer memory 60 for temporarily storing a processing request, and a scheduler 70 for allocating the processing request to each computer 1, 2, 3. And process execution monitor 8 that monitors the processing status
0 is provided. The respective computers 1, 2, 3 are connected via a channel 90 or a communication line, and the shared memory 110 and the shared direct storage device 120 can be accessed from the respective computers 1, 2, 3. The shared memory 110 is provided with a process execution computer priority table 140 that defines the priority order of the computers that execute processes, and a process execution time limit table 150 that defines the process execution time limit of each computer. It has been.

FIG. 2 is a diagram showing a configuration example of the processing execution computer priority table in FIG. This table 1
In a field 40, the order of computers that execute processing in the multiple computer system and the identification information of the computers are registered.
There is a rudder. Then, in this embodiment, the computing ability of the computer is used as the priority of execution of the computer.
That is, the computer identification ids are set on the processing execution computer priority table 140 in the order of the execution priority 1 to the computers 1, 2 and 3 having the highest calculation capabilities. Figure 3
It is a figure which shows the structural example of the process execution time limit table in FIG. In this table 150, a file indicating the execution time limit of each computer from the execution order 1 of the computers to the order 3 which is the number of computers in order corresponding to the processing execution class.
There is a rudder. For example, the process execution class A,
For execution order 1 for each of B and C, a1,
b and c1 are a2 and b, respectively, for execution order 2.
2 and c2 are a3 and b, respectively, for execution order 3.
3, c3 are specified. As the execution time limit, a numerical value from 0 to the maximum value can be described in microsecond units.

FIG. 4 is an operation flowchart of the distributed processing method of the computer system according to the present invention. Now, in the multi-computer system shown in FIG. 1, the process execution computer order table 140 is set to the content shown in FIG. 2, and the process execution time limit table 150 is set to the content shown in FIG. And It is assumed that the processing request T is issued from the terminal system 32 connected to the computer 2 (step 101). The processing request T is first stored in the buffer memory 60 of the computer 2 to which the terminal system 32 is connected. The processing scheduler 70 is a buffer memory 60.
The content of the highest priority is read by reading the contents from the one with the highest processing request priority, and referring to the processing execution class of the processing and the processing execution computer priority table 140 arranged in the shared memory 110. It is known that the computer with the highest number is computer 1 (step 102). Further, it is assumed that the processing execution class 131 of this processing request T is B. By referring to the process execution time limit table 150, the process scheduler 70 knows that the execution time limit of the process class is b1 microseconds (step 103). The scheduler 70 allocates the process T to the computer 1 and executes it (step 1
04, 105), the processing execution state monitoring mechanism 80 of the computer 1
Request to monitor the execution status of the process. If the process T is a job that requires a short execution time,
Since the computer 1 has the highest computing capacity in the multi-computer system, the processing is completed in a time close to the shortest processing time expected for execution in the system (step 107).

In this case, the scheduler 70 recognizes the end of the processing T and reports the processing result to the terminal 32 which is the processing request source (step 119). On the other hand, if the process T is a job that requires a large amount of execution processing time, the process execution state monitoring mechanism 80 uses the timer 50 to monitor the execution state of the process T (step 106), and if the process T ends. If the processing time previously exceeded b1 microseconds (step 10
8), report this to Schedule 70. Schedule
LA 70 interrupts the execution of the process T in the computer 1,
Referring again to the processing execution computer priority table 140,
Next, the process T is assigned to the computer 2 having the highest priority (step 109). Then, the process T is executed (step 110) and the process execution state monitoring mechanism 80 of the computer 2 is requested to monitor the process execution state. The processing execution state monitoring mechanism 80 of the computer 2 monitors the execution state of the processing T using the timer 50 (step 111). If the processing T ends the processing within b2 microseconds (steps 112 and 113), the scheduler 70 recognizes that and reports this to the terminal 32 that is the processing request source (step 119). .. If the processing time exceeds b2 microseconds before the processing T ends, the computer 2
The processing execution state monitoring mechanism 80 of (1) reports to the scheduler 70.

The scheduler 70 is a processing T in the computer 2.
Execution is suspended, the processing execution computer priority table 140 is again referenced, and the processing T is assigned to the computer 3 having the next highest priority (step 114). And
The process T is executed (step 115) and the process execution state monitoring mechanism 80 of the computer 3 is requested to monitor the process execution state. The processing execution state monitoring mechanism 80 of the computer 3 monitors the execution state of the processing T using the timer 50 (step 116). If the processing T finishes the processing within b3 microseconds (step 117), the scheduler 70 recognizes that and the terminal 3 which is the processing request source.
2 (step 119). If the processing time exceeds b3 microseconds before the processing T ends (step 118), the processing execution state monitoring mechanism 80 of the computer 3 reports this to the scheduler 70. Schedule
The processor 70 interrupts the execution of the processing T on the computer 3 and refers to the processing execution computer priority table 140 again to learn that there is no computer to which the processing T is assigned, and the time of the processing T is overridden. -As a baera, it reports to the terminal 32 which is the processing request source (step 120). In addition,
When the processing T is completed within the execution time of the computer, the processing execution state monitoring mechanism 80 reports to the scheduler 70 that the processing is completed normally, and at the same time, the processing result is sent to the terminal 32 which is the processing request source. Report (step 119).

As described above, in the present embodiment, in the case of a process having a short execution time, the process can be completed only by the computer having the first priority. On the other hand, in the case of a process with a long execution time, the time limit is exceeded, and the process is dropped to a computer with a lower priority. Therefore, only a process requiring a long execution time is delayed. However, due to this, processing with short execution time is
It does not interfere with the execution of processing. As a result, a process with a short execution time can obtain a stable response of the processing time. In particular, when constructing a multi-computer system using a plurality of computers having different computing capacities, the execution priority is set in the descending order of the computers having the high computing capacities as in the present embodiment, so that the execution processing time is short. For processing, it is possible to obtain the shortest response time that can be expected in a multiple computer system.

[0012]

As described above, according to the present invention,
The computer that executes the processing is dynamically changed according to the amount of execution of the requested processing, so a stable processing time response can be obtained, especially for processing with a short execution processing time. The operation of the computer system can be realized.

[0013]

[Brief description of drawings]

FIG. 1 is a configuration diagram of a multi-computer system showing an embodiment of the present invention.

FIG. 2 is a diagram showing an example of contents of a processing execution computer priority table in FIG.

FIG. 3 is a diagram showing an example of contents of a process execution time limit table in FIG.

FIG. 4 is an operation flowchart of distributed processing according to an embodiment of the present invention.
It's a chart.

[Explanation of symbols]

 1,2,3 Computer 21,22,23 Direct storage device 31,32,33 Terminal system 40 Memory 50 Timer 60 Buffer 70 Schedule 80 Processing execution state monitoring mechanism 90 Channel 110 Shared memory 120 Shared direct storage device 140 Processing execution Computer priority table 141 Execution priority 142 Computer identification id 150 Process execution time limit table 151 Process execution class 200 Execution time

Claims (1)

[Claims] 1. In a multi-computer system in which a plurality of independent computers and a memory or a direct storage device shared by the computers are connected to each other, the execution priority order of processing is assigned to each computer in the order of the processing capability of the computer. In addition to assigning, set the execution time limit of the process for each computer,
When there is a processing request, the requested processing is executed by the computer with the highest priority first, and the execution time of the processing is monitored by the timer of the execution status monitoring device. As a result of the monitoring, the execution time limit is exceeded. Sometimes, the processing in the above computer is interrupted, the computer with the next highest priority takes over the processing, and the processing is continued. Similarly, when the execution time limit for each computer is exceeded, the priority of the computer is changed. A distributed processing method for a multi-computer system, which is characterized by taking over to a low computer.