KR101651114B1 - Apparatus for balancing load - Google Patents

Abstract

A load balancing apparatus capable of preventing unnecessary load balancing from being executed. According to one embodiment of the present invention, the load balancing apparatus detects a target task, determines whether to perform load balancing according to the state of the detected target task, and performs load balancing according to the determination.

Description

{APPARATUS FOR BALANCING LOAD}

To a load balancing technique that moves some of the tasks assigned to a plurality of processors to another processor.

Recently, as the amount of data increases, a single processor may not be able to process large amounts of data. To solve this problem, a technique of processing a large amount of data by mounting a plurality of processors on one chip has been developed.

However, as the number of processors increases, the performance of the system does not improve proportionately. This is because a load imbalance occurs in which all tasks are not evenly distributed to all of the processors but are executed and allocated to a specific processor in comparison with other processors.

Therefore, in order to solve such a load imbalance, researches on load balancing that distributes allocated tasks equally are actively under way.

A load balancing apparatus capable of preventing unnecessary load balancing from being executed.

A load balancing apparatus according to an embodiment of the present invention includes a determination unit that determines whether to perform the load balancing according to a status of a task to be load balanced among tasks allocated to a plurality of processors, And an execution unit for executing the program.

Here, the determination unit may determine to perform load balancing if the state of the target task is the first task state that increases the utilization rate of the processor below the threshold value.

Here, the determination unit can determine not to perform load balancing when the state of the target task is the second task state that increases the usage rate of the processor to a threshold value or more.

The load balancing apparatus according to another embodiment of the present invention includes a determining unit for determining load balancing using at least one of the number of tasks allocated to each processor and the usage rate for each processor, and an execution unit for executing the load balancing according to the determination .

Here, the determination unit may decide not to perform load balancing when at least one of the number of tasks allocated per processor is larger than the reference task number.

Here, the number of reference tasks can be calculated using the actual processing time of the task, the task limit processing time, and the number of processors.

Here, the determination unit can decide not to perform load balancing when the utilization rate per processor is smaller than the reference utilization rate.

Here, the determination unit may change the usage rate for each processor based on the time at which the usage rate is maintained.

Here, if the usage rate of the processor is smaller than the reference usage rate, the determination unit may change the usage rate of the corresponding processor based on the time when the usage rate is maintained.

Here, the determination unit may determine to perform load balancing when the number of tasks allocated per processor is smaller than the number of reference tasks and at least one of the utilization ratios for each processor is greater than a reference utilization ratio.

According to the disclosed contents, unnecessary load balancing is not performed by executing load balancing according to the state of the target task.

1 is a block diagram of a load balancing apparatus according to an embodiment of the present invention.
2 is a diagram for explaining a target task related to an embodiment of the present invention.
3 is a diagram for explaining a state of a task according to an embodiment of the present invention.
4 is a view for explaining a load balancing apparatus according to another embodiment of the present invention.
5 is a view for explaining a reference utilization rate according to an embodiment of the present invention.
6 is a flowchart illustrating a load balancing method according to an embodiment of the present invention.
7 is a flowchart illustrating a load balancing method according to another embodiment of the present invention.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of a load balancing apparatus according to an embodiment of the present invention.

The load balancing apparatus 100 includes a determining unit 101 and an executing unit 102. [ Or at least one processor 110, 120 to which tasks for performing load balancing are assigned.

The determination unit 101 detects a status of a task to be load balanced among the tasks assigned to the plurality of processors 110 and 120. Here, the target task means a candidate task to be moved to a queue of another processor by load balancing. Then, the determination unit 101 determines whether to perform load balancing according to the state of the target task.

2 is a diagram for explaining a target task related to an embodiment of the present invention. 2, task 1 and task 2 are allocated to a queue of the first processor 110, and tasks 3, 4, 5, and 6 are assigned to queues of the second processor 120 . Currently, there is a need to perform load balancing because the task assigned to the queue of the second processor 120 is larger than the task assigned to the queue of the first processor 110. [ At this time, the determining unit 101 may select a task ('target task') to be moved to the queue of the first processor 110 among the tasks assigned to the second processor 120. For example, the determination unit 101 may select the task 6 200 of the tasks assigned to the queue of the second processor 120 as the target task.

The determination unit 101 can determine the state of the target task based on the state of the task shown in Fig.

3 is a diagram for explaining a state of a task according to an embodiment of the present invention. The state of a task according to an exemplary embodiment of the present invention may include a new task state, an idle task state, a new idle task state, a running task state, Can be distinguished. The new task state means that the task is newly created. The idle task state indicates a state in which the task is waiting to be allocated to the processor, and the idle task is assigned to the processor in the order of priority among the idle tasks. The new idle task state means a state in which an input / output interrupt is generated and then an input / output completion signal is input. For example, when an input / output completion signal is input, the new idle task transitions to the idle task. A running task state means that a task is assigned to a processor and a task is being processed by the processor. Other task states may include tasks whose execution is terminated. Although the state of the task will be described with reference to Fig. 3 in the present embodiment, the state of the task can be divided into more tasks or divided into fewer tasks according to the setting of the user or the like.

The determination unit 101 determines whether to perform load balancing according to the state of the target task. For example, when the state of the target task is the first task state, the determination unit 101 determines to perform load balancing. Here, the first task state means a state in which the usage rate of the processor is increased from 0% to less than the threshold value. When a task is assigned to and executed by a processor, the utilization of the processor is increased. For example, the state of a task that increases the processor utilization from 0% to 20% is the first task state. Here, the threshold value is 20%. Referring to FIG. 3, the first task state may be another task. That is, in the case of other tasks, the usage rate of the processor increases to less than 20%. Here, the range in which the usage rate of the processor is changed according to the state of the task can be determined through experiments.

On the other hand, when the state of the target task is the second task state, the determination unit 101 determines not to perform load balancing. Here, the second task state means a state in which the use rate of the processor is increased to a threshold value or more. For example, the state of the task increasing the usage rate of the processor to 20% or more becomes the second task state. Here, the threshold value is 20%. 3, the first task state may be a new task state, an idle task state, a new idle task state, and a running task state . That is, in the case of a new task state, an idle task state, a new idle task state, and a running task state, the usage rate of the processor increases to 20% or more.

The execution unit 102 performs load balancing according to the determination of the determination unit 101. [ For example, when the determination unit 101 determines not to perform load balancing, the execution unit 102 does not move the target task to another processor. On the other hand, when the determination unit 101 determines to perform load balancing, the execution unit 102 moves the target task to another processor.

Load balancing devices can perform load balancing according to the state of the target task, thereby reducing unnecessary load balancing and improving system performance.

4 is a view for explaining a load balancing apparatus according to another embodiment of the present invention.

The load balancing apparatus 400 includes a determination unit 401 and an execution unit 402. Or at least one processor 410, 420 to which tasks for performing load balancing are allocated.

The determining unit 401 determines whether or not the load balancing is performed using at least one of the number of tasks allocated to the processors 410 and 420 and the usage rates of the processors.

Hereinafter, a case where the determination unit 401 determines load balancing using the number of tasks allocated for each processor will be described.

The determining unit 401 determines not to perform load balancing when at least one of the number of tasks allocated per processor is greater than the reference task number. In other words, there is at least one processor to which the number of tasks is larger than the reference task number, and the decision unit 401 decides not to perform load balancing.

Here, the number of reference tasks is calculated by the following equation (1).

Where n: the number of reference tasks

        C: Actual processing time of task

        T: Task limit processing time

        m: Number of processors

The actual processing time of the task means the time it takes to process the task in the processor. Task limit processing time refers to the time allocated to process tasks in the processor. That is, when the task processing time exceeds the task limit processing time, the processor suspends the processing of the task and prioritizes another task.

For example, when the task actual processing time C is 300 占 퐏, the task limit processing time is 10000 占 퐏, and the number m of processors is 4, the number of reference tasks is 76 according to the equation (1). Alternatively, the number of reference tasks may be set to 75, 74, or the like, depending on the setting of the user. That is, the number of tasks obtained by subtracting one or two from the number of tasks obtained in Equation (1) may be set as the number of reference tasks.

The determination unit 401 determines not to perform load balancing when at least one of the number of tasks allocated per processor is larger than the calculated number of reference tasks. Hereinafter, the case where the number of reference tasks is 75 will be described as a reference. For example, when the number of tasks allocated to the first processor 410 is 80 and the number of tasks allocated to the second processor 420 is 70, the determining unit 401 determines whether the number of tasks allocated to the first processor 410 Since the number of tasks is larger than the number of reference tasks, it is decided not to perform load balancing.

In another example, when the number of tasks allocated to the first processor 410 is 72 and the number of tasks allocated to the second processor 420 is 70, the determination unit 410 determines the number of tasks allocated to the first processor 410 and the second processor 420, Since the task assigned to the second processor 420 is smaller than the reference task number, it is determined to perform load balancing.

Hereinafter, a case where the determination unit 401 determines whether or not load balancing is to be performed using a processor-specific usage rate will be described.

For example, the determination unit 401 determines to perform load balancing if at least one of the utilization ratios per processor is larger than the reference utilization ratios. Here, the reference utilization rate is a utilization rate that is a reference for whether load balancing is to be executed or not, and can be set by a user or the like. If the usage rate of the first processor 410 is 80%, the usage rate of the first processor 410 is 85%, and the usage rate of the second processor 420 is 70% Since it is larger than the usage rate, it is decided to perform load balancing.

5 is a view for explaining a reference utilization rate according to an embodiment of the present invention.

Referring to FIG. 5, the usage rate of the processor can be divided into a first area, a second area, and a third area.

If the utilization rate of the processor corresponds to the first area, the determination unit 401 determines that load balancing is to be executed. The first area and the second area are classified based on the reference use rate (500).

When the utilization rate of the processor corresponds to the second area, the determination unit 401 determines that load balancing is not performed. However, after dividing the second area into 2-1 area, 2-2 area, 2-3 area and 2-4 area, the determining part 401 can change the usage rate of the processor based on the time when the usage rate is maintained have. For example, when the occupancy rate of the processor is maintained in the 2-4 region for 3 seconds or longer, the determination unit 401 can change the occupancy rate of the processor from the 2-4 region to the third region. Although the second embodiment has been described with reference to the second area, a method of changing the usage rate according to the weight may be applied to other areas as well. By performing load balancing in advance before the processor utilization reaches the third area, the task can be evenly distributed to the processors.

When the utilization rate of the processor corresponds to the third area, the determination unit 401 determines that load balancing is not performed. That is, the third area is an area in which load balancing need not be performed.

Hereinafter, a case where the determination unit 401 determines load balancing using the number of tasks allocated to each processor and the usage rate for each processor will be described.

For example, the determination unit 401 determines to perform load balancing when the number of tasks allocated per processor is smaller than the number of reference tasks and at least one of the utilization ratios for each processor is larger than the reference utilization ratio. Hereinafter, it is assumed that the reference usage rate is 80%, the first processor 410 is used 85%, the second processor 420 is used 70%, the reference task number is 75, and the first processor 410 The case where the number of allocated tasks is 73 and the number of tasks allocated to the second processor 420 is 70 will be described. The determination unit 401 determines that load balancing is to be executed because the number of tasks of the first processor 410 and the second processor 420 is smaller than the number of reference tags and the utilization rate of the first processor 410 is larger than the reference utilization rate do. Otherwise, the determination unit 401 determines not to perform load balancing.

Load balancing devices use processor utilization and the number of processor tasks to determine whether to perform load balancing, thereby reducing unnecessary load balancing and improving system performance.

In yet another embodiment, the load balancing device described in Figures 1 and 4 may be implemented as a single device. For example, the load balancing apparatus 100 shown in Fig. 1 is used to preliminarily determine whether or not to perform load balancing, and finally, the load balancing is performed using the load balancing apparatus 100 shown in Fig. 4 Or not. However, only when the load balancing apparatus 100 determines to perform load balancing, the load balancing apparatus 400 finally makes a determination. That is, when the load balancing apparatus 100 determines not to perform load balancing, the load balancing apparatus 400 is not executed.

For example, if the load balancing apparatus 100 described in FIG. 1 preliminarily determines that load balancing is to be executed, and then determines not to perform load balancing in the load balancing apparatus 400 shown in FIG. 4, Do not perform balancing. In this case, the determination as to whether or not the load balancing is determined by the load balancing apparatus 400 is given priority.

6 is a flowchart illustrating a load balancing method according to an embodiment of the present invention.

Referring to FIG. 6, the determination unit determines whether a load imbalance has occurred at a predetermined time (600). When load unbalance occurs, the determination unit selects a task to be migrated to another processor among the tasks assigned to the plurality of processors (610). The determination unit detects the state of the target task (620). The determining unit determines whether to perform load balancing according to the detected status of the target task (630). If the state of the target task is the first task state, the determination unit 640 determines to perform load balancing. On the other hand, when the state of the target task is the second task state, the determination unit determines not to perform the load balancing (650). The execution unit executes load balancing according to the determination result of the decision unit. The first task state means a state in which the usage rate of the processor is increased below a threshold value. The second task state means a state in which the usage rate of the processor is increased to a threshold value or more.

7 is a flowchart illustrating a load balancing method according to another embodiment of the present invention.

The determination unit detects the number of tasks allocated per processor (700). In addition, the determination unit detects the usage rate per processor (710). The determination unit determines whether to perform load balancing using at least one of the number of tasks allocated per processor and the usage rate per processor (720). For example, the determining unit 401 determines (730) if load balancing is to be performed when at least one of the utilization ratios per processor is greater than the reference utilization ratios ('first condition'). In another example, when the number of tasks allocated to each processor is smaller than the number of reference tasks and at least one of the usage ratios for each processor is larger than the reference usage rate ('first condition'), the determination unit 401 determines to perform load balancing (730). In another example, the determining unit 401 determines 740 that the load balancing should not be executed when at least one of the number of tasks allocated to each processor is larger than the reference task number (the 'second condition').

The embodiments described may be constructed by selectively combining all or a part of each embodiment so that various modifications can be made.

It should also be noted that the embodiments are for explanation purposes only, and not for the purpose of limitation. In addition, it will be understood by those of ordinary skill in the art that various embodiments are possible within the scope of the technical idea of the present invention.

Further, according to an embodiment of the present invention, the above-described method can be implemented as a code that can be read by a processor on a medium on which the program is recorded. Examples of the medium that can be read by the processor include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, etc., and may be implemented in the form of a carrier wave (e.g., transmission over the Internet) .

Claims (10)

delete delete delete Balancing is performed in accordance with the status of the load balancing task among the tasks allocated to the plurality of processors, and the load balancing is determined by using at least one of the number of tasks allocated to each processor and the usage rate per processor A decision unit for finally determining And
And an execution unit for executing the load balancing according to the final determination,
Wherein,
Wherein the processor is configured to perform load balancing if the state of the target task is a first task state that increases a usage rate of the processor to less than a threshold value and if the state of the target task increases a usage rate of the processor to a threshold value or more, Preliminarily decides not to perform load balancing in the state,
The first task state includes a new task state in which the task is newly created, an idle task state waiting to be allocated to the processor, a new idle task waiting to input the input / output completion signal after the input / output interrupt occurs, Wherein the second task state includes at least one of an execution task state in which the task is being processed, and the second task state includes other task state in which the execution of the task is terminated.
5. The method of claim 4,
Wherein,
And finally determines not to perform load balancing when at least one of the number of tasks allocated to each processor is greater than a reference task number.
6. The method of claim 5,
The number of reference tasks,
Task Load balancing device calculated using actual process time, task limit process time, and number of processors.
5. The method of claim 4,
Wherein,
And finally determines not to perform load balancing when the per-processor utilization rate is smaller than the reference utilization rate.
5. The method of claim 4,
Wherein,
And changes the usage rate per processor based on a time when the utilization rate is maintained.
9. The method of claim 8,
Wherein,
A load balancing device that alters the utilization of a corresponding processor based on the time the utilization is maintained when the utilization of the processor is less than a reference utilization.
5. The method of claim 4,
Wherein,
The number of tasks allocated to each processor is smaller than the number of reference tasks,
And if the at least one of the processor-specific utilization ratios is greater than a reference utilization ratio, the load balancing apparatus finally determines to perform load balancing.

Images