JPH11338719A - Computer system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To deal with the save/restoration of a job without reserving any process ID at the time of executing a job provided with plural processor elements and composed of one or plural processes. SOLUTION: This system is provided with a managing means 10 for managing the relation of correspondence among a job ID allocated to the job, processor ID of a virtual processor element allocated to the job and processor ID of a real processor element for executing the job and an allocating means 12 for generating the process ID defined by coupling the bits of a local process ID defined in the real processor element for executing a process composing of the job when generating that process, virtual processor ID provided for that job instructed by that real processor element and job ID allocated to that job and for allocating the process ID to that process.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a computer system having a plurality of processor elements and executing a job composed of one or a plurality of processes. More particularly, the present invention relates to a system for saving a job without reserving a process ID. -It relates to a computer system that can deal with restoration.

[0002]

2. Description of the Related Art In a computer system, mutually unique job IDs are assigned to jobs in order to guarantee uniqueness of the jobs.

On the other hand, in a computer system extended in parallel, a job is composed of one or a plurality of processes, and unique to both processes belonging to a job and processes not belonging to a job. Process I
Assign D.

In a parallel processor system composed of a plurality of processor elements, a job ID and a process I
D must guarantee uniqueness across processor elements.

[0005] Under these conditions, after a job composed of a process group is saved to an external storage device and another operation is performed, the process ID assigned to the process group belonging to the saved job becomes later. May overlap with the process ID of the process generated by the process described above, and a collision of the process ID may occur when the saved job is restored.

[0006] This process ID collision has a high probability of occurrence when there are many saved processes or when processes are frequently created and terminated during saving. Conventionally, in order to avoid the collision of the process IDs, a configuration is adopted in which a reservation table for registering process IDs reserved for use is prepared, and when a process is saved to an external storage device, the process is allocated to the process. Process I
By registering D in the reservation table, the process ID of the process is prevented from being allocated to a process generated later during the process of saving the process.

[0007]

However, according to such a conventional technique, since the process ID registered in the reservation table cannot be used, the numbering range of the process ID assigned to the process generated later is not available. There was a problem that was pressed.

Furthermore, there is a problem that it is necessary to prepare a reservation table having a memory capacity corresponding to the number of processes to be saved. The present invention has been made in view of such circumstances, and has a configuration in which a plurality of processor elements are provided to execute a job including one or a plurality of processes, and a process ID is reserved. Without
It is an object of the present invention to provide a new computer system that can deal with saving and restoring jobs.

[0009]

FIG. 1 illustrates the principle configuration of a computer system equipped with the present invention. As shown in this figure, a computer system equipped with the present invention is composed of a plurality of processor elements 1-i (i = 1, 2,...) And an external storage device 2, and an operating system 3
The configuration includes a management unit 10, a registration unit 11, an allocation unit 12, and a specification unit 13.

[0010] The management means 10 stores a job I assigned to a job (consisting of one or a plurality of processes).
It manages the correspondence between D, the processor ID of the virtual processor element assigned to the job, and the real processor ID of the processor element 1-i that executes the job.

The registration means 11 generates data managed by the management means 10 when a job is generated, and registers the data in the management means 10. When a process is generated, the allocating unit 12 generates a process ID serving as an identifier of the process, and allocates the process ID to the process.

The specifying means 13 specifies a process as a communication destination when a communication request is issued by designating a process ID. In the computer system having the present invention configured as described above, the management unit 10 manages the correspondence between the job ID, the virtual processor ID, and the real processor ID according to the registration processing of the registration unit 11.

For example, for the job developed in the processor elements 1-1 and 2 in FIG.
As shown in FIG. 2, a job ID (JID) of “1”
And has two virtual processor IDs “VP0” and “VP1”, and Pex of the processor element 1-1 assigned to “VP0” of the virtual processor ID as a real processor ID, and “VP” of the virtual processor ID.
PE of processor element 1-2 assigned to 1 "
It manages having y.

As shown in FIG. 2, the management means 10 has a job ID (JID) of “2” and “VP0” for the job developed in the processor elements 1-2 and 3 of FIG. And the virtual processor ID “VP1”, and the PEy of the processor element 1-2 assigned to the virtual processor ID “VP0” and the virtual processor ID “VP” as the real processor ID.
PE of processor element 1-3 assigned to "1"
It manages having z.

On the other hand, when a process constituting a job is generated, the allocating means 12 determines a local process ID defined in the processor element 1-i executing the process and a local process ID indicated by the processor element 1-i. The virtual processor ID of the job and the job I of the job
A process ID as shown in FIG. 3A defined by bit combination with D is generated and assigned to the process.

When a process that does not constitute a job is generated, the allocating unit 12 determines the local process ID defined in the processor element 1-i that executes the process, and the actual processor of the processor element 1-i. A process ID as shown in FIG. 3B defined by bit combination with the ID is generated and assigned to the process.

Here, the "code" shown in FIG. 3 is an identification code indicating whether or not the process is a job constituting process. By using the process ID of the present invention, a job is saved to the external storage device 2 and then another operation is performed, whereby the process generated later is allocated to the process of the saved job. Even if the same local process ID as the current one is assigned, no inconvenience occurs.

Further, by using the process ID of the present invention, when a job is moved between the processor elements 1-i, the same local process as the one assigned to the process of the job is set in the transfer destination. Even if the process having the ID is deployed, no inconvenience occurs.

That is, the process ID of the present invention has a data structure obtained by extending the local process ID used in the prior art. The process can be identified at all times.

When the process ID of the present invention is used and a communication request is issued by designating the process ID,
When the “sign” of the process ID indicates that the process does not constitute a job, the specifying unit 13 expands the processor element 1-i indicated by the real processor ID of the process ID. The process indicated by the local process ID of the ID is specified as a communication destination.

On the other hand, when the "sign" of the process ID designated as the communication request destination indicates that the process is a process constituting a job, the specifying means 13 determines the job ID / virtual processor ID of the process ID. Is used as a search key to search the management means 10 to specify the real processor ID, which is expanded in the processor element 1-i indicated by the specified real processor ID, and is indicated by the local process ID indicated by the process ID. Is specified as a communication destination.

As described above, when the process ID of the present invention is used, the communication request destination process can be easily specified by referring to the management data of the management means 10.

[0023]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail according to embodiments. FIG. 4 shows an embodiment of a computer system equipped with the present invention.

The computer system equipped with the present invention according to this embodiment is composed of PEx, PEy, PEz, and PEα.
And an external storage device 2.

On each processor element, an operating system operates, and for the operation of the system, a plurality of jobs constituted by a single or a plurality of process groups and a plurality of processes which do not constitute a job operate. doing.

In each processor element, a job management table 100 notified from the operating system is developed. The job management table 100 manages the correspondence between a job ID assigned to a job, a virtual processor ID assigned to the job, and a real processor ID of a processor element that executes the job.

When a job is submitted to the system, the operating system assigns a unique job ID to the job and a virtual processor defined by a serial number relative to the processor element executing the job. An ID is assigned and registered in the job management table 100 developed for each processor element. Here, the job ID and the virtual processor ID are:
While the job exists, it is treated as an invariant value.

To explain with reference to the embodiment of FIG. 4, a job I "1"
D, a virtual processor ID “VP0” is assigned to the processor element PEx executing the job, and a virtual processor ID “VP1” is assigned to the processor element PEy executing the job. It is registered in the job management table 100.

A job ID “2” is assigned to a job input to the system.
A virtual processor ID “VP0” is assigned to the processor element PEy executing the job, and the processor element PE executing the job is assigned.
A virtual processor ID “VP1” is assigned to z and registered in the job management table 100.

Then, when a job is operated, the operating system performs processing to rewrite the management data of the job management table 100 as necessary according to the operation.

For example, when the job shown in FIG. 4 is moved as shown in FIG. 5, when the processor element executing the job moves from PEy / PEz to PEz / PEα, the processor element PEz executing the job moves to PEz / PEα. A virtual processor ID “VP0” is assigned to the processor element P for executing the job.
The virtual processor ID “VP1” is assigned to Eα, and the management data of the job management table 100 is rewritten according to the virtual processor ID.

The job management table 100 is obtained by extending the job management table of a standard UNIX system shell in parallel using array variables. Further, when the process is created, the operating system
, A process ID characteristic of the present invention is assigned to the process.

Next, the process of assigning the process ID will be described with reference to the process flow shown in FIG. When the process is created, the operating system firstly performs step 1 as shown in the processing flow of FIG.
For the generated job, a 16-bit process ID (hereinafter, referred to as a local process ID) used in a standard UNIX system having a single processor configuration is generated.

Subsequently, in step 2, it is determined whether or not the created process constitutes a job.
When judging to configure a job, step 3
Then, the job ID of the job to which the generated process belongs is specified, and in the subsequent step 4, the real processor ID of the processor element that executes the generated process is specified.

Subsequently, in step 5, the job management table 100 is searched by using the job ID specified in step 3 and the real processor ID specified in step 4 as search keys, thereby corresponding to the generated process. The virtual processor ID to be assigned is specified.

Subsequently, in step 6, the local process ID generated in step 1, the job ID specified in step 3, and the virtual processor ID specified in step 5
7 and a flag value “1” (job determination flag) indicating the process ID of the process constituting the job is added to the head of the
A 32-bit process ID as shown in (a) is generated and assigned to the generated process.

On the other hand, if it is determined in step 2 that the created process does not constitute a job, the process proceeds to step 7 where the actual processor ID of the processor element that executes the created process is specified.

Subsequently, in step 8, the local process ID generated in step 1 and the real processor ID specified in step 7 are bit-coupled, and the process ID of the process that does not constitute a job at the beginning is determined. By adding the indicated flag value “0” (job determination flag), a 32-bit process ID as shown in FIG. 7B is generated and assigned to the generated process.

As described above, in the standard UNIX system having a single processor configuration, a 16-bit process ID (local process ID according to the present invention) as shown in FIG. 7C is assigned to a process. In contrast to the configuration, in the computer system including the present invention,
When assuming a parallel processor system, UNIX
In consideration of compatibility to take over the rich assets of the system, the lower 2 bytes have a local process ID assigned by the UNIX system, and the upper 2 bytes
In the case of a process belonging to a job, a process ID having a job determination flag / job ID / virtual processor ID is assigned, and for a process not belonging to a job, a process ID having a job determination flag / real processor ID is assigned.

In this way, for example, for a process developed in the processor element PEx of the job shown in FIG. 4, [1,1, VP0, aaaa] where "aaaa" is a 32-bit local process ID The process ID of FIG.
[1, 1, VP1, bbbb] where "bbbb" is a 32-bit process ID called a local process ID, which is assigned to the process expanded in the processor element PEy of the job shown in FIG.
[0, PEx, cccc] where "cccc" is a 32-bit process ID called a local process ID.

By using this process ID, in the computer system equipped with the present invention, the job is stored in the external storage device 2.
Performing other operations after saving to the same process as the local process ID assigned to the process of the saved job
Nothing happens when something happens to be assigned.

Also, by using this process ID, in the computer system equipped with the present invention, when a job is moved between the processor elements, the same destination as that assigned to the process of the job is assigned to the destination. No inconvenience occurs even if a process having the local process ID of “.

That is, the process ID used in the computer system having the present invention has a data structure obtained by extending the local process ID used in the standard UNIX system. Process can be identified even if it has the local process ID.

When a communication request is issued by designating a process ID from an in-job process or an out-of-job process, the operating system performs a process of specifying a process of a communication request destination according to the process flow of FIG.

Next, the process of specifying the communication request destination will be described. The operating system uses the process ID
When a communication request is issued by designating, as shown in the processing flow of FIG. 8, first, in step 1, the first bit of the designated process ID is read.

Subsequently, in step 2, it is checked from the read bit value whether the process indicated by the specified process ID is an in-job process or an out-of-job process, and it is determined that the process is an in-job process. When a determination is made, the process proceeds to step 3, in which a cutout process based on a bit operation is executed to cut out a job ID / virtual processor ID / local process ID from a designated process ID.

Subsequently, in step 4, the actual processor ID is specified by searching the job management table 100 using the cut-out job ID and virtual processor ID as search keys. In step 5, the actual processor ID of the specified real processor ID is specified. The process indicated by the local process ID extracted in step 3 and expanded in the indicated processor element is specified as a communication destination.

As described above, when the process indicated by the process ID specified as the communication request destination is an in-job process, the process ID specified by the specified process ID and the management data in the job management table 100 are used as shown in FIG. The communication destination is specified.

On the other hand, if it is determined in step 2 that the process indicated by the designated process ID is a non-job process from the read bit value, the process proceeds to step 6
To execute the cutout processing by the bit operation, and the actual processor ID /
Cut out the local process ID.

Subsequently, in step 7, the data is expanded into the processor element indicated by the extracted real processor ID.
The process indicated by the extracted local process ID is specified as a communication destination.

In this way, when the process indicated by the process ID specified as the communication request destination is a non-job process, the communication destination is specified from the specified process ID as shown in FIG.

Process ID specified as communication request destination
Since the process of specifying the communication destination when the process indicated by is a job process can be executed in a 5τ machine cycle, the communication destination can be specified at high speed.

That is, in the process of specifying the communication destination, it is determined whether the specified process ID belongs to the job by referring to the bit value of the first one bit (job determination flag). When judging
This is executed by searching the job management table 100. Whether the bit value of the first one bit is “1” or “0” is executed by comparison processing executed in 3τ, The search of table 100 is performed at 2τ2
Since the processing is executed by specifying the dimension array, the processing can be executed in a total of 5τ machine cycles.

As described above, the process ID used in the computer system equipped with the present invention is the same as the process ID used in the standard UNIX system adopting a single processor configuration.
D is extended.

Therefore, it is possible to take over the abundant assets of the standard UNIX system. Further, a migrating operation (migration operation) from a standard UNIX system having a single processor configuration to a computer system having a parallel processor configuration according to the present invention can be performed, and a computer having a parallel processor configuration having the present invention can be realized. From the system,
It is characterized in that a migrating operation (moving operation) to a standard UNIX system adopting a single processor configuration becomes possible.

That is, when a processor element is added to the computer system equipped with the present invention, the process ID of the process operating on the existing UNIX system is extended (upper two bytes are added as shown in FIG. 7). This makes it possible to move a process operating on the existing UNIX system to a computer system equipped with the present invention. This allows the system to be easily expanded.

When one of the processor elements of the computer system equipped with the present invention goes down, the process I
By deleting the upper 2 bytes of D (there is no need to delete it when the destination ignores it), it becomes possible to move the process running on that processor element to a standard UNIX system. This makes it easy to deal with a system down.

FIG. 11 shows a processing flow of an operation process executed by the computer system equipped with the present invention. FIG. 12 shows a process executed by a standard UNIX system adopting a single processor and cooperating with the computer system equipped with the present invention. 2 illustrates a processing flow of an operation process to be performed.

Next, this processing flow will be described.
When the computer system equipped with the present invention is started, it enters the processing shown in the processing flow of FIG. 11, and when a job (process) is input by a user in accordance with an operation on the job (process), a local process ID is set. Numbering, job determination flag / job ID / virtual processor I
D and attribute information such as job determination flag / real processor ID are appropriately arranged and combined, and the
Job (process) submitted by combining with D
, A 32-bit process ID is assigned.

On the other hand, when a signal is issued by a user or a communication instruction between processes is issued by a user in accordance with an operation on a job (process), attribute information is extracted by bit operation, and a process to be processed is specified from a process ID. Then, control and processing for it are performed.

On the other hand, when saving to the external storage device 2 and restoration from the external storage device 2 are performed in accordance with the operation for the job (process), when saving to the external storage device 2, the job (process) is ) Is saved in the external storage device 2, another job is input according to a user's instruction, and the system is restarted. In the case of restoration from the external storage device 2, it is determined whether the instruction from the user is non-migration restoration (restoration to return to the original processor element) or movement restoration. Attribute information is extracted by the process ID
To identify the process to be processed, perform control and processing on it, and at the time of migration restoration, determine whether the migration destination is the local system or the standard UNIX system. , Bit information to extract attribute information, specify a process to be processed from a process ID, and perform control and processing on the process.

On the other hand, when the user performs a my-grade operation (move operation) in accordance with an operation on a job (process), whether the destination is the own system or not, a standard UN
It is determined whether the system is an IX system. If the destination is the own system, attribute information is extracted by bit operation, a process to be processed is specified from the process ID, and control and processing for the process are performed.

When this processing configuration is adopted, if the destination is a standard UNIX system, FIGS.
Job (process) to be moved, as shown in A of FIG.
To a standard UNIX system. At this time, in a standard UNIX system, a 32-bit process ID is used.
Since only the process ID portion of the lower 16 bits is extracted as the process ID, the job (process) to be moved is moved to the standard UNIX system as it is.

When the UNIX system cooperating with the computer system having the present invention is started, it enters the processing shown in the processing flow of FIG. 12, and the user operates the job (process) according to the operation on the job (process).
Is input, a 16-bit process ID is assigned to the input job (process) by assigning the 16-bit process ID shown in FIG. 7C.

On the other hand, when a signal is issued by a user or a communication instruction between processes is issued according to an operation on a job (process), a process to be processed is specified from a 16-bit process ID, and control for the process is specified. And processing.

On the other hand, when the user performs a my-grade operation (move operation) in accordance with the operation on the job (process), as shown in FIGS. At the time, and at that time, attribute information such as 16-bit job determination flag / job ID / virtual processor ID and job determination flag / real processor ID are appropriately arranged and combined, and added to them by the UNIX system. By combining the transferred 16-bit process ID, a 32-bit process ID is assigned to the job (process) to be moved.

As described above, the computer system provided with the present invention is a standard UNIX system having a single processor configuration.
Data processing can be performed while inheriting the rich assets of the system.

In this embodiment, a configuration is adopted in which processes belonging to a job are distinguished from processes not belonging to the job. However, a method may be adopted in which a process not belonging to a job is regarded as constituting a job. It is. When this method is adopted, the process IDs assigned to the processes are unified to those shown in FIG.

[0069]

As described above, according to the present invention,
In a computer system adopting a parallel processor configuration, it is possible to deal with saving and restoring a job without reserving a process ID while taking over abundant assets of a standard UNIX system as it is.

[Brief description of the drawings]

FIG. 1 is a principle configuration diagram of the present invention.

FIG. 2 is an explanatory diagram of data managed by a management unit.

FIG. 3 is an explanatory diagram of a process ID according to the present invention.

FIG. 4 is an embodiment of the present invention.

FIG. 5 is an embodiment of the present invention.

FIG. 6 is a process flow of a process of assigning a process ID.

FIG. 7 is an embodiment of a process ID according to the present invention.

FIG. 8 is a processing flow of a communication destination specifying process.

FIG. 9 is an explanatory diagram of a communication destination specifying process.

FIG. 10 is an explanatory diagram of a communication destination specifying process.

FIG. 11 is an explanatory diagram of operation processing of the present invention.

FIG. 12 is an explanatory diagram of operation processing of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Processor element 2 External storage device 3 Operating system 10 Management means 11 Registration means 12 Assignment means 13 Identification means

Claims (3)

[Claims] 1. A method comprising: a plurality of processor elements;
In a computer system that executes a job constituted by one or a plurality of processes, a job ID assigned to the job and a processor ID of a virtual processor element assigned to the job
Management means for managing a correspondence relationship between a processor ID of a real processor element executing a job, and a local process defined in the real processor element executing the process when a process configuring the job is generated An allocation means for generating a process ID defined by a bit combination of an ID, the virtual processor ID of the job indicated by the real processor element, and a job ID assigned to the job, and assigning the process ID to the process; A computer system characterized by the following. 2. The computer system according to claim 1, wherein, when a process that does not form a job is generated, the allocating unit includes: a local process ID defined in a real processor element that executes the process; A computer system characterized by generating a process ID defined by a bit combination with a processor ID of a processor element and assigning the process ID to the process. 3. The computer system according to claim 1, wherein when a communication request is issued by specifying a process ID,
A computer system comprising: a specification unit that specifies a process to be a communication destination from the process ID and management data of the management unit.