JPH11259431A - Distributed processing method and computer readable storage medium for recording network distributed processing program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To process processing units for making uniform a whole processing time on each computer system by classifying the designation of the processing units in a descending order by using a processing unit quantity or processing time related with plural processing units as a key, and operating the program processing of the plural processing units on plural computer systems in the classified order. SOLUTION: Information such as the name of a module to be compile- processed and the file size is collected so that a compile management table can be prepared (S101). Then, the designation of processing units is classified in a descending order by using a processing unit quantity (the file size, the number of steps, the number of sentences, the number of words, or the number of characters) or processing time (a processing lapse time, a CPU using time, or the number of times of input and output) related with the processing units as a classification key (S102 and S103). The distributed processing of the processing units is operated in the classified order by a program operating on plural systems connected through a network (S104, 105, and 110).

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 for distributing processing by a program to a plurality of computers connected to a network. In recent years, as the cost of communication facilities such as LANs has become lower, the processing time irrelevant to the processing order has been allocated on a plurality of computer systems connected by a network, and distributed processing has been performed. Is being done. However,
In processes that involve large variations in each process elapsed time, if simply scheduling the order of processing, there are computers that finish processing early and computers that do not finish processing easily, and the entire process is not well distributed. There is a problem that the whole processing is not completed in a well-balanced manner.

[0002] From such a viewpoint, a processing unit is allocated to each computer system in a well-balanced manner, the processing of all the processing units on the network is scheduled to end at the same time, and the processing elapsed time by the distributed processing is further reduced. Provision is desired.

[0003]

2. Description of the Related Art In a conventional distributed translation process, a source program is translated (compiled) on a plurality of computer systems connected to a network in a predetermined processing order. However, the number of steps differs greatly for each source program, and the variation in the translation elapsed time of each source program was remarkable. Therefore, for example, a total of 900 source programs can be stored in an arbitrary 300
When a book is set as one set and distributed translation processing is started simultaneously by compilers operating on three computer systems connected by a network, the time at which each set ends varies, and the translation that ended last Time became the total processing time.

FIG. 9 is a diagram for explaining the compilation CPU usage time and allocation order of each module. This figure shows that there is a large variation in the compile CPU usage time. For example, if the compilation order of each module is fixedly compiled by three hosts,
As indicated by 0, the computer of M2 completes the translation of three programs in a short time, but the computer of M1 may require an extremely long time. As described above, there is a problem that the total compilation processing elapsed time becomes the computer processing time of M1 which requires the longest processing elapsed time.

[0005] However, if the compile processing elapsed time of each module is known in advance, as shown in FIG.
To M3 to perform the translation processing of each module, it is possible to reduce the elapsed time of the entire compilation processing. As described above, in the porting work or the system development that requires compiling a large number of programs at a time, the work of translating a large number of modules occurs at a time, and it is desired to reduce the compile time by distributed translation processing.

[0006]

As described above, the distributed translation processing shortens the entire translation processing time on the network system, but the compile time variation limits the translation processing time. SUMMARY OF THE INVENTION It is an object of the present invention to perform processing units in distributed processing performed on a plurality of computer systems connected by a network such that the total processing time on each computer system becomes uniform.

[0007]

The processing elapsed time greatly changes depending on the operating state of the system, such as waiting for CPU resource allocation in the computer system to be processed and waiting for input / output conflict. That is, the processing elapsed time becomes longer if a lot of resource contention waiting occurs due to the execution of many programs, and the processing elapsed time becomes shorter if the waiting time is small. For this reason, the CPU usage time and the number of times of input / output, or a value obtained by combining the CPU usage time and the number of times of input / output, which change little, may be referred to as the predicted value of the processing elapsed time. In the present invention, the processing unit is data processed by an application or the like,
It means sentences, source programs, etc., and the processing unit amount per processing unit is the file size of the processing unit, the number of steps,
It means the number of sentences, the number of words or the number of characters. Also,
The processing time per processing unit includes CPU usage time, processing elapsed time (time required from start to end of processing), and CP
The estimated processing time based on the U use time and the number of input / output times (for example, CPU use time * coefficient 1 + input / output number * coefficient 2) is included. Further, the plurality of computer systems connected by a network include the own computer system.

According to the present invention, in a distributed processing method for performing distributed processing by a plurality of computer systems connected via a network, the names of the processing units are sorted in descending order using the processing unit amounts or processing times of the plurality of processing units as keys. And a processing procedure for causing a plurality of processing units to perform program processing on a plurality of computer systems in the order classified by the name classification means. A computer-readable recording medium recording a network distributed processing program for performing distributed processing by a computer system,
A name classification procedure for classifying the names of the processing units in descending order using the processing unit amount or the processing time related to the plurality of processing units as a key,
A computer-readable recording medium recording a network distributed processing program for executing a processing procedure for causing the plurality of processing units to perform program processing on the plurality of computer systems in the order classified by the name classification unit. It is means for solving the above-mentioned problems.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a diagram for explaining the flow of the first embodiment of the present invention. This will be described with reference to FIG. In the first embodiment of the present invention, the execution multiplicity of the compile process in each computer is 1. Therefore, the module names (2) to (4) during compilation shown in FIG. 7 are not used.

In step (S101) shown in FIG. 1, the module name (the name of the translation unit) to be compiled and information on the file size, the number of steps, the number of sentences, the number of words, and the number of characters for each module name are stored. Each is collected and the compilation management table shown in FIG. 6 is created. In the compilation management table, a module name 61 is set as the file name. Furthermore, the file size 62, the number of steps 63, the number of sentences 64, the number of words 65, and the number of characters 6 for each module name
6 are set respectively. The set information is used as a classification key for determining the compilation order. However, if the first compile process has been executed, in step (S122), the compile process elapsed time 68, the CPU usage time 69, and the number of input / output times 70 in the previous compile process are set in the compile management table. ing.

In the step (S102), for example, an instruction is given as a parameter at the time of starting the program to perform the compilation in the order of the file size of each module. The module names are sorted in descending order by file size, and the order is determined so that distributed compilation is performed in order from the largest value in the column of file size 62. In addition, any of file size, number of steps, number of sentences, number of words, number of characters, compile processing elapsed time, CPU usage time, and number of input / output times can be specified as the classification key for determining the compilation order.

In step (S103), the compile order determined in step (S102) is stored in the field of the processing order number 67 in the order of distributed compilation (distributed translation) from the largest file size to the smallest file size. Is set. That is, in FIG. 6, 1 is set as the value set in the processing order number 67 corresponding to the line of the module name to be compiled first, and 2 is set as the value of the module to be processed next. For subsequent module names, values that are sequentially added one by one are set. The largest value among the values set in the processing sequence number 67 is set in a final processing sequence number counter (not shown).

In step (S104), all the module names (1) 72 being compiled in the compilation monitoring table shown in FIG. 7 are initialized with spaces. FIG. 7 is a view for explaining the compilation monitoring table. In the compilation monitoring table, the names of the computers connected via the network are set in advance in the computer name 71, and the module name translated by the operating compiler on the computer name is the module name being compiled in the line corresponding to the computer name. (1) 72
Is set to

In step (S105), 1 is set to a compile process management counter (not shown). In addition,
The compile processing management counter sets and holds the processing order of the next module to be compiled and is referred to when a module to be compiled is selected. In step (S110), it is determined whether the translation of all modules has been completed. That is, the value of the compile process management counter exceeds the value of the final process sequence number counter, and at least one module name is set in the module name (1) 72 being compiled in the compile monitoring table shown in FIG. If not, it is determined that all modules have been compiled.

In step (S111), if all the module names are set in the compiling module name (1) 72 of the compilation monitoring table shown in FIG. 7, it is determined that all the computers are executing the compiling process. I do. If at least one module name has not been set, the flow branches to step (S112); otherwise, the flow branches to step (S121).

In step (S112), the compile processing management counter is read out, and step (S1)
In 13), the next one line is read from the compilation management table shown in FIG. 6, and the module to be compiled next is selected. Step (S114) is equivalent to step (S20).
1, S121), and each step will be described in more detail with reference to FIG. In the figure, in step (S114), a subprocess is started as a main process using a UNIX (Solaris 1.x) fork function. Next, the rsh command is executed by the execute function from the subprocess,
The host name is specified by the hostname operand of the rsh command, and a time command and a compiler command to be executed by the host are also specified. Thereby, in step (S201), a time command is activated on the host. Next, the time command activates a compiler and executes a compiling process for a module name specified as an operand of the compiler command. When the compile processing is completed, control returns to the time command from the compiler. The time command outputs the compile processing elapsed time, the CPU usage time, and the number of times of input / output of the compiler to the file system of the calling computer via standard output and NFS (Network File System). Then, the process returns to the calling sub-process in step (S114) and executes the exit function. This exit
By executing the function, wa of step (S121) is obtained.
The wait state by the wait process of the it function is released, and the process proceeds to the next step (S122).

After executing the fork function in step (S114), the process immediately proceeds to the next step (S115). In step (S115), step (S11
The name of the module requested to be compiled in step (S114) is set to the module name (1) 72 being compiled in the compilation monitoring table that matches the computer name (host name) requested to be compiled in 4).

In step (S116), the value of the processing sequence number 67 to be processed next is set in the compile processing management counter. Then, the process proceeds to step (S110). In step (S121), step (S114)
Compile request to the host, and waits for completion of this compile processing in this step. That is, in this step, the wait state is entered by executing the wait function. Then, when the exit function is executed in step (S114), the wait state by the wait function is released, and the process proceeds to the next step (S122).

In step (S122), t
Compile processing elapsed time by CPU
The usage time and the number of times of input / output are acquired via the NFS (file sharing system). Then, respective values are set to the translation elapsed time 68, the CPU usage time 69, and the number of input / output times 70 in the compilation management table. Step (S1
In 23), the name of the module whose translation has been completed is replaced with the name of the module being compiled in the compilation monitoring table (1) 72.
And initialize the corresponding module name with blanks.
Then, the process proceeds to step (S110).

A second embodiment of the present invention will be described with reference to the drawings. FIG. 2 is a view for explaining the flow (No. 1) of the second embodiment of the present invention. This will be mainly described with reference to FIG. In the second embodiment of the present invention, the execution multiplicity of the compile process in each computer is 1 to 4. Therefore,
Module names (1) to (4) being compiled shown in FIG.
Use

Steps (S101 to S103) perform the same processing as the steps (S101 to S103) of the first embodiment of the present invention, and determine and set the compile order based on the classification key. Since it has already been described, the description is omitted. In step (S104), the module name (1) 72 being compiled in the compilation monitoring table shown in FIG.
Initialize all module names (4) 75 being compiled with spaces.

In step (S105), 1 is set to a compile process management counter (not shown). In addition,
The compile processing management counter sets and holds the processing order of the next module to be compiled and is referred to when a module to be compiled is selected. In step (S106), all the compile execution numbers 83 of the multiplicity management table (see FIG. 8) for managing the execution multiplicity of the compiler in each computer are initialized to 0 (zero). In this multiplicity management table, the computer name 81 is set in advance to each computer name, and the compilation allowable multiplicity 82 is set in advance to the compilation allowable multiplicity for each computer name.
Also, a compile execution number 83 is provided in which a value of the compile execution number at the time of executing the compiler is set.

In step (S107), 1 is set to an instruction counter (not shown). Step (S108)
Then, the first row is read from the multiplicity management table. In step (S110), it is determined that all compilations have been completed, and the value of the compilation processing management counter exceeds the value of the final processing sequence number counter, and the name of the module being compiled is displayed in the compilation monitoring table. Is not set (all module names are blank), it is determined that all translations have been completed.

Step (S110a) is performed when the current allowable multiplicity is equal to or larger than the currently executed multiplicity for each computer name column of the multiplicity management table.
10b), otherwise (step S110)
Proceed to c). In step (S110b), it is determined whether the number of compiles in the multiplicity management table has the same value as the instruction counter.
112), and if different, step (S110)
Proceed to C).

The step (S112) reads the compile processing management counter, and the step (S113) reads the compile management table. Step (S114) operates in cooperation with step (S201, S121).
Steps S116, S201, S121, S115, S11 in the figure for explaining the flow of the first embodiment of the present invention, as well as the setting of the translation module name and the processing number in S116).
This is the same process as 6), and has already been described, and thus will not be described.

In step (S117), the computer name 81 of the multiplicity management table is set by adding 1 to the value of the number of compiles executed in the row corresponding to the computer name requested to be compiled in step (S114). Step (S110
In c), when the processing of all the computer names (the first and second computers) set in the computer name 71 is completed, step (S1)
Proceed to 19); otherwise, step (S118)
Proceed to.

In step (S118), the next row in the multiplicity management table is read, and the flow advances to step (S110). In step (S119), 1 is added to the value of the instruction counter. In step (S110d), the compile allowable multiplicity number 82 of the row read from the multiplicity management table is set.
When the value is equal to or smaller than the instruction counter, the process proceeds to step (S121); otherwise, the process proceeds to step (S108).

Steps (S121, S122, S12)
3) issues a wait function that waits for a notification of the result of requesting translation processing to the host, sets the compile processing elapsed time after the wait state is released, and initializes the module name. For the detailed description, refer to the steps (S121, S122, S122) of the flow of the first embodiment of the present invention.
123).

In step (S124), in the computer name 81 of the multiplicity management table, 1 is subtracted from the value of the compile execution number 83 of the row that matches the computer name that has been compiled in step (S121). Step (S125)
Is the name of the module being compiled (1) 72 to the name of the module being compiled (4) in the compilation monitoring table.
When all blanks are set in 75, it is determined that compilation of all modules has been completed, and the processing is terminated. Otherwise, go to step (126).

In step (S126), a line is read from the compile management table. Step (S12
Step 7) requests the host to perform compile processing, and performs the same processing as described in the above-described step (S114). Steps (S128 and S129) set the translation module name in the compilation monitoring table and the next processing sequence number in the compilation processing management counter.

In the step (S130), 1 is added to the value of the compile execution count 83 in the row of the computer name 81 corresponding to the computer name 81 of the multiplicity management table and the host name of the translation request in the step (S127). Then, step (S
Go to 121).

[0032]

As described above, according to the present invention, the processing unit amount (file size, number of steps,
The names of processing units are sorted in descending order using the number of sentences, any of words, and the number of characters) or the processing time (one of elapsed processing time, CPU usage time, and number of input / output times) as a classification key, and the classification order By distributing the processing units using a program operating on a plurality of systems connected by a network, the processing elapsed time in each computer system can be averaged, and as a result, the total processing time can be reduced effective.

[Brief description of the drawings]

FIG. 1 is a view for explaining the flow of a first embodiment of the present invention;

FIG. 2 is a flowchart (part 1) of a second embodiment of the present invention;
Diagram explaining

FIG. 3 is a flowchart (part 2) of the second embodiment of the present invention;
Diagram explaining

FIG. 4 is a flowchart (part 3) of the second embodiment of the present invention;
Diagram explaining

FIG. 5 is a view for explaining data compilation such as a compile instruction and CPU usage time for a host.

FIG. 6 is a view for explaining a compilation management table;

FIG. 7 is a view for explaining a compilation monitoring table;

FIG. 8 illustrates a multiplicity management table.

FIG. 9 is a diagram illustrating the elapsed processing time and the CPU usage time of each module.

FIG. 10 is a view for explaining processing elapsed time by fixed order compilation;

FIG. 11 is a view for explaining reduction of processing elapsed time by distributed compilation in the order of CPU use time.

[Explanation of symbols]

 61 Module name 62 File size 63 Number of steps 64 Number of sentences 65 Number of words 66 Number of characters 67 Processing sequence number 68 Compile processing elapsed time 69 CPU usage time 70 Number of input / output times 71 Computer name 72 Module name during compilation (1) 73 Compiling Module name (2) 74 Module name being compiled (3) 75 Module name being compiled (4) 81 Computer name 82 Compilable multiplicity 83 Compilation execution number

Claims (4)

[Claims] 1. A distributed processing method for performing distributed processing by a plurality of computer systems connected via a network, comprising: a name classification procedure for classifying processing unit names in descending order by using a processing unit amount relating to a plurality of processing units as a key; And a processing procedure for causing a plurality of processing units to perform program processing on the plurality of computer systems in the order classified by the name classification means. 2. A distributed processing method for performing distributed processing by a plurality of computer systems connected via a network, comprising: a name classification procedure for classifying processing unit names in descending order by using processing times relating to a plurality of processing units as a key; A processing procedure for causing a plurality of processing units to perform program processing on the plurality of computer systems in the order classified by the name classification means. 3. A computer-readable recording medium storing a network distributed processing program for performing distributed processing by a plurality of computer systems connected via a network, wherein a processing unit amount related to a plurality of processing units is used as a key. A name classification procedure for classifying processing unit names in descending order, and a processing procedure for causing the plurality of processing units to perform program processing on the plurality of computer systems in the order classified by the name classification means. A computer-readable recording medium recording a network distributed processing program. 4. A computer-readable recording medium on which a network distributed processing program for performing distributed processing by a plurality of computer systems connected by a network is recorded, wherein the processing is performed by using a processing time relating to a plurality of processing units as a key. A network for executing a name classification procedure for classifying unit names in descending order, and a processing procedure for causing the plurality of processing units to perform program processing on the plurality of computer systems in the order classified by the name classification unit. A computer-readable recording medium recording a distributed processing program.