JPH08305671A - Job scheduling system - Google Patents

Abstract

PURPOSE: To effectively use a system by the scheduling system allowing for the cluster constitution in multi-job circumstances in the hierarchy sharing distributed memory type multi-processor system. CONSTITUTION: Secondary job queuing means 61 to 6n correspond to job assignment units 31 to 3n. A job schedule means 5 refers to a load state inquiry means 7 with respect to the job received from a primary job queuing means 4 to detect the job assignment unit of the lightest load and throws the job into the secondary job queue corresponding to the detected job assignment unit. A processor 11 of each cluster 10 successively processes jobs in the secondary job queuing means corresponding to the job assignment unit to which the cluster belongs. Only when jobs belonging to different job assignment units are processed, they are accessed through a network 2.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a job scheduling method in a hierarchical shared distributed memory multiprocessor system in which a conventional shared memory multiprocessor system is used as one cluster and a plurality of clusters are connected by a network.

[0002]

2. Description of the Related Art Conventional job scheduling methods in a shared memory multiprocessor system include a method having one common job queue in the entire system and a method having each processor having its own job queue.

[0003]

In an attempt to construct a system using more processors than in the past, the above-mentioned 1
When a method with a common job queue for books is adopted, memory access conflicts frequently occur. In addition, the number of processors that access the common job queue via the network is increased, which causes a lot of network access competition. Further, when each processor has its own job queue, the access contention problem for accessing the common job queue is eliminated, but there is a problem that the load state of each processor becomes uneven. In order to solve this problem, there is a method in which a processor that has no jobs to process checks the load status of other processors and moves the job from the processor with a high load and executes it, but the number of processors increases. Therefore, this operation causes a lot of access contention of the network and the common job queue.

In view of the above problems, the present invention provides a shared memory type multiprocessor system as one cluster, divides a plurality of clusters into groups for each job allocation unit, and connects each cluster via a network. An object of the present invention is to provide a job scheduling method by which a memory-type multiprocessor system is configured so that access competition of a network and a common job queue does not easily occur even if many processors are used.

[0005]

According to a job scheduling method of the present invention, a shared memory type multiprocessor system is used as one cluster, a plurality of clusters are divided into groups for each job allocation unit, and each cluster is connected via a network. A job scheduling method of a hierarchical shared distributed memory type multiprocessor system connected by means of a primary job queuing means for managing jobs submitted to the system in FIFO order, and a cluster configuration of the hierarchical shared distributed memory type multiprocessor system. A secondary job queuing unit that manages jobs assigned to each job allocation unit in a FIFO order, one for each job allocation unit that is set in consideration,
From the primary job queuing means, referring to the load status inquiry means that stores the load status of each job allocation unit and the load status that the load status inquiry means stores,
Job scheduling means for appropriately distributing jobs to the next job queuing means.

Further, the job to be submitted to the system can be identified by the job class, the job class is stored by the job class inquiry means, and the job scheduling means sorts the jobs into different job allocation units according to the job class. Is preferred.

[0007]

A secondary job queuing means is associated with each job allocation unit. The job scheduling means refers to the load status inquiry means for the job received from the primary job queuing means, detects the job allocation unit with the least load,
The job is submitted to the secondary job queue corresponding to the detected job allocation unit. The processor of each cluster sequentially processes the jobs of the secondary job queuing means corresponding to the job allocation unit to which the cluster belongs. Only when processing jobs belonging to different job allocation units,
Access via network.

[0008]

Embodiments of the present invention will now be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a job scheduling system of the present invention which employs a hierarchical shared distributed memory type multiprocessor system. Each cluster 10
Has a configuration similar to that of a conventional shared memory multiprocessor system, and includes a plurality of processors 11 and a shared memory 12 shared by the processors 11. Each of the clusters 10 is connected to the network 2 and constitutes a hierarchical shared distributed memory multiprocessor system as a whole. In the hierarchical shared distributed memory multiprocessor system, information transmission between processors belonging to the same cluster is performed via the shared memory 12.
Information transmission between processors belonging to different clusters is performed by communication via the network 2.

The job allocation units 31 to 3n are set in consideration of the cluster configuration of the hierarchical shared distributed memory type multiprocessor system.
Secondary job queuing means 61 to 6n are associated with n. Jobs submitted to the primary job queuing means 4 are processed by the job scheduling means 5 to 2
It is passed to the next job queuing means 61-6n.

At this time, the job scheduling means 5 acquires the load status of each allocation unit 31 to 3n and the job class of the scheduling target job by means of the load status inquiry means 7 and the job class inquiry means 8 and allocates this information to each job. It is used for distribution of jobs into units 31 to 3n. Each job allocation unit 31
Each of the processors 11 to 3n receives jobs from the secondary job queuing means 61 to 6n and processes them.

As shown in FIG. 2, the primary job queuing means 4 comprises a primary job queue 41, a primary job enqueue means 42, and a primary job dequeue means 43. The primary job queue 41 manages jobs submitted to the system in FIFO order. The primary job enqueuing means 42 performs an operation of adding the job submitted to the system to the end of the primary job queue 41.
The primary job dequeuing means 43 uses the primary job queue 41.
Retrieve the job from the beginning of.

Similarly, the secondary job queuing means 61
6n are composed of a secondary job queue 71, a secondary job enqueue means 72, and a secondary job dequeue means 73. The secondary job queue 71 manages the jobs delivered from the job scheduling unit 5 in FIFO order. The secondary job enqueue means 72 performs an operation of adding a job to the end of the secondary job queue 71. The secondary job dequeuing means 73 performs an operation of taking out a job from the head of the secondary job queue 71.

Next, the operation of the job scheduling means 5 will be described with reference to FIG. The job scheduling means 5 takes out from the primary job queue 41 the job that has been submitted to the system and placed in the primary job queue 41 (step S1). Next, in order to select a job allocation unit for executing the fetched job, the load status inquiry unit 7 acquires the load status of each job allocation unit (step S2). In conventional job scheduling where each processor has its own job queue, access conflicts occur because the load status of other processors is checked when the number of processors increases, but in this scheduling method, the unit for checking the load status is the job allocation unit. As a result, access competition is reduced, and system performance degradation can be suppressed. That is, the job allocation unit with the least load is selected (step S3), and the job is submitted to the secondary job queue 71 of the job allocation unit selected by the enqueue means 72 (step S3).
4). On the other hand, each processor retrieves and executes a job from the job queue corresponding to the job allocation unit to which it belongs. In this case, the common area called the secondary job queue 71 is accessed by a plurality of processors.
The processors that access the one common area (secondary job queue 71) are limited to the processors within the job allocation unit. Therefore, it is possible to suppress system performance degradation by reducing access contention as compared with the conventional scheduling method in which the entire system has only one job queue.

Next, the job scheduling method when the job can be identified by the job class is shown in FIG.
And it demonstrates with reference to FIG. The job allocation unit is set in consideration of the cluster configuration, and different job allocation units 301 to 30n are given for each job class. Further, the secondary job queue 91 existing in the secondary job queuing means 81 to 8n corresponding to each job allocation unit is also changed for each job class. For example, the secondary job queue 91 for jobs having low priority is set long. As a result, although the number of jobs waiting to be executed in the queue increases, jobs can be scheduled in units of job allocation for other job classes, and interference with the execution of other jobs can be prevented.

Next, the operation of the job scheduling means 50 will be described with reference to FIGS. The job extracted from the primary job queue 4 in step S11 acquires the job class by using the job class inquiry unit 8 in step S12. In step S13, the load status is acquired from the load status inquiry means 7 for the job allocation unit given to the job class acquired in step S12. Step S1
In 4, the job allocation unit with the least load is selected, and in step S15, the job is submitted to the secondary job queue. Here, since the unit of job allocation for inquiring about the load state is limited to that given to the job class, the access contention that occurs for inquiring about the disabled state is reduced, and the overhead for scheduling is reduced.

[0016]

As described above, according to the present invention, the clusters are divided into job allocation units, and each cluster is connected via the network to form a hierarchical shared distributed memory multiprocessor system. It is possible to reduce the contention of memory access and the contention of network access, and to perform job scheduling that maximizes system performance.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of a job scheduling system of the present invention which employs a hierarchical shared distributed memory type multiprocessor system.

FIG. 2 is a block diagram showing in detail primary and secondary job queuing means of the embodiment of FIG.

FIG. 3 is a flowchart showing a process of a job scheduling unit of the embodiment of FIG.

FIG. 4 is a block diagram showing another embodiment of the present invention in which a job can be identified by a job class.

FIG. 5 is a block diagram showing in detail primary and secondary job queuing means of the embodiment of FIG.

FIG. 6 is a flowchart showing a process of a job scheduling unit in FIG.

[Explanation of symbols]

2 Network 4 Primary Job Queuing Means 5,50 Job Scheduling Means 7 Load Status Inquiring Means 8 Job Class Inquiring Means 10 Clusters 11 Processors 12 Shared Memory 31-3n Job Allocation Unit 41 Primary Job Queues 42, 72, 92 Enqueuing Means 43, 73, 93 Dequeuing means 61-6n, 81-8n Secondary job queuing means 71, 91 Secondary job queue

Claims (3)

[Claims] 1. A hierarchical shared distributed memory multiprocessor system in which a shared memory multiprocessor system is defined as one cluster, a plurality of clusters are divided into groups for each job allocation unit, and each cluster is connected via a network. Job scheduling method that manages jobs submitted to the system in FIFO order 1
One job assignment unit is set in consideration of the next job queuing unit and the cluster configuration of the hierarchical shared distributed memory multiprocessor system, and the jobs assigned to each job assignment unit are managed in FIFO order. 2 Referring to the next job queuing means, the load status inquiry means for storing the load status of each job allocation unit, and the load status stored by the load status inquiry means,
A job scheduling method comprising: a job scheduling unit that appropriately distributes jobs from a primary job queuing unit to a secondary job queuing unit. 2. A job submitted to the system can be identified by a job class, the job class is stored in a job class inquiry unit, and the job scheduling unit sorts the job into different job allocation units depending on the job class. The job scheduling method according to claim 1. 3. The job scheduling method according to claim 2, wherein the job class is at least one of priority, distinction between interactive and batch, and presence / absence of parallel processing.