JP7355456B2 - CPU resource management device - Google Patents

Description

The present invention relates to a CPU resource management device, a CPU resource management method, and a program.

In recent computer systems, system administrators set workload classes, which are classes to which processes belong, and allocate available CPU resources to each workload class. In such a computer system, if resources are not allocated appropriately to each workload class, system performance may drop significantly and system operation may be hindered. Therefore, techniques have been disclosed for devising CPU resource allocation to each workload class.

For example, in Patent Document 1, according to the distribution ratio of CPU resources to be allocated to each workload class, the maximum CPU usage time to allocate CPU resources to each workload class is set at each fixed time interval, and this maximum CPU A technique is described that allocates CPU resources to processes belonging to each workload class according to usage time. For example, if there are three workload classes a, b, and c, and the CPU resource distribution ratio is 4:3:3 and the interval time is 100ms, then in one interval, workload classes a, b, and c A maximum of 40 ms, 30 ms, and 30 ms of CPU usage time will be allocated to each process belonging to .

Patent No. 4427802

By the way, if there is a low-load workload class that does not require the allocated maximum CPU usage time within the interval, it is desirable to end the interval earlier by the excess CPU time and start the next interval. For example, workload class a is in a high load state and requires more CPU resources than the allotted maximum CPU usage time, while workload classes b and c are in a low load state and require more CPU resources than the allotted maximum CPU usage time. It shall not require any usage time. At this time, if workload classes b and c use only 15ms of the maximum CPU usage time of 30ms allocated to them, the interval will end early by the 30ms of extra CPU time and the next interval will start. do. As a result, the previous interval is effectively shortened to 70 ms. If a similar situation occurs in the next interval, the next interval will also be effectively shortened to 70 ms.

In this way, by shortening the interval by the surplus CPU time of a low-load workload class that does not require the allocated maximum CPU usage time, you can reduce the interval by the surplus CPU time of a low-load workload class that does not require the allocated maximum CPU usage time. can improve turnaround time for specific workload classes. However, as the interval is shortened, the number of process switches per unit time increases, resulting in a decrease in throughput.

An object of the present invention is to provide a CPU resource management device that solves the above-mentioned problems.

A CPU resource management device according to one embodiment of the present invention includes:
A CPU resource management device comprising a storage device that stores a process, and a processor that executes the process,
The processor includes:
When an interval starts, CPU resources are allocated and executed to processes belonging to a workload class that satisfies the condition that the CPU usage time is less than or equal to the maximum CPU usage time, and when there are no more processes that belong to a workload class that satisfies the condition. Execution control means for recording the number of times CPU resource allocation is aborted due to reaching the maximum CPU usage time and the total allocation time that is the time from the start to the end of the interval. and,
If the total allocation time is shorter than the base interval time and the number of times the CPU resource allocation is aborted is 1 or more, the next interval time is determined to be longer than the previous interval time, and adjustment means for determining the next maximum CPU usage time for each workload class based on the distribution ratio of CPU usage time for each workload class;
It is configured to include.

Further, a CPU resource management method according to another embodiment of the present invention includes:
A CPU resource management method executed by a computer including a storage device that stores a process and a processor that executes the process,
When an interval starts, CPU resources are allocated and executed to processes belonging to a workload class that satisfies the condition that the CPU usage time is less than or equal to the maximum CPU usage time, and when there are no more processes that belong to a workload class that satisfies the condition. and record the number of times CPU resource allocation was aborted due to reaching the maximum CPU usage time and the total allocation time that is the time from the start to the end of the interval;
If the total allocation time is shorter than the base interval time and the number of times the CPU resource allocation is aborted is 1 or more, the next interval time is determined to be longer than the previous interval time, and determining the next maximum CPU usage time for each workload class based on the CPU usage time distribution ratio for each workload class;
It is configured as follows.

Further, a program according to another embodiment of the present invention is
A computer including a storage device that stores a process, and a processor that executes the process,
When an interval starts, CPU resources are allocated and executed to processes belonging to a workload class that satisfies the condition that the CPU usage time is less than or equal to the maximum CPU usage time, and when there are no more processes that belong to a workload class that satisfies the condition. terminating the interval and recording the number of times CPU resource allocation was aborted due to reaching the maximum CPU usage time and the total allocation time that is the time from the start to the end of the interval;
If the total allocation time is shorter than the base interval time and the number of times the CPU resource allocation is aborted is 1 or more, the next interval time is determined to be longer than the previous interval time, and determining the next maximum CPU usage time for each workload class based on the CPU usage time distribution ratio for each workload class;
is configured to perform the following.

By having the configuration as described above, the present invention can improve the throughput of a computer system.

FIG. 1 is a block diagram of a CPU resource management device according to a first embodiment of the present invention. FIG. 3 is a diagram illustrating a configuration example of a workload definition in the CPU resource management device according to the first embodiment of the present invention. FIG. 3 is a diagram illustrating a configuration example of an executable process queue in the CPU resource management device according to the first embodiment of the present invention. FIG. 6 is a diagram illustrating a configuration example of a determination rule used in the next interval time review process (1) in the CPU resource management device according to the first embodiment of the present invention. FIG. 6 is a diagram showing a calculation formula used in the next interval time review process (1) in the CPU resource management device according to the first embodiment of the present invention. FIG. 7 is a diagram showing a conditional expression used in the next interval time review process (2) in the CPU resource management device according to the first embodiment of the present invention. FIG. 7 is a diagram showing a calculation formula used in the next interval time review process (2) in the CPU resource management device according to the first embodiment of the present invention. 3 is a flowchart showing the operation of the CPU resource management device according to the first embodiment of the present invention. FIG. 2 is an explanatory diagram of the effects of the CPU resource management device according to the first embodiment of the present invention. FIG. 2 is a block diagram of a CPU resource management device according to a second embodiment of the present invention. FIG. 3 is a block diagram of a CPU resource management device according to a third embodiment of the present invention.

Next, embodiments of the present invention will be described in detail with reference to the drawings.
[First embodiment]
FIG. 1 is a block diagram of a CPU resource management device 100 according to a first embodiment of the present invention. Referring to FIG. 1, the CPU resource management device 100 includes a storage device 110 and a CPU 120 connected to the storage device 110.

The storage device 110 is composed of a hard disk, a memory, etc., and is configured to store processing information and programs 114 necessary for various processes executed by the CPU 120. The program 114 is a program that implements various processing units by being read and executed by the CPU 120, and is connected to an external device (not shown) or a storage medium (not shown) via a data input/output function such as a communication interface unit (not shown). (not shown) is read in advance and stored in the storage device 110. The processing information stored in the storage device 110 includes a workload definition 111, an executable process queue 112, and dispatching information 113.

The workload definition 111 represents a class definition that defines the class (workload class) to which a process (or job) belongs, and a distribution definition that defines the distribution ratio (or distribution rate) of CPU resources to be allocated to each workload class. It is information. The workload definition 111 is mainly set by a system administrator. FIG. 2 is a configuration example of the workload definition 111. The class definition of this workload definition 111 defines that there are three word load classes named a, b, and c, respectively. Word load classes are mainly divided into business units. Further, in the distribution definition of this workload definition 111, it is defined that the distribution ratio of CPU resources for workload classes a, b, and c is 4:3:3.

The executable process queue 112 is a storage means in which executable processes are queued. FIG. 3 shows an example of the configuration of the executable process queue 112. This executable process queue 112 stores information about the process number of the queued process and the workload class to which it belongs. The executable process queue 112 may store information such as the execution priority of each process in addition to the process number and the workload class to which it belongs. Furthermore, the workload class to which each process belongs may be stored in a location other than the executable process queue 112.

The dispatching information 113 is information used to control the allocation of CPU resources to processes. Referring to FIG. 1, the dispatching information 113 includes a base interval time 131, average process switching time 132, threshold value 133, next interval time 134, next maximum CPU usage time for each workload class 135, and previous interval time 136. , the previous total allocation time 137, and the previous CPU resource allocation abort count 138.

The base interval time 131 is a basic interval time and is a fixed value. For example, the base interval time 131 may be 100 ms.

The process switching average time 132 is an average value of the time required to switch a process executed on the CPU 120 to another process. The process switching average time 132 is the average time required for one process switching.

The threshold value 133 is a threshold value used in determination of the next interval time review process (2), which will be described later.

The next interval time 134 is the interval time to be used in the next interval. This next interval time 134 is a variable value.

The next maximum CPU usage time for each workload class 135 is the maximum CPU usage time for each workload class in the next interval. The maximum CPU usage time is also referred to as allocation time.

The previous interval time 136 is the interval time used in the previous interval.

The previous total allocation time 137 is the time from the start to the end of the previous interval.

The previous CPU resource allocation abort count 138 is the number of times CPU resource allocation was aborted due to reaching the maximum CPU usage time from the start to the end of the previous interval.

The CPU 120 is composed of an arithmetic processing unit and its peripheral circuits, and is configured to read the program 114 from the storage device 110 and execute it, thereby causing the hardware and the program 114 to work together to implement various processing units. ing. The processing section implemented by the CPU 120 includes an initialization means 121, an execution control means 122, and an adjustment means 123.

The initialization means 121 is configured to initialize the workload definition 111 and the dispatching information 113. Specifically, the initialization unit 121 receives the workload definition 111 from the system administrator via a data input/output function such as a communication interface unit (not shown), and stores it in the storage device 110 . The initialization means 121 also sets a predetermined fixed time (for example, 100 ms) to the base interval time 131. Further, the initialization unit 121 sets the average time required for one process switching (for example, 10 μs) determined by prior measurement to the process switching average time 132 . The initialization unit 121 also sets the threshold value 133 to a predetermined value (for example, 5%). The initialization means 121 also sets the next interval time 134 to the same time as the base interval time 131. In addition, the initialization means 121 distributes the CPU usage time for each workload class described in the workload definition 111 to the interval time set as the next interval time 134 to the maximum CPU usage time 135 for each next workload class. Set the maximum CPU usage time for each workload class calculated by distributing it according to the ratio. The initialization unit 121 also sets an initial value (for example, 0) to the previous CPU resource allocation abort count 138. However, if the previous history is left in the previous interval time 136, the previous total allocation time 137, and the previous CPU resource allocation abort count 138, the initialization means 121 does not initialize them. Good too.

The execution control means 122 is configured to allocate CPU resources to processes and execute them. Specifically, the execution control unit 122 reads the dispatching information 113 from the storage device 110 and, according to the dispatching information 113, determines the interval at which processes of each workload class queued in the executable process queue 112 are executed. Start. Furthermore, when the interval starts, the execution control means 122 resets the CPU usage time and the number of CPU resource allocation aborts for each workload class, which are internal variables, to zero, and starts subtracting the timer set to the next interval time 134. do. The execution control unit 122 also controls processes belonging to workload classes whose CPU usage time after the start of the interval is less than or equal to the maximum CPU usage time set in the next maximum CPU usage time 135 for each workload class to an executable process. It is taken out from the queue 112, assigned to the CPU 120, and executed. Furthermore, the execution control unit 122 adds the CPU usage time of the workload class to which the process belongs, depending on the execution time of the process. Furthermore, if the CPU usage time of the workload class to which the process belongs exceeds its maximum CPU usage time 135 while the process is being executed, the execution control means 122 terminates the allocation of CPU resources to the process and executes the process. It is stored in the possible process queue 112. At this time, the execution control means 122 increments the number of times CPU resource allocation is aborted. Furthermore, when an event occurs that causes a process that is being executed to release itself from the CPU (such as an input/output request), the execution control unit 122 terminates the allocation of CPU resources to the process, and places the process in an execution queue (not shown) or the like. Store. At this time, the execution control means 122 does not increment the number of times the CPU resource allocation is aborted because the reason for aborting the CPU resource allocation is not due to the maximum CPU usage time.

On the other hand, the execution control unit 122 determines that processes belonging to workload classes whose CPU usage time after the start of the interval is less than or equal to the maximum CPU usage time set in the next maximum CPU usage time 135 for each workload class are executable. If it does not exist in the process queue 112, the timer that set the next interval time 134 stops decrementing, and the current interval ends. In addition, the execution control means 122 determines that a process belonging to a workload class whose CPU usage time after the start of the interval is less than or equal to the maximum CPU usage time set in the next maximum CPU usage time 135 for each workload class is an executable process. Even if the data exists in the queue 112, when the value of the timer that sets the next interval time 134 becomes zero and times out, the current interval ends.

Further, when the interval ends, the execution control unit 122 performs the following update process on the dispatching information 113, and then writes back the updated dispatching information 113 to the storage device 110. First, the execution control means 122 stores the time set as the next interval time 134 in the above-mentioned interval time 136. In addition, the execution control means 122 calculates the time from the start to the end of the current interval as the total allocated time by subtracting the value of the timer that has stopped or expired from the next interval time 134, and calculates the time from the start to the end of the current interval as the total allocated time. 137. For example, if the next interval time 134 is 100ms, the timer starts subtracting from 100ms when the interval starts, so if the interval ends when the timer value is 30ms, the total allocated time will be 100ms - 30ms = 70ms. Become. Furthermore, the execution control unit 122 stores the number of CPU resource allocation aborts counted from the start to the end of the interval in the previous CPU resource allocation abort count 138.

The adjustment means 123 is configured to adjust the next interval time and the next maximum CPU usage time for each workload class based on the execution status of the process of each workload class in the previous interval by the execution control means 122. ing. Specifically, the adjustment means 123 reads out the dispatching information 113 updated by the execution control means 122 from the storage device 110, and reviews the next interval time as follows.

<Next interval time review process (1)>
First, the adjusting means 123 compares the previous total allocated time 137 and the base interval time 131, and determines whether the previous total allocated time 137 is longer, the same as, or shorter than the base interval time 131. The adjustment unit 123 also determines whether the previous CPU resource allocation abort count 138 is 1 or more, that is, whether there is a workload class for which CPU resource allocation has been aborted. Next, the adjusting means 123 determines whether to change the next interval time from the previous interval time 136 or not, based on the results of the above two determinations and a predetermined determination rule.

FIG. 4 is a configuration example of the determination rule. In this judgment rule, in the following two cases, the next interval time will be changed from the previous interval time, and in other cases, the next interval time will not be changed and will be the same as the previous interval time. .
<Case 1>
There is a workload class for which CPU resource allocation has been discontinued, and the previous total allocation time is shorter than the base interval.
<Case 2>
The previous total allocation time is longer than the base interval. It does not matter whether there is a workload class for which CPU resource allocation has been discontinued.

If the adjustment unit 123 determines in the above determination that the next interval time is to be changed from the previous interval time, in case 1, the adjustment unit 123 changes the next interval time to a time longer than the previous interval time, and in case 2. In this case, change the next interval time to a shorter time than the previous interval time. Specifically, the adjusting means 123 calculates the next interval time according to calculation formula (1) shown in FIG. That is, the adjusting means 123 sets the time obtained by multiplying the previous interval time 136 by the ratio of the base interval time 131 to the previous total allocated time 137 as the next interval time.

<Next interval time review process (2)>
The adjusting means 123 executes the next interval time review process (2) following the next time interval review process (1). In the next interval time review process (2), the adjusting means 123 first determines whether conditional expression (2) shown in FIG. 6 is satisfied. That is, the adjusting means 123 determines whether the ratio of the total time (process switching average time x n) due to process switching performed due to reaching the maximum CPU usage time to the base interval time is equal to or greater than the threshold value. do. Here, in conditional expression 2 and calculation expression 3, which will be described later, n is the number of times 138 of previous CPU resource allocation was aborted.

Next, if conditional expression 2 does not hold, the adjusting means 123 determines the next interval time determined in the next interval time review process (1) as the final next interval time. On the other hand, when conditional expression 2 is satisfied, the adjusting means 123 adjusts the next interval time determined in the next interval time review process (1) so that it becomes even longer. Specifically, the adjusting means 123 calculates the next interval time according to formula (3) shown in FIG. That is, the adjusting means 123 determines, in the next interval time determined in the next interval time review process (1), the proportion of the total time due to process switching that was performed due to reaching the maximum CPU usage time with respect to the base interval time. The value obtained by multiplying the value by the threshold value 133 is set as the final next interval time.

After determining the final next interval time, the adjusting means 123 sets the final next interval time as the next interval time 134 of the dispatching information 113. Further, the adjustment means 123 calculates the maximum CPU usage for each workload class by distributing the interval time set as the next interval time 134 according to the CPU usage time distribution ratio for each workload class described in the workload definition 111. A time 135 is set in the dispatching information 113. Then, the adjusting means 123 writes back the updated dispatching information 113 to the storage device 110.

FIG. 8 is a flowchart showing the operation of the CPU resource management device 100 according to this embodiment. The operation of the CPU resource management device 100 according to this embodiment will be described below with reference to FIG. 8.

When the CPU resource management device 100 is started, the initialization means 121 operates to perform initialization (step S1), and then the execution control means 122 controls the execution of processes at one interval (step S2). Furthermore, each time one interval is ended by the execution control means 122, the adjustment means 123 adjusts the next interval time and the maximum CPU usage time for each workload class as soon as possible based on the situation of the previous interval. (Step S3). Once the processing of the adjustment means 123 is finished, the execution control means 122 again starts an interval based on the adjusted interval time and the CPU usage time for each workload class to control the execution of the process. (Step S2). This loop of steps S2 and S3 is continuously executed until the system operation is stopped.

In the initialization of step S1, the initialization means 121 initializes the workload definition 111 and dispatching information 113 on the storage device 1100 as described above.

In the execution control in step S2, the execution control means 122 executes the processes of each workload class queued in the executable process queue 112 according to the dispatching information 113. Specifically, when an interval is started, the execution control means 122 allocates CPU resources to a process belonging to a workload class whose CPU usage time after the start of the interval is less than or equal to the next maximum CPU usage time of 135 for each workload class. Assign and execute. Furthermore, the execution control unit 122 ends the interval when there is no longer a process belonging to the workload class whose CPU usage time after the start of the interval is less than or equal to the maximum CPU usage time. Furthermore, at the end of the interval, the execution control means 122 determines the number of times CPU resource allocation is aborted due to reaching the maximum CPU usage time between the start and end of the interval, and the time from the start to the end of the interval. The total allocation time is set to the previous CPU resource allocation abort count 138 and the previous total allocation time 137.

In the adjustment in step S3, the adjusting means 123 performs the above-described next interval time review process (1) and The next interval time 134 is determined by the time review process (2), and the next maximum CPU usage time 135 for each workload class is determined based on the determined next interval time.

Next, the effects of the CPU resource management device 100 according to this embodiment will be explained.

For example, if there are three workload classes a, b, and c, and the CPU resource allocation ratio is 4:3:3, and the interval time and base interval time are 100 ms, then the Thus, each process belonging to workload classes a, b, and c is allocated a maximum CPU usage time of 40 ms, 30 ms, and 30 ms, respectively. However, if workload classes a and b have a low load and workload class c has a high load, there will be a surplus in the maximum CPU usage time allocated to workload classes a and b, for example, as shown in the middle part of Figure 9. On the other hand, in workload class c, CPU resource allocation is discontinued because the maximum CPU usage time has been reached.

In the example of FIG. 9, the low-load workload classes a and b use only 30 ms and 20 ms, respectively, and 10 ms of the maximum CPU usage time is unused. On the other hand, in a high-load workload class, CPU resource allocation is terminated after the maximum CPU usage time of 30 ms is used. As a result, the total allocated time is 30+20+30=80ms.

In the next interval time review process (1) performed by the adjusting means 123 after execution of such an interval, if the previous total allocation time of 80 ms is smaller than the base interval time of 100 ms, and the CPU resource allocation is terminated. Since the workload class c exists, it is determined based on the determination rule in FIG. 4 that the next interval time should be changed from the previous interval time. Then, the adjusting means 123 calculates 125 ms as the next interval time by substituting 100 ms for the previous interval time, 100 ms for the base interval time, and 80 ms for the previous total allocated time on the right side of formula 1 in FIG. do. Next, the adjustment means 123 determines whether the conditional expression of FIG. 6 in the next interval time review process (2) is satisfied or not, but assuming that it is not satisfied here, the next interval time will be It is determined to be 125ms.

Next, the adjusting means 123 allocates the next interval time of 125 ms according to the allocation ratio, so that the next maximum CPU usage time for each workload class is set as the maximum CPU usage time for workload classes a, b, Assign 50ms, 38ms, and 38ms to c, respectively. Then, the execution control means 122 executes the next interval under the conditions of the next interval time of 125 ms and the next maximum CPU usage time of workload classes a, b, and c of 50 ms, 38 ms, and 38 ms. Start.

If there is no change in the load status of workload classes a, b, and c, in the next interval, workload classes a and b will not use all of the maximum CPU usage time of 50ms and 38ms, and some of them will remain unused. . On the other hand, workload class c uses the entire maximum CPU usage time of 38 ms. As a result, the expected actual CPU usage time for the next interval is 37ms (30x125/100) for workload class a and 25ms (20x125/100) for workload class b, as shown in Figure 9. /100), workload class c has a maximum CPU usage time of 38 ms, and the total (that is, total allocated time) is 100 ms. In other words, the adjusting means 123 sets the interval time longer based on the total allocated time of the previous interval so that the total allocated time of the next interval becomes the base interval time, and based on this, This means determining the maximum CPU usage time for each workload class.

The number of process switches in the previous and current intervals is 3, so if you combine the previous and current intervals, a total of 6 process switches will be performed in a period of 80ms + 100ms = 180ms, and the number of processes per unit time will be 3. The number of switching times is 6/180=0.0333.

On the other hand, if the next interval time is not adjusted and the interval time is fixed at 100 ms, the total allocated time for the next interval will be 80 ms, the same as the previous interval, as shown by the broken line in the lower part of FIG. Therefore, combining the previous and current intervals, a total of 6 process switches will be performed in a period of 80ms + 80ms = 160ms, and the number of process switches per unit time is 6/160 = 0.0375 (>0. 333).

By dynamically adjusting the next interval time according to the CPU usage status of each workload class as described above, it is possible to reduce the number of process switching times per unit time. As a result, the overhead of process switching is reduced and the throughput of the entire system is improved.

However, if the next interval time is longer than the base interval time, the waiting time until CPU resources are allocated to each workload class becomes longer, resulting in a side effect of longer response time. Therefore, according to the determination rule in FIG. 4 and calculation formula 1 in FIG. 5, if the previous total allocation time is longer than the base interval, the adjustment means 123 changes the next interval time to a time shorter than the previous interval time. By doing so, the above side effects are suppressed.

Furthermore, as the number of workload classes increases, the number of process switching within an interval increases, so the overhead due to process switching increases. Therefore, based on the conditional expression in FIG. 6, the adjustment unit 123 prevents the throughput from decreasing by further lengthening the next interval time when the overhead due to process switching exceeds the threshold value.

In the above description, the dispatching information 113 was configured to store the previous interval time 136 and the previous total allocation time 137. However, the dispatching information 113 may be configured to store a history of the interval time and total allocated time for the previous multiple times. Further, the adjusting means 123 uses the average value of the previous interval time and total allocated time stored in the dispatching information 113 as the previous interval time and the previous total allocated time on the right side of calculation formula 1. You can also do this.

[Second embodiment]
FIG. 10 is a block diagram of a CPU resource management device 200 according to the second embodiment of the present invention. Referring to FIG. 10, the CPU resource management device 200 includes a storage device 210 and a plurality of CPUs 220-0 to 220-m connected to the storage device 210.

The storage device 210 is composed of a hard disk, memory, etc. shared by a plurality of CPUs 220-0 to 220-m, and stores processing information and programs 214-0 to 214-0 necessary for various processes executed by the CPUs 220-0 to 220-m. 214-1. The programs 214-0 to 214-1 are programs that implement various processing units by being read and executed by the CPUs 220-0 to 220-m, and are implemented through data input/output functions such as a communication interface unit (not shown). The information is read in advance from an external device (not shown) or a storage medium (not shown) and stored in the storage device 210. The processing information stored in the storage device 210 includes a workload definition 211, an executable process queue 212, and m pieces of dispatching information 213-1 to 213 that correspond one-to-one to the CPUs 220-1 to 220-m. - There is m. These are configured similarly to the workload definition 111, executable process queue 112, and dispatching information 113 described with reference to FIG.

The CPU 220-0 is composed of an arithmetic processing unit and its peripheral circuits, and by reading and executing the program 214-0 from the storage device 210, the CPU 220-0 causes the hardware and the program 214-0 to work together to operate various processing units. It is configured to accomplish this. The processing section implemented by the CPU 220-0 includes an initialization means 221 and an adjustment means 223. These are configured similarly to the initialization means 121 and adjustment means 123 described with reference to FIG. That is, the initialization means 221 is configured to initialize the workload definition 211 and the dispatching information 213-1 to 213-m. Further, the adjusting means 223 determines the next interval time and the maximum CPU usage time for each workload class based on the total allocation time and the number of aborts of CPU resource allocation in the previous interval stored in the dispatching information 213-1 to 213-m. is configured to update the dispatching information 213-1 to 213-m.

The CPUs 220-1 to 220-m are composed of an arithmetic processing unit and its peripheral circuits, and read and execute the storage device 210 to the program 214-1, thereby causing the hardware and the program 214-1 to cooperate. It is configured to implement various processing units. The processing units realized by the CPUs 220-1 to 220-m include execution control means 222-1 to 222-m. These are configured similarly to the execution control means 122 described with reference to FIG. That is, the execution control means 222-1 to 222-m execute the processes of each workload class queued in the executable process queue 112 according to the corresponding dispatching information 213-1 to 213-m. It is configured.

Next, referring to FIG. 8, the operation of the CPU resource management device 200 according to this embodiment will be described.

When the CPU resource management device 200 is started, the initialization means 221 first operates to initialize the workload definition 211 and each dispatching information 213-11 to 213-m (step S1). Next, each execution control means 222-1 to 222-m starts an interval according to the corresponding dispatching information 213-1 to 213-m, and transfers the processes stored in the executable process queue 212 to the CPUs 220-1 to 213-m. 220-m (step S2). Further, each time an interval is ended by any execution control means 222-i (i is any one from 1 to m), the adjustment means 223 immediately adjusts the dispatch corresponding to that execution control means 222-i. The next interval time and the maximum CPU usage time for each workload class are adjusted based on the previous interval status (previous total allocation time, previous CPU resource allocation abort count) stored in the processing information 213-i. Step S3). Then, again, the execution control means 222-i starts an interval based on the dispatching information 213-i in which the adjusted interval time and CPU usage time for each workload class are stored, and starts the process. Execution control is performed (step S2). This loop of steps S2 and S3 is continuously executed until the system operation is stopped.

As described above, according to the CPU resource management device 200 according to the present embodiment, the overhead of process switching is reduced and the throughput of the entire system is improved for the same reason as the CPU resource management device 100 according to the first embodiment. .

Further, according to the CPU resource management device 200 according to the present embodiment, since the plurality of CPUs 22-1 to 220-m execute the plurality of execution control means 222-1 to 222-m in parallel with each other, The number of process executions increases and the overall system throughput improves.

In the above description, the CPU 220-0 is not equipped with an execution control means, but the CPU 220-0 may be equipped with an execution control means similar to the execution control means on the CPUs 220-1 to 220-m. Furthermore, in the above explanation, although the CPUs 220-1 to 220-m are not equipped with adjustment means, each CPU 220-1 to 220-m is equipped with an adjustment means similar to the adjustment means on the CPU 220-0, and the CPU 220-i ( i is any one from 1 to m), when the processing of the execution control means on the CPU 220-i is completed, the adjustment means on the CPU 220-i performs adjustment processing, and then the adjustment means on the CPU 220-i The execution control means may be arranged to start the next interval.

[Third embodiment]
FIG. 11 is a block diagram of a CPU resource management device 300 according to the third embodiment of the present invention. Referring to FIG. 11, the CPU resource management device 300 includes a storage device 310 and a processor 320 connected to the storage device 310.

Storage device 310 is configured to store workload definitions 311, executable process queues 312, and dispatching information 313. The workload definition 311 stores information representing the distribution ratio of CPU usage time for each workload class. The executable process queue 312 stores a plurality of processes belonging to any workload class. The dispatching information 313 includes the maximum CPU usage time for each workload class, the interval time that is the sum of the maximum CPU usage time for each workload class, and the fact that the maximum CPU usage time was reached between the start and end of the interval. The number of times CPU resource allocation has been terminated, the total allocation time that is the time from the start to the end of the interval, and the base interval time are stored.

Processor 320 includes one or more CPUs. The processor 320 also includes an execution control means 321 and an adjustment means 322.

When starting an interval according to the dispatching information 313, the execution control means 321 allocates CPU resources to a process belonging to a workload class whose CPU usage time after the start of the interval is less than or equal to the maximum CPU usage time, executes the process, and starts the interval. The interval is configured to end when there is no longer a process belonging to a workload class whose subsequent CPU usage time is less than or equal to the maximum CPU usage time. The execution control means 321 also determines the number of times CPU resource allocation is aborted due to reaching the maximum CPU usage time from the start to the end of the interval, and the total allocation time that is the time from the start to the end of the interval. is configured to be stored in the dispatching information 313.

When the total allocation time stored in the dispatching information 313 is shorter than the base interval time and the number of times of aborting CPU resource allocation stored in the dispatching information 313 is 1 or more, the adjustment means 322 adjusts the dispatching information 313 to the dispatching information 313. The stored interval time is configured to be changed to a longer time. The adjusting means 322 also distributes the interval time after changing the maximum CPU usage time for each workload class stored in the dispatching information 313 to a longer time according to the CPU usage time distribution ratio for each workload class. The CPU usage time is changed to the maximum CPU usage time calculated for each workload. Further, the adjustment means is configured to cause the execution control means 321 to start the next interval according to the changed dispatching information 313.

The CPU resource management device 300 configured in this manner operates as follows. That is, when the execution control unit 321 starts an interval according to the dispatching information 313, it allocates CPU resources to a process belonging to a workload class whose CPU usage time after the start of the interval is less than or equal to the maximum CPU usage time, and executes the process. The interval ends when there are no more processes belonging to the workload class whose CPU usage time after the start of is less than or equal to the maximum CPU usage time. The execution control means 321 also determines the number of times CPU resource allocation is aborted due to reaching the maximum CPU usage time from the start to the end of the interval, and the total allocation time that is the time from the start to the end of the interval. is stored in the dispatching information 313.

Next, when the total allocation time stored in the dispatching information 313 is shorter than the base interval time and the number of times of aborting the CPU resource allocation stored in the dispatching information 313 is 1 or more, the adjustment means 322 adjusts the dispatching information The interval time stored in 313 is changed to a longer time. The adjusting means 322 also distributes the interval time after changing the maximum CPU usage time for each workload class stored in the dispatching information 313 to a longer time according to the CPU usage time distribution ratio for each workload class. Change it to the maximum CPU usage time calculated for each workload. The adjustment means also causes the execution control means 321 to start the next interval according to the changed dispatching information 313.

According to the CPU resource management device 300 configured and operated as described above, the throughput of the computer system can be improved. The reason for this is that the previous interval ended in a shorter time than the base interval due to the existence of a workload class that did not require CPU resources for the maximum CPU usage time, and it also required CPU resources for more than the maximum CPU usage time. If a workload class exists, the next interval will be changed to a longer time than the previous interval, and the maximum CPU usage time for each workload class will also be changed accordingly. If the load of be.

Although the present invention has been described above with reference to the embodiments described above, the present invention is not limited to the embodiments described above. The configuration and details of the present invention may be modified in various ways within the scope of the present invention by those skilled in the art.

The present invention can be applied to applications such as CPU resource management, which allocates CPU resources used by each workload according to a predetermined CPU resource utilization ratio.

100...CPU resource management device 110...Storage device 111...Workload definition 112...Executable process queue 113...Dispatching information 114...Program 120...CPU
121...Initialization means 122...Execution control means 123...Adjustment means 131...Base interval time 132...Average process switching time 133...Threshold value 134...Next interval time 135...Next maximum CPU usage time for each workload class 136...Previous time Interval time 137...Previous total allocation time 138...Previous CPU resource allocation abort count 200...CPU resource management device 210...Storage device 211...Workload definition 212...Executable process queues 213-1 to 213-m...Dispatching Information 214-0~214-1...Program 220-0~220-m...CPU
221...Initialization means 222-1 to 222-m...Execution control means 223...Adjustment means 300...CPU resource management device 310...Storage device 311...Workload definition 312...Executable process queue 313...Dispatching information 320...Processor 321 ...Execution control means 322...Adjustment means

Claims (9)

A CPU resource management device comprising a storage device that stores a process, and a processor that executes the process,
The processor includes:
When an interval starts, CPU resources are allocated and executed to processes belonging to a workload class that satisfies the condition that the CPU usage time is less than or equal to the maximum CPU usage time, and when there are no more processes that belong to a workload class that satisfies the condition. Execution control means for recording the number of times CPU resource allocation is aborted due to reaching the maximum CPU usage time and the total allocation time that is the time from the start to the end of the interval. and,
If the total allocation time is shorter than the base interval time and the number of times the CPU resource allocation is aborted is 1 or more, the next interval time is determined to be longer than the previous interval time, and adjustment means for determining the next maximum CPU usage time for each workload class based on the distribution ratio of CPU usage time for each workload class;
A CPU resource management device comprising: When the total allocation time is longer than the base interval time, the adjustment means determines the next interval time to be shorter than the previous interval time, and adjusts the next interval time and the CPU usage time for each workload class. determining the next maximum CPU usage time for each workload class based on the distribution ratio;
The CPU resource management device according to claim 1. The adjustment means calculates the next interval time by multiplying the previous interval time by a ratio of the base interval time to the total allocated time.
The CPU resource management device according to claim 1 or 2. The adjustment means calculates a value obtained by multiplying the previous interval time by the ratio of the base interval time to the total allocation time, and calculates the total time taken for process switching due to the termination of the CPU resource allocation with respect to the base interval time. If the ratio exceeds the threshold, the value obtained by multiplying the value by the ratio is further divided by the threshold, and the value is calculated as the next interval time.
The CPU resource management device according to claim 1 or 2. A CPU resource management method executed by a computer including a storage device that stores a process and a processor that executes the process,
When an interval starts, CPU resources are allocated and executed to processes belonging to a workload class that satisfies the condition that the CPU usage time is less than or equal to the maximum CPU usage time, and when there are no more processes that belong to a workload class that satisfies the condition. and record the number of times CPU resource allocation was aborted due to reaching the maximum CPU usage time and the total allocation time that is the time from the start to the end of the interval;
If the total allocation time is shorter than the base interval time and the number of times the CPU resource allocation is aborted is 1 or more, the next interval time is determined to be longer than the previous interval time, and determining the next maximum CPU usage time for each workload class based on the CPU usage time distribution ratio for each workload class;
CPU resource management method. If the total allocation time is longer than the base interval time, the next interval time is determined to be shorter than the previous interval time, and based on the next interval time and the distribution ratio of CPU usage time for each workload class. to determine the next maximum CPU usage time for each workload class.
The CPU resource management method according to claim 5. calculating the next interval time by multiplying the previous interval time by the ratio of the base interval time to the total allocated time;
The CPU resource management method according to claim 5 or 6. A value obtained by multiplying the previous interval time by the ratio of the base interval time to the total allocation time is calculated, and the ratio of the total time taken for process switching due to the termination of the CPU resource allocation to the base interval time exceeds a threshold value. If so, calculate the value obtained by multiplying the value by the ratio and further dividing the value by the threshold value as the next interval time.
The CPU resource management method according to claim 5 or 6. A computer including a storage device that stores a process, and a processor that executes the process,
When an interval starts, CPU resources are allocated and executed to processes belonging to a workload class that satisfies the condition that the CPU usage time is less than or equal to the maximum CPU usage time, and when there are no more processes that belong to a workload class that satisfies the condition. terminating the interval and recording the number of times CPU resource allocation was aborted due to reaching the maximum CPU usage time and the total allocation time that is the time from the start to the end of the interval;
If the total allocation time is shorter than the base interval time and the number of times the CPU resource allocation is aborted is 1 or more, the next interval time is determined to be longer than the previous interval time, and determining the next maximum CPU usage time for each workload class based on the CPU usage time distribution ratio for each workload class;
A program to do this.

Images