JP5769224B2 - Data processing system and data processing method - Google Patents

Description

  The present invention relates to a data processing system and a data processing method for managing main memory by virtual memory management.

  In a data processing system that manages main memory by virtual memory management, because the main memory capacity is insufficient, there may be a state in which page replacement between the main memory and the virtual memory device frequently occurs. Such a state is called thrashing. When such thrashing occurs, not only the job being executed simultaneously but also the processing capability of the operating system itself is reduced. In such a thrashing state, there is a possibility that the worst system will be stopped, and it is necessary to stop / end the job being executed or increase the main storage capacity.

  As a system for preventing such a thrashing state, in Patent Document 1, thrashing occurs depending on whether the ratio of the accumulated page fault processing elapsed time to the accumulated job execution elapsed time of a job is larger than a preset threshold value. It has been proposed that if thrashing occurs, the job is recorded in the thrashing mode recording means as a thrashing mode state.

JP 08-272684 A

  As described above, when such thrashing occurs, not only the job being executed simultaneously but also the processing capability of the operating system itself is reduced. When thrashing occurs and the job being executed is stopped / terminated, the job selected as the target to be stopped / terminated is a job in which page replacement processing between the main memory and the virtual storage device is frequently performed. There may be other jobs such as a job with a large main storage capacity used or a job submitted last.

  In Patent Document 1, when a memory (thrashing memory pool) reserved in advance is shared among programs determined to be in a thrashing state, the allocated memory is used for execution, or the allocated memory is insufficient. Does not cause page fault processing by waiting for memory. However, in the case of the one shown in Patent Document 1, the amount of memory allocated to the thrashing job increases while the program using the memory reserved in advance is not terminated because the thrashing state is determined. Therefore, the memory waiting state of the job waiting for the memory is not released. Further, in Patent Document 1, it is not performed to limit the multiplicity of execution by collecting only jobs that are prone to thrashing.

  In view of the above-described problems, an object of the present invention is to provide a data processing system and a data processing method capable of preventing thrashing when the main memory is managed by virtual memory management. And

In order to solve the above-described problem, the present invention provides a data processing apparatus that manages main memory by virtual memory management, a means for detecting a job that frequently occurs, and a job that frequently causes missing. The means for registering as a class and jobs belonging to the freeze job class are limited in accordance with the execution multiplicity described in the freeze class definition information corresponding to the identification information of the freeze job class. Means for executing, means for determining whether or not the free memory capacity is larger than a memory request amount of a job waiting for execution in a job belonging to the freeze job class, and the free memory capacity belongs to the freeze job class. If it is determined that the amount of memory required for a job that is waiting to be executed is greater than the required amount of memory To, for the job that has been executed on the basis of the execution multiplicity managed as freeze job classes, and characterized in that it comprises a means for executing a original job class to delete the registration as the freeze job class To do.

The present invention provides a data processing method in which main memory is managed by virtual memory management, a step of detecting a job that frequently occurs, a step of registering the job that frequently occurs as a freeze job class, and the freeze A step of executing a job belonging to a job class by limiting the execution multiplicity according to the execution multiplicity described in the freeze class definition information corresponding to the identification information of the freeze job class, and a free memory capacity A step of determining whether or not the memory request amount of a job waiting for execution in a job belonging to the freeze job class is larger than the memory request amount, and a job belonging to the freeze job class is waiting for execution If it is determined to be greater than the memory requirements of the job, and freeze job classes For the job which has been performed based on the execution multiplicity and supervises as, characterized in that to remove the registration as the freeze job classes and a step to be executed as the original job class.

  According to the present invention, a job in which page replacement with the virtual storage device frequently occurs because the main storage capacity is insufficient is added to the job class to which the job originally belongs, and the main storage capacity is increased. Due to the lack of main storage capacity by managing even the job class for managing the multiplicity of jobs that frequently cause page replacement with the virtual storage device due to lack of virtual Thrashing can be prevented by suppressing the execution multiplicity of jobs that frequently cause page replacement with the storage device.

1 is a block diagram illustrating a computer system according to a first embodiment of this invention. It is explanatory drawing of the information of an example of a system information provision part. It is explanatory drawing of the information of an example of a policy information provision part. It is explanatory drawing of an example of the information of a job information provision part. It is explanatory drawing of an example of the information of a job class information provision part. It is a flowchart which shows the flow of the process of registration and execution to a freeze job class in the 1st Embodiment of this invention. It is a flowchart which shows the flow of a process of deletion of registration from the freeze job class in the 1st Embodiment of this invention. It is a block diagram which shows the computer system of the 2nd Embodiment of this invention. It is explanatory drawing of the other example of the information of a job information provision part. It is a flowchart which shows the flow of a process of deletion of registration from the freeze job class in the 2nd Embodiment of this invention. It is a flowchart which shows the flow of a process of deletion of registration from the freeze job class in the 2nd Embodiment of this invention.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a computer system according to the first embodiment of this invention. As shown in FIG. 1, a computer system 1 according to the first embodiment of the present invention includes a system information providing unit 10, a policy information providing unit 11, a job information providing unit 12, a job class information providing unit 13, An empty memory detection unit 20, a missing number excess detection unit 21, a job management unit 22, a freeze job class management unit 23, a freeze job class registration / deletion unit 24, and a freeze job class job execution unit 25 are included.

  2 is an explanatory diagram of an example of information of the system information providing unit, FIG. 3 is an explanatory diagram of an example of information of the policy information providing unit, and FIG. 4 is an explanatory diagram of an example of information of the job information providing unit. FIG. 5 is an explanatory diagram of an example of information of the job class information providing unit.

  As shown in FIG. 2, the system information providing unit 10 provides free memory amount information INF101. The free memory amount information INF101 is a main storage capacity that can be allocated to a job.

  As shown in FIG. 3, the policy information providing unit 11 provides missing count threshold information INF111. Missing frequency threshold information INF111 is a missing frequency threshold per unit time of a job, which is determined by the system administrator and should be determined that job missing frequently occurs.

  As shown in FIG. 4, the job information providing unit 12 provides memory request amount information INF121, missing count information INF122 per unit time, and freeze class attribute information INF123 for each job. The memory request amount information INF121 is the amount of memory requested by the job. Missing frequency information INF122 per unit time is the number of missing times per unit time of the job. The freeze class attribute information INF123 is information indicating whether or not the job belongs to the freeze class.

  As shown in FIG. 5, the job class information providing unit 13 provides freeze class definition information INF 131. The freeze class definition information INF 131 has multiplicity information INF 1311 for managing the number of jobs that are frequently executed for page replacement processing between the main storage and the virtual storage device, as determined by the system administrator.

  In FIG. 1, the free memory detection unit 20 monitors the amount of free memory and the memory request amount of a job waiting for execution in a job whose freeze class attribute belongs to the freeze job class. Determine whether the amount is exceeded.

  The missing number excess detection unit 21 monitors the number of missing times per unit time for each job, and determines whether or not the number of missing times per unit time of the job has been exceeded. Detect whether or not

  The job management unit 22 performs job management. More specifically, when the missing number excess detection unit 21 determines that the number of missing times per unit time of the job exceeds the threshold value, the job management unit 22 causes frequent missing in the job. The job is changed to the freeze job class. If the free memory detection unit 20 determines that the free memory capacity is larger than the memory request amount of the job waiting for execution in the job belonging to the freeze job class, the job management unit 22 The job is deleted from the frozen job class and returned to the original job class for execution.

The freeze job class management unit 23 manages freeze job classes. More specifically, when there is a job to be changed to a freeze job class because missing is frequently occurring, the freeze job class management unit 23 remains in the job class to which the job originally belongs and the job is changed to a freeze job. Change it to belong to a class. In addition, when there is a job to be changed to the frozen job class on the assumption that missing frequently occurs, the frozen job class management unit 23 executes the job as a job of the frozen job class.

  The freeze job class registration / deletion unit 24 performs processing of changing a job to a freeze job class and deleting a job from the freeze job class.

  The freeze job class job execution unit 25 executes a job of the freeze job class. When a job of the freeze job class is executed, the job is executed with execution multiplicity suppressed according to the multiplicity information INF 1311 of the freeze class definition information INF 131 of the job class information providing unit 13.

  In the first embodiment of the present invention, the missing number excess detection unit 21 determines whether or not the number of missing times per unit time of the job is larger than a threshold value of the number of missing times per unit time of the job. If it is determined that the number of missing times is greater than the missing number threshold per unit time of the job, the job is changed to a freeze job class and executed. Jobs belonging to the freeze job class are executed with multiplicity according to the definition of the freeze job class.

  Next, the operation of the first exemplary embodiment of the present invention will be described with reference to a flowchart. As described above, in the first embodiment of the present invention, a job that frequently occurs is changed to belong to the freeze job class while belonging to the job class to which the missing job originally belongs, and the definition of the freeze job class is defined. The execution multiplicity is limited according to FIG. 6 is a flowchart showing the flow of registration and execution processing for a freeze job class according to the first embodiment of the present invention.

  In FIG. 6, the missing number excess detection unit 21 always monitors the missing number information INF122 per unit time of the job information providing unit 12 for each job, and the job of the policy information providing unit 11 determined by the system administrator. It is determined whether or not the missing count threshold value information INF111 per unit time is exceeded (step S101). If the missing count excess detection unit 21 determines that the missing count information INF122 per unit time of the job is smaller than the missing count threshold information INF111, the missing job is not frequently generated and the job is left as it is. .

  In step S101, when the missing number excess detection unit 21 determines that the missing number information INF122 per unit time of the job is larger than the missing number threshold information INF111 per unit time of the job, the missing frequently occurs in the job. If the job has occurred, the job management unit 22 is activated to change the job to the freeze job class and execute the job (step S102).

  Next, the job management unit 22 activates the freeze job class management unit 23 in order to change the job to the freeze job class (step S103).

  Next, the freeze job class management unit 23 activates the freeze job class registration / deletion unit 24, records information belonging to the freeze job class in the freeze class attribute information INF123 of the job information providing unit 12, and the job Is changed to belong to the freeze job class (step S104).

  Finally, the freeze job class management unit 23 activates the freeze job class job execution unit 25 to execute the job as a job of the freeze job class (step S105). At this time, since the job is executed according to the definition of the freeze job class, the execution multiplicity is suppressed according to the multiplicity information INF1311 of the freeze class definition information INF131 of the job class information providing unit 13.

  Further, in the first embodiment of the present invention, as described above, the amount of free memory more than that requested by a job waiting for execution due to the restriction of the execution multiplicity in the freeze job class in the main memory. Occurs, the job registered in the freeze job class and waiting to be executed is deleted from the freeze job class and executed in the original job class. FIG. 7 is a flowchart showing the flow of processing for deleting registration from the freeze job class in the first embodiment of the present invention.

  In FIG. 7, the free memory detection unit 20 is waiting to be executed by a job whose free class attribute information INF 123 belongs to the freeze job class in the free memory amount information INF 101 of the system information providing unit 10 and the job information providing unit 12. The memory request amount information INF121 of the job is monitored, and it is determined whether or not the free memory amount information INF101 exceeds the memory request amount information INF121 (step S201). If the free memory detection unit 20 determines that the free memory amount information INF101 is smaller than the memory request amount information INF121 of a job waiting for execution in a job belonging to the freeze job class, the free memory amount information INF101 remains as it is for the job. And

  Next, when it is determined that the free memory amount information INF101 is larger than the memory request amount information INF121 of the job waiting for execution in the job belonging to the freeze job class, the free memory detection unit 20 Is deleted from the freeze job class, and the job management unit 22 is activated to restore the original job class to be executed (step S202).

  Next, the job management unit 22 activates the freeze job class management unit 23 to delete the job from the freeze job class (step S203).

  Finally, the freeze job class management unit 23 activates the freeze job class registration / deletion unit 24 and records information indicating that it does not belong to the freeze job class in the freeze class attribute information INF123 of the job class information providing unit 13, The class to which the job belongs is changed only to the original job class (step S204). At this time, the job is executed according to the definition of the original job class.

  As described above, in the first embodiment of the present invention, a freeze job class is configured only by a job group in which missing frequently occurs, and multiplicity is suppressed for a job in which missing frequently occurs. Thrashing is prevented. For this reason, thrashing can be prevented without affecting jobs other than jobs that frequently occur.

  Next, a second embodiment of the present invention will be described. FIG. 8 is a block diagram showing a computer system according to the second embodiment of this invention. As shown in FIG. 8, a computer system 201 according to the second embodiment of the present invention includes a system information providing unit 210, a policy information providing unit 211, a job information providing unit 212, a job class information providing unit 213, Free memory detection unit 220, missing number excess detection unit 221, job management unit 222, freeze job class management unit 223, freeze job class registration / deletion unit 224, freeze job class job execution unit 225, page out An information recording / reading unit 226 and a page-in control unit 227 are included.

  System information providing unit 210, policy information providing unit 211, job information providing unit 212, job class information providing unit 213, free memory detecting unit 220, missing number excess detecting unit 221, job managing unit 222, freeze job class managing unit 223, The freeze job class registration / deletion unit 224 and the freeze job class job execution unit 225 are the system information providing unit 10, the policy information providing unit 11, the job information providing unit 12, and the job class information providing unit according to the first embodiment of the present invention. 13, a free memory detection unit 20, a missing number excess detection unit 21, a job management unit 22, a freeze job class management unit 23, a freeze job class registration / deletion unit 24, and a freeze job class job execution unit 25. .

  However, as shown in FIG. 9, the job information providing unit 212 has page-out information INF124 in addition to the information of the job information providing unit 12 (see FIG. 4) of the first embodiment. In addition to the first embodiment, the second embodiment of the present invention further includes a page-out information recording / reading unit 226 and a page-in control unit 227.

  In the second embodiment of the present invention, a job in which missing frequently occurs is changed to belong to the freeze job class while belonging to the job class to which it originally belongs, and the execution multiplicity is defined according to the definition of the freeze job class. Is limited. FIG. 10 is a flowchart showing the flow of processing for deleting registration from the freeze job class in the second embodiment of the present invention.

  In FIG. 10, the missing number excess detection unit 221 always monitors the missing number information INF122 per unit time of the job information providing unit 212 for each job, and the policy information providing unit 211 determined by the system administrator It is determined whether or not the number of missing times threshold information INF111 per unit time has been exceeded (step S301). If it is determined that the missing count information INF122 per unit time of the job is smaller than the missing count threshold information INF111, the missing count excess detection unit 221 determines that missing has not occurred frequently and keeps the job as it is. To do.

  Next, when it is determined that the missing count information INF122 per unit time of the job is larger than the missing count threshold information INF111 per unit time of the job, the missing number excess detection unit 221 frequently performs missing in the job. Assuming that the job has occurred, the job management unit 222 is activated to change the job to the freeze job class and execute the job (step S302).

  Next, the job management unit 222 activates the freeze job class management unit 223 to change the job to the freeze job class (step S303).

  Next, the freeze job class management unit 223 activates the freeze job class registration / deletion unit 224, records information belonging to the freeze job class in the freeze class attribute information INF123 of the job information providing unit 212, and the job Is changed to belong to the freeze job class (step S304).

  Next, the freeze job class management unit 223 activates the freeze job class job execution unit 225 to execute the job as a job of the freeze job class (step S305). At this time, since the job is executed according to the definition of the freeze job class, the job is executed according to the multiplicity information INF 1311 of the freeze class definition information INF 131.

  Finally, the page-out information recording / reading unit 226 outputs the paged-out information to the page-out information INF 124 of the job information providing unit 212 for the job that is waiting for the execution multiplicity when the freeze job class is set. Recording is performed (step S306).

  Further, in the second embodiment of the present invention, when a free memory amount larger than that requested by a job waiting for execution in the freeze job class due to the restriction of the execution multiplicity is generated in the main memory, the freeze is generated. Jobs registered in the job class and waiting to be executed are deleted from the frozen job class and executed in the original job class. FIG. 11 is a flowchart showing the flow of processing for deleting registration from the freeze job class in the second embodiment of the present invention. .

  In FIG. 11, the free memory detecting unit 220 is a job waiting for execution in the free memory amount information INF 101 of the system information providing unit 210 and a job whose freeze class attribute information INF 123 belongs to the freeze job class in the job information providing unit 12. The memory request amount information INF121 is monitored, and it is determined whether or not the free memory amount information INF101 exceeds the memory request amount information INF121 (step S401). If the free memory amount detection unit 220 determines that the free memory amount information INF101 is smaller than the memory request amount information INF121 of a job waiting for execution in a job belonging to the freeze job class, the free memory amount information INF101 is left as it is for the job. And

  Next, when it is determined that the free memory amount information INF101 is larger than the memory request amount information INF121 of the job waiting for execution in the job belonging to the freeze job class, the free memory detection unit 220 determines that the job Is deleted from the freeze job class, and the job management unit 222 is activated (step S402).

  Next, the job management unit 222 activates the page-in control unit 227 before execution, and activates the freeze class registration / deletion unit 224 to delete the job from the freeze job class (step S403).

Next, the freeze job class management unit 223 activates the page-out information recording / reading unit 226 to acquire the page-out information INF124 of the job that has been waiting for multiplicity in the freeze job class from the job information providing unit 212. The page-out information is read (step S404).

  Next, the page-in control unit 227 uses the page-out information INF 124 acquired by the page-out information recording / reading unit 226 to perform page-in of a page necessary for the job asynchronously (step S405).

  Finally, the freeze job class registration / deletion unit 224 records information that does not belong to the freeze job class in the freeze class attribute information INF123 of the job information providing unit 212, and changes the class to which the job belongs to only the original job class. (Step S406). At this time, the job is executed according to the definition of the original job class.

  As described above, in the second embodiment of the present invention, asynchronous page-in processing is performed before execution. Therefore, when a job that is in the execution waiting state in the freeze job class is set to the execution state, It is possible to suppress the occurrence of missing immediately after execution, which occurs due to the main memory being used being paged out.

1, 201: Computer system 10, 210: System information providing unit 11, 211: Policy information providing unit 12, 212: Job information providing unit 13, 213: Job class information providing unit 20, 220: Memory detecting unit 21, 221: Missing number excess detection unit 22, 222: job management unit 23, 223: freeze job class management unit 24, 224: freeze job class registration / deletion unit 25.225: freeze job class job execution unit 226: record / read page-out information Unit 227: Page-in control unit

Claims (4)

In a data processing apparatus that manages main memory by virtual memory management,
Means for detecting jobs that frequently occur,
Means for registering a job that frequently occurs as a freeze job class;
Means for executing a job belonging to the freeze job class with the execution multiplicity being limited according to the execution multiplicity described in the freeze class definition information corresponding to the identification information of the freeze job class;
Means for determining whether or not the free memory capacity is larger than the memory request amount of the job waiting for execution in the job belonging to the freeze job class;
When it is determined that the free memory capacity is larger than the memory request amount of a job waiting for execution in a job belonging to the freeze job class, it is managed as a freeze job class and executed based on the execution multiplicity. A data processing system comprising: means for deleting the registration as the freeze job class and executing the job as the original job class. In addition, it has means for recording paged-out information for jobs that are waiting for multiplicity of execution when it is set as a freeze job class,
The data processing according to claim 1, wherein when the job is returned from the freeze job class to the original job class and executed, the job is paged in based on the recorded page-out information. system.   The means for detecting a job in which missing frequently occurs detects whether or not missing frequently occurs in the job by determining whether or not the missing frequency threshold per unit time is exceeded for each job. The data processing system according to claim 1 or 2, characterized by the above. In a data processing method for managing main memory by virtual memory management,
Detecting a job that frequently occurs,
Registering the job that frequently occurs as a freeze job class;
The job belonging to the freeze job class is executed with the execution multiplicity limited according to the execution multiplicity described in the freeze class definition information corresponding to the identification information of the freeze job class;
Determining whether the free memory capacity is larger than the memory requirement of a job waiting to be executed in a job belonging to the freeze job class;
When it is determined that the free memory capacity is larger than the memory request amount of a job waiting for execution in a job belonging to the freeze job class, it is managed as a freeze job class and executed based on the execution multiplicity. And a step of deleting the registration as the freeze job class and executing the job as the original job class.

Images