JPH10333929A - Job execution system and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent the collision of processes by registering virtual processes respectively allocated to the plural processes in a virtual PID control table, allocating the empty real PID of a real PID control table, registering it in correspondence and making the respective processes communicate with each other by using the virtual process and perform processings within a job. SOLUTION: Plural processors PE1 and PE2 are constituted of the real PID control table 3 and a job execution means 7, etc. Further, the job execution means 7 provides the virtual PID control table 4 for the respective jobs and registers unique virtual process IDs respectively allocated to the plural processes for executing the jobs in the virtual PID control table 4. Then, the empty real PID of the real PID control table 3 is allocated and registered in correspondence and the respective processes communicate with each other by using the virtual process IDs and execute the processing. Thus, the collision between the process IDs is eliminated without the reservation of the process ID.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a job execution system and a recording medium for executing a job by an arbitrary number of processes on a plurality of processors.

[0002]

2. Description of the Related Art Conventionally, a so-called parallel job executed by a group of processes spanning a plurality of PEs (processors) is interrupted at an arbitrary point, its state is stored in a file or the like, and the state is frozen. After that, there is a job freeze / thaw function that reads the frozen job from the file again and continues the execution of the job.

The following points are problematic when freezing and thawing a job. In a process belonging to a job, data is exchanged between processes in the job using the process identifier (process ID). When the job is frozen, information of the process ID is stored in the program space. When the job is decompressed, it is necessary to reproduce the job and avoid the collision of the process IDs in the form of the process in the job at the time of decompressing the job.

At this time, if the same process ID as the process ID is already being used in a process outside the job, the job cannot be resumed without waiting for the end of the process.

To solve this problem, information on a process ID is stored when a process is frozen, a process ID is reserved based on the stored information when the system is restarted, a job is decompressed and reproduced, and another job is executed. Eliminates the wait caused by using the process ID undesirably, starts data exchange between processes using the previously reserved process ID, and resumes the process.

[0006]

As described above, in the conventional job freezing / thawing function, in order to avoid the collision of the process ID at the time of thawing, the process ID is set when starting the system.
Therefore, it is necessary to reserve storage information for reservation according to the job to be decompressed, and to reserve and secure it when the system is started.

In such a conventional job freezing / thawing function, a process ID must be reserved in advance, and there is a problem that a frozen job cannot be thawed.

In a parallel program, since the number of programs constituting a job is equal to the number of PEs, it is necessary to reserve a large number of process IDs, and there are process IDs that cannot be always used in the system. Also, there is a problem that the numbering range of the process ID of another process is pressed. In particular, a large-scale job requiring a job freezing / thawing function tends to have a high degree of parallelism and require a large number of processes.
There is a problem that can not be used.

[0009] The present invention solves these problems,
A virtual PI that uniquely manages a virtual process ID in a job in addition to a real PID control table that manages a unique real process ID
A D control table is provided to correlate the real process ID with the virtual process ID, realize the assignment of any unique virtual process ID for each job, and without reserving the process ID when the frozen job is thawed. An object of the present invention is to enable a process for exchanging data between processes in a job to be immediately resumed without a process ID collision.

[0010]

Means for solving the problem will be described with reference to FIG. In FIG. 1, processors PE1 and PE2 are a plurality of processors for generating a process and executing a job, and here are constituted by an actual PID control table 3, a job execution means 7, and the like. Things.

The real PID control table 3 is for registering and managing a real process ID uniquely assigned to a process. The job execution means 7 executes a job by an arbitrary number of processes on a plurality of processors, and as shown in the figure, a virtual PID control table 4, a virtual PID correspondence table 5,
And a plurality of processes.

The virtual PID control table 4 registers and manages a real process ID in association with a virtual process ID uniquely assigned to a process in a job. Virtual PI
The D correspondence table 5 registers which processor has a virtual process ID.

Next, the operation will be described. Virtual P for each job
A unique virtual process ID assigned to each of a plurality of processes that execute a job by providing an ID control table 4 is registered in the virtual PID control table 4 and an empty real PID in the real PID control table 3 is assigned and associated. After registration, each process in the job communicates with each other using the virtual process ID to execute processing.

At this time, the virtual process I is stored in each processor.
A virtual P that registers which processor D exists in
An ID correspondence table 5 is provided to determine which processor has a virtual process ID in a job, transmit data to that processor, and process in the job mutually transmits and receives data to execute processing. I am trying to do it.

Further, in response to the instruction to freeze the job, the virtual PID control table 4 of the designated job and the processes in the job, or the virtual PID correspondence table 5 is further saved and saved.

In response to the job decompression instruction, the virtual PID control table 4 of the saved job and the processes in the job or the virtual PID correspondence table 5 are restored, and then the virtual PID control table 4 is restored. Empty real process ID in the real PID control table 3 for the virtual process ID registered in
After allotment and registration, each process in the job resumes processing.

Accordingly, in addition to the real PID control table 3 for managing a unique real process ID, a virtual PID control table 4 for uniquely managing a virtual process ID in a job is provided to correspond to the real process ID and the virtual process ID. By assigning a unique virtual process ID to each job, it is possible to immediately assign a unique virtual process ID to each job. Thus, the processing for exchanging data can be restarted.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, an embodiment and operation of the present invention will be described in detail with reference to FIGS. Here, the program read from the recording medium or the program read from the hard disk device as the external storage device, or the program read from the external storage device of the center and transferred via the line is loaded into the main storage and activated. , The following various processes are performed.

FIG. 1 is a block diagram showing one embodiment of the present invention.
In FIG. 1, PE1 and PE2 represent two of a plurality of processors, and are composed of an operating system 2 and a user program 6 as shown.

The operating system (OS) 2 comprises:
As shown in the figure, a real PID control table 3 for allocating and managing a real process ID (real PID) and a virtual process ID (virtual PID) and a job execution unit 7 as shown in FIG. Are composed of a virtual PID control table 4 and a virtual PID correspondence table 5.

The actual PID control table 3 contains the user program 6
The real process ID (real PID) which is uniquely assigned to the process created in the section is registered and managed. The virtual PID control table 4 uniquely assigns a virtual process I to a process created in the user program 6 in a job.
D (virtual PID) is registered and managed.

The virtual PID correspondence table 5 contains a virtual process ID.
Is registered as to which processor (PE) the predetermined range is in (see FIG. 5). The process is created in the portion of the user program 6 and executes various processes, and is assigned a unique process ID (real PID or virtual PID) for mutually communicating data.

Next, a procedure for assigning process IDs (real PID and virtual PID) under the configuration of FIG. 1 will be described in detail according to the order shown in the flowchart of FIG. FIG. 2 is a flowchart illustrating the operation of the present invention.
Here, PE1 and PE2 are P, which is the processor of FIG.
E1 and PE2.

In FIG. 2, S1, S2 are PE1, P
E2 performs IPL of the OS. Here, during the IPL of the OS, the PE1 and PE2 create real PID control tables # 4 and # 5 for uniquely assigning, registering, and managing the actual process ID of the process.

At S3, a parallel job # 10 is input to PE1. With this extension, the parallel job # 10 using PE1 and PE2 is executed. S4 and S5 are S
When a parallel job # 10 is submitted in step 3, virtual PID control tables # 6 and # 7 for each job are created. As a result, virtual PID control tables # 6 and # 7 for registering and managing the correspondence between the virtual process ID and the real process ID for each job have been created.

S6 and S7 correspond to the master processor PE1.
Creates a virtual PID correspondence table # 8 and assigns each PE, here P
Distribution to E2 and PE2 corresponding to virtual PID correspondence table # 9
Create According to the virtual PID correspondence tables # 8 and # 9,
In each PE, the correspondence between the virtual process ID and the PE is recognized, and it is possible to determine to which PE the data should be transmitted based on the virtual process ID. Here, the virtual process IDs 001 to 100 PE1 101 to 200 PE2 are registered in the virtual PID correspondence tables # 8 and # 9.
Up to 00 exist in PE1 and from 101 to 200
In S8, the parent process # 2 is generated in PE1. And
The parent process # 2 generates a child process # 3 in PE2 in S9.

In S10, in response to the creation of the parent process in S8, PE1 assigns and registers "1997" as an empty real process ID in the real PID control table # 4. In S11, as in S10, corresponding to the generation of the child process in S9, the PE2 executes the real PID control table #
“2038” is assigned and registered as an empty real process ID of No. 5.

In step S12, a free virtual process ID (3) is assigned, and the real process ID (1997) and the virtual process ID (3) are associated with each other and registered in the virtual PID control table # 6. Thereby, the virtual PID control table #
6 has a real process ID (3) and a virtual process ID (1
997) is registered.

In step S13, as in step S12, the virtual process ID (101) is assigned and the real process ID (101) is assigned.
D (2038) is associated with the virtual process ID (101) and registered in the virtual PID control table # 7. As a result, the real process ID (20
38) and the corresponding relationship between the virtual process ID (101) are registered.

By the above procedure, an internal data structure for controlling the real process ID and the virtual process ID has been constructed. And, to the user program, S
12, the virtual process ID assigned in S13 is notified, and the user program is notified by itself or the parent process #
2. Recognize the process ID of child process # 3. Here, the parent process # 2 is “3” (actually, the actual process ID,
1997), the child process # 3 recognizes it as “101” (actually, a real process ID, 2038).

Next, in accordance with the order shown in the flowchart of FIG. 3, under the configuration of FIG. 1, the internal data structure for controlling the real process ID and the virtual process ID is constructed by the procedure of FIG. The operation when executing a job will be described in detail.

FIG. 3 is a flowchart illustrating the operation of the present invention. Here, PE1 and PE2 are PE1 and PE2 which are the processors of FIG. In FIG. 3, S21
Sends data from process # 2 (parent process) to PID = 101.

In S22, the operating system 2 of the PE 1 determines the communication destination PE by referring to the virtual PID correspondence table # 8. Here, PID = 101 is determined as PE2.
In step S23, since PE2 is determined in step S22, data is sent to the process of PID = 101 of the parallel job # 10 to PE2. Since the communication destination PE is determined to be PE2 in S22, the data is transmitted to the process of the virtual process ID (101) under the same job # 10 as PE1.

In step S24, the operating system 2 of the PE 2, which has received the data transmitted in step S23, looks at the virtual PID control table # 3 of the parallel job # 10 and processes the process # 3.
Pass the data to Then, the job is executed.

As described above, in response to the data transmission from the process # 2 in the parallel job # 10 of PE1 of FIG. 1 to PID = 101, the data is passed to PE2 with reference to the virtual PID correspondence table # 8. , Further virtual PID in PE2
Process # in parallel job # 11 with reference to control table # 7
3, the data has been passed. Thus, the data is transmitted to the corresponding PE by referring to the virtual PID correspondence table, and then the data is passed to the corresponding process by referring to the virtual PID control table in the PE, and the process in the different PE in the job is executed. This means that the data has been successfully transmitted.

Next, in accordance with the order shown in the flowchart of FIG. 4, freezing and thawing of a job under the configuration of FIG. 1 will be described in detail. FIG. 4 is a flowchart illustrating the operation of the present invention.

In FIG. 4, S31 receives a job freeze instruction. In this case, the OS receives a job freeze instruction from a system administrator or a user. Then, a job freeze instruction is notified to each PE that is executing the job.

At S32, the PE 2 receives the job freeze instruction notified at S31. In S33 and S34, the virtual PID correspondence tables # 8 and # 9, the virtual PID control tables # 6 and # 7, and the processes # 2 and # 3 of the parallel job # 10 are saved and saved in files in each PE.

In step S35, the job freeze is completed. As described above, when the job freeze instruction is received, each PE executing the job that received the freeze instruction saves the virtual PID correspondence table, the virtual PID control table, and the process of the job in a file, respectively. I do.

In step S40, job decompression is started. S41
Receives the job decompression instruction. In this case, the OS receives a job decompression instruction from a system administrator or a user. Then, a job decompression instruction is notified to each PE to which the job is assigned.

In step S42, the job decompression instruction notified by the PE 2 in step S41 is received. In S43 and S44, the virtual PID correspondence tables # 8 and # 9 and the virtual PID control tables # 6 and # 7 of the parallel job # 10 are restored in each PE. Then, the information of the processes # 2 and # 3 is restored to PE1 and PE2 together.

S45 and S46 correspond to the actual PID control table # 4,
The reassignment of the empty process ID of # 5 and registration in the virtual PID control tables # 6 and # 7 are performed. As a result, the empty real process IDs of the real PID control tables # 4 and # 5 are assigned, and the correspondence between the real process IDs and the virtual process IDs is registered in the virtual PID control tables # 6 and # 7, and the real process IDs are registered. And the virtual process ID are determined.

In steps S47 and S48, the processes # 2 and # 3 are restored, and the execution of the job is resumed. As mentioned above,
After restoring the virtual PID correspondence table, virtual PID control table, and process information, an empty real process ID is obtained from the real PID control table.
After the real process ID is registered in the virtual PID control table to determine the correspondence with the virtual process ID,
The process can be restored and job execution can be resumed. As described above, as the assignment of the real process ID, an empty real process ID is registered in the virtual PID control table in association with the virtual process ID with reference to the real PID control table. Process ID
Can be assigned, and the conventional process does not use the same real process ID as at the time of freezing and waits until the process terminates. This eliminates the occurrence of the situation, and quickly changes the real process ID to the virtual process ID. Job execution can be resumed by assigning.

FIG. 5 shows an example of a virtual PID correspondence table according to the present invention. The virtual PID correspondence table 5 registers processor IDs in association with the range of the virtual process ID.
Here, when the range of the virtual process ID is 001 to 100, the virtual process ID is registered as the processor PE1 and the virtual process ID is registered.
It turns out that the processes 001 to 100 operate on the processor PE1. Therefore, in the case of a data transmission request to a virtual process ID whose virtual process ID is in the range of 001 to 100, the data may be transmitted to the PE1. Then, the PE 1 refers to the virtual PID control table 4 and passes the received data to the process of the corresponding virtual process ID.

[0045]

As described above, according to the present invention,
In addition to the real PID control table 3 for managing a unique real process ID, a virtual P for uniquely managing a virtual process ID in a job
By providing an ID control table 4, the real process ID and the virtual process I
D, a unique virtual process ID can be assigned to each job, and when a frozen job is thawed, the conventional process I
The process of exchanging data between the processes in the job can be restarted immediately without the process ID collision without the reservation of D. Also, the process ID is automatically assigned when the process ID is reassigned at the time of decompression, without being aware of the user program configuration (process configuration).
It is possible to reliably and quickly restart the job by avoiding the collision of D, and to continue the processing.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of one embodiment of the present invention.

FIG. 2 is a flowchart for explaining the operation of the present invention (process I).
D).

FIG. 3 is a flowchart for explaining the operation of the present invention (data transmission / reception using a virtual process ID).

FIG. 4 is a flowchart for explaining the operation of the present invention (assignment of process IDs when a job is frozen and a job is thawed).

FIG. 5 is an example of a virtual PID correspondence table according to the present invention.

[Explanation of symbols]

 2: Operating system (OS) 3: Real PID control table 4: Virtual PID control table 5: Virtual PID correspondence table 6: User program 7: Job execution means PE, PE1, PE2: Processor

Claims (5)

[Claims] 1. A job execution system for executing a job by an arbitrary number of processes on a plurality of processors, wherein a unique real process ID assigned to each of the processes executing processes on the plurality of processors is registered and managed. A PID control table and a unique virtual process ID assigned to each of a plurality of processes executing a job on a plurality of processors in the job are registered and managed in association with the unique real process ID. A job execution system, comprising: a virtual PID control table provided in each processor; and in the job, each process communicates with each other using a virtual process ID to execute processing. 2. A virtual PID correspondence table for registering which processor has a virtual process ID for each job in each processor, and determines which processor has a virtual process ID in the job. 2. The job execution system according to claim 1, wherein each of the processes in the job transmits and receives data to and from the processor by transmitting data to the processor. 3. The virtual PID control table of the designated job and a process in the job in response to the freeze instruction of the job.
3. The job execution system according to claim 1, wherein the virtual PID correspondence table is saved and saved. 4. In response to a job decompression instruction, after restoring the saved virtual PID control table of the job and the process in the job, or further restoring the virtual PID correspondence table,
An empty real process ID in the real PID control table for the virtual process ID registered in the virtual PID control table
The job execution system according to any one of claims 1 to 3, wherein each process in the job resumes processing after assigning and registering the job. 5. A unique real process I assigned to a process that executes a process on each of a plurality of processors.
Means for registering and managing D in the real PID control table; and a unique virtual process ID assigned to each of a plurality of processes executing a job on a plurality of processors in the job. Means for registering and managing in the virtual PID control table in association with each other, means for communicating with each other using the virtual process ID in the job and executing processing, and recording a program to function as Computer-readable recording medium.