JPH06250984A - Distributed processing system - Google Patents

Abstract

PURPOSE:To provide the effective and high-speed distributed processing system by accurately selecting WS for distributed processing and performing link processing between unit processing programs. CONSTITUTION:A processor 1 divides an AP5 into unit processing programs 9 and 10 by a program division routine 6 which are processed by WSs 3 and 4. An optimal WS allocation routine 7 is provided with a WS composition management table 8 storing individual WS resource information 11 and 12 provided on the WSs 3 and 4, specifying the WSs 3 and 4 most suitable for respective programs in prior to the execution of the unit processing programs 9 and 10. In executing the unit processing program 9, the unit processing program and data are down loaded from the processor 1 to the WS 3. At the same time, the address of the WS 4 with the execution of the next unit processing program 10 allocated is added. Thus, the execution of the unit processing program can be directly shifted between the WSs without through the processor 1.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a distributed processing system, and more particularly, to a distributed processing system suitable for application to an application system that requires repetitive processing on a large amount of data such as large-scale numerical processing, image processing, and document retrieval. Regarding processing system.

[0002]

2. Description of the Related Art Generally, a distributed processing system connects a plurality of workstations (hereinafter, referred to as WS) to each other through a communication line so that a processing device as a host can perform various processes on these WSs. Is configured. When the processing device allocates the processing program to the WS, depending on the load on the WS at that time,
The optimum WS for the processing of the program is selected and the execution of the program processing is requested.

As a conventional technique relating to this type of distributed processing system, for example, a technique described in Japanese Patent Laid-Open No. 2-120968 is known.

[0004]

In the distributed processing system according to the prior art described above, the WS to which the processing program is assigned is determined only by the magnitude of the load on the WS. There is a problem in that the processing efficiency of the entire system may be deteriorated due to lack of the processing performance of 1), and it is not always possible to perform optimal WS allocation.

Further, the above-mentioned conventional technique has a problem that when there are a plurality of programs, the unit processing programs cannot be linked with each other, so that the processing efficiency in such a case is lowered. There is.

The object of the present invention is to solve the above-mentioned problems of the prior art, to make the selection of WSs for distributed processing more accurate, and to perform cooperative processing between unit processing programs. An object of the present invention is to provide a distributed processing system that enables the execution and is more efficient and has a higher execution speed.

[0007]

According to the present invention, the above-described object is to provide a configuration management table provided in a processing device that manages a plurality of WSs connected to a network, with an operating rate, an address, a CPU processing performance of each WS. , Hardware capacity information such as memory capacity and external storage capacity, and software composition information are stored, and the most optimal WS for executing the unit processing program can be selected based on these information. It is achieved by doing so.

[0008] Further, the above-mentioned object is to enable continuous processing between unit processing programs by adding the address of the WS to which the unit processing program to be executed next is assigned to the unit processing program being executed. It is achieved by

[0009]

The WS connected to the network stores its own resource information in the storage unit and transfers the resource information to the processing device connected to the network. The processing device collectively manages the information as a configuration management table. This resource information is updated at any time.

The processing apparatus divides an application program (hereinafter referred to as AP) into unit processing programs according to its purpose by a division routine, and based on the resource information in the configuration management table, a unit is optimized by an allocation optimization routine. WS is assigned to execute the processing program.

Then, the processing device adds the address of the WS to the unit processing program and also adds the address of the WS to which the program to be executed next to the unit processing program is assigned at the same time. Is downloaded to WS.

According to the first embodiment of the present invention, therefore, continuous processing between unit processing programs can be performed without passing through a processing device, and distributed processing can be efficiently executed.

[0013]

Embodiments of the distributed processing system according to the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention. In FIG. 1, 1 is a processor, 2 is a network, 3 is a WS, 5 is an AP, 6 is a program division routine, 7 is an optimal WS allocation routine, and 8 is a WS.
Configuration management table, 9, 10 are unit processing programs, 1
Reference numerals 1 and 12 are WS resource information, and 13 is an address addition routine.

A distributed processing system according to an embodiment of the present invention comprises a processing apparatus 1 for managing the entire system, and a plurality of WS3 and WS4 accommodated in a network 2 such as a LAN. Each WS 3, 4 is configured to include its own resource information 11, 12 that it manages. Further, the processing device 1 includes the AP 5 on which the distributed processing is performed,
5 is provided with a program division routine 6 for dividing 5, unit processing programs 9 and 10 after division, an optimum WS allocation routine 7, and an address addition routine 13 for adding an address of the next processing WS. 7, each WS accommodated in the network 2
The WS configuration management table 8 for managing the resource information of the above is provided.

In the above-described embodiment of the present invention, WS
Resource information 11 and 12 managed in 3 and 4
Is hardware resource information such as WS address, CPU operating rate, CPU processing performance, internal memory capacity, and external storage capacity, and software resource information. These resource information are updated every time information changes, or Every predetermined time, as shown by the one-dot chain line, it is reported to the processor 1 via the network 2 and the W in the optimum WS allocation routine 7 is reported.
It is stored in the S configuration management table 8 and managed by the processing device 1.

When the AP 5 is to be distributed to the WSs, the processor 1 first causes the program division routine 6 to divide the AP 5 into a plurality of unit processing programs 9 and 10 according to the purpose. After that, the optimum WS allocation routine 7 determines the WS to execute each of the unit processing programs 9 and 10 based on the content of the resource information of each WS stored in the WS configuration management table 8 and indicates it by the dotted line. Then, the unit processing programs 9 and 10 and the data are downloaded to the WSs 3 and 4.

Now, it is assumed that the processes of the unit processing programs 9 and 10 are executed in order of the programs 9 and 10 by WS3 and WS4, respectively. In this case, the optimum WS allocation routine 7 sends the unit processing program 9 and the data to the WS
When downloading to 3, the address adding routine 13 adds the address of the WS 4 to which the unit processing program 10 to be executed next is assigned to the unit processing program 9 and the data and downloads.

Then, in parallel with the execution of the unit processing program 9 by the WS 3, the unit processing program 10 is executed by the processing device 1.
From WS to WS4. When the processing of the unit processing program 9 ends in WS3, the WS3 performs the next processing that has been added to the processed individual processing program, in this case, by the address of WS4, the processing result is obtained without going through the processing device 1. And necessary data, WS4
Transfer to. As a result, the WS 4 can immediately execute the unit processing program 10.

In the above description, the WS allocation by the optimum WS allocation routine 7 requires the memory capacity, the external storage capacity, the required processing time, etc. required by the unit processing program to be processed, and the WS configuration management table 8 This is done by selecting the most suitable WS for executing the unit processing program based on the resource information of.

According to the above-described embodiment of the present invention, since the resource information of the WS accommodated in the network serving as the reference for allocating the optimum WS for executing the unit processing program is increased, AP when processing
It is possible to finely and optimally perform the WS allocation for executing the unit processing program obtained by dividing.

According to the embodiment of the present invention, since the address of the WS to which the unit processing program to be executed next is assigned is added to the unit processing program currently being processed, the next unit At the time of shifting to the execution of the processing program, the processing can be directly shifted between the WS without going through the processing device, and the overhead can be reduced.

In the above-described one embodiment of the present invention, the distributed processing system is described as being configured by two WSs and processing devices, but the present invention configures the system by accommodating more WSs in the network. According to the embodiment of the present invention, the AP is divided into two individual processing programs. However, the present invention allows the WS to individually perform a plurality of APs that are not divided, or Can also be applied to the case where each WS is divided into a plurality of parts and these are individually performed.

[0024]

As described above, according to the present invention, it is possible to more reliably optimize the WS for allocating the unit processing program obtained by dividing the application when performing the distributed processing, and the processing between the programs. At the time of the shift, the process can be directly shifted to the next WS without passing through the processing device, and the overhead can be reduced.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention.

[Explanation of symbols]

 1 Processor 2 Network 3, 4 Workstation (WS) 5 Application Program (AP) 6 Program Division Routine 7 Optimal WS Allocation Routine 8 WS Configuration Management Table 9, 10 Unit Processing Program 11, 12 WS Resource Information 13 Address Addition Routine

Claims (2)

[Claims] 1. In a distributed processing system in which a processing device and a plurality of workstations are housed in a network, the processing device comprises a configuration management table for storing resource information of each of the plurality of workstations. , A distributed processing system characterized by allocating workstations for performing distributed processing based on the contents of this table. 2. The workstation, wherein the processing device divides an application program into unit processing programs according to its purpose, and when a unit processing program and data are downloaded to an assigned workstation, a work station is executed next. The distributed processing system according to claim 1, wherein the address is added.