KR101384311B1 - Apparatus and method for reducing power consumption in wireless communication system - Google Patents

Abstract

The present invention relates to an apparatus and method for reducing power consumption in a mobile communication system, and more particularly, to an apparatus and method for maintaining performance while minimizing operation of hardware required for data processing.

Mobile communication system, data processing, work time interval (TTI), system call

Description

APPARATUS AND METHOD FOR REDUCING POWER CONSUMPTION IN WIRELESS COMMUNICATION SYSTEM}

The present invention relates to an apparatus and method for reducing power consumption in a mobile communication system, and more particularly, to an apparatus and method for maintaining performance while minimizing operation of hardware required for data processing.

A mobile communication system is a very complex system operated by many processing processes, which perform operations at various levels of supply voltage and clock frequency.

The primary purpose of processor design is to reduce power consumption while providing improved performance. Modern processors may have several different levels of performance, depending on the requirements of the application being run for them. Such processors can significantly reduce power consumption by reducing the clock frequency and hence the operating voltage. However, this reduction in processor performance can be accepted without affecting or hardly affecting performance as determined by the user. However, in order to reduce the power consumption, it may cause a reduction in the operating frequency or a decrease in the operating voltage, or a decrease in the performance of the system due to a malfunction. Moreover, due to the decrease in the performance level of such a processor, You should not hand it over. Also, it is more effective to save power by performing the task slowly to the execution time limit, rather than ending the task execution time and entering the idle mode.

Conventional power saving methods average the load of the work being performed for a certain period of time to predict the performance level of the next job to be performed, or improve it to use the process of the work performed between task time intervals defined for each job. There is a method to measure the load by reflecting the rate and to set the performance level of the work performed based on the rate.

However, even if the protocol stack software of mobile communication system is the same task, the past workload is not related to the workload to be performed in the future, and thus the method that uses the policy established for the previously performed task is applied as it is. If you do, the system will not meet the performance requirements.

For example, the same task, but the load was small due to the small amount of data in the past, but when large data is instantaneously received, the performance level is set lower because it is not predicted, and in this case, the execution time is exceeded because the requirement is not satisfied. Will occur. In the opposite case, the power saving effect is lost by setting a higher performance level than the actual one.

Accordingly, the present invention, which was devised to solve the problems of the prior art operating as described above, proposes a method and apparatus for increasing power saving efficiency and system stability by setting a performance level of a task considering the characteristics of a mobile communication system. .

Specifically, the above problems caused by incorrectly predicting the performance level of the existing work are improved by extracting the factors affecting the work load and predicting the optimally planned work load by using the process utilization rate of the work accordingly. The present invention proposes an apparatus and method for improving power saving performance.

In addition, we apply the method of setting the execution time limit considering the tasks to be performed in the transmission time interval (hereinafter referred to as TTI), which is the transmission unit of the protocol stack. An apparatus and method for improving the occurrence of system errors are presented.

An apparatus and method for reducing power consumption in a mobile communication system are provided.

According to the present invention operating as described above in detail, by extracting the factors affecting the workload of the task in the mobile communication system and setting the performance level in consideration of the transmission time interval of the task processing time limit, power saving efficiency and This increases the stability of the system.

Hereinafter, the operation principle of the present invention will be described in detail with reference to the accompanying drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unnecessary. The following terms are defined in consideration of the functions of the present invention, and may be changed according to the intentions or customs of the user, the operator, and the like. Therefore, the definition should be based on the contents throughout this specification.

1 is a configuration diagram of a power management apparatus of a data processing system according to the present invention.

Referring to FIG. 1, a kernel module (hereinafter, referred to as a 'kernel module') 110 that performs basic functions, a system call (hereinafter, referred to as a 'system call'), a scheduler (hereinafter, referred to as a 'scheduler'), and The OS kernel 145 performs functions such as power management such as sleep / wake.

According to the present invention, the proposed APM (hereinafter referred to as 'APM') 115 is configured inside the kernel and is referred to as a main controller (hereinafter referred to as 'main controller') 140 and a workload meter (hereinafter referred to as 'workload'). estimator ') (130), policy stack manager (hereinafter referred to as' policy stack manager') 120, policy stack (hereinafter referred to as' policy stack ') 125, performance level controller (hereinafter referred to as' performance level controller') It consists of five blocks of 135). The APM 115 is called at the time when a task is changed in the OS kernel 145, the previous workload and the task completion time (hereinafter referred to as 'deadline') information, the task workload factor (hereinafter referred to as 'deadline') The performance level of the next task to be executed is determined by using 'TWF') information and policy (hereinafter referred to as 'policy') information, and this is called Dynamic Voltage Frequency Scaling (DVFS) device. Transfer to the controller 150.

The main controller 140 is called by a hooking function (hereinafter referred to as a 'hooking function') that is performed when a task is started in the OS kernel 145. The main controller 140 is a block that manages the entire flow of the IEM SW and calls the workload estimator 130 and the performance level controller 135 as necessary. In addition, the status information and the task identifier of the performed task (hereinafter referred to as "task") is received from the OS kernel 145, and task information is collected and calculated for each task operation time.

Workload estimator 130 is a block that calculates the workload for the period after the task is executed. The current workload value and previous value are used to calculate the average value of the next workload to be executed, calculate the average value of the children, and store the values in the TWF location of each task in the policy stack 125.

The performance level controller 135 uses the TWF received from the protocol stack manager 120 to obtain an average workload value and deadline information of the task from the policy stack 125. The final performance level is calculated and delivered to the DVFS controller 150 using the workload and deadline for the task to be performed.

The policy stack manager 120 manages input and output of task related information stored in the policy stack 125. The policy stack 125 stores the average and deadline values of the previously performed workload according to the TWF of each task.

Hereinafter, an example of the apparatus and method according to the present invention will be described with respect to a case of receiving data in a mobile communication system.

The mobile communication system has a hierarchical (Layer) structure for data processing, and delivers data to the next layer according to the processing in each layer.

Assuming each layer is a task,

First 1. Measure and save the workload of each task by TWF. At this time, TWF is a factor that has a major influence on the workload of the task, and the number of MAC Service Data Units (SDUs) to be processed in the Media Access Control (MAC) layer. And the number and size of radio link control (RLC) SDUs and the number of packet data convergence protocol (PDCP) SDUs. .

2. Extract TWF and number of tasks to be executed through system call or task occurrence event.

3. Read the stored workload corresponding to the extracted TWFs.

4. The idle time for determining the performance level of the task to be performed is calculated as shown in Equation 1 below.

I = TTI-t _elapse -tp _task -tap _task -t _ISR

Where I is the idle time, t _elapse is the time spent in the TTI before the task to be executed (0 to 1 ms), and tp _task is the tasks that are expected to run and the workload corresponding to that TWF. Shows the sum of the expected times when running at CPU full speed.

In addition, the tap _task is the expected time when the task is executed at CPU full speed when a task that is executed aperiodically is triggered. Set to '0' if non-periodic task is not triggered. t _ISR is the sum of the ISR execution times occurring periodically in the TTI.

5. The performance level of the task to be performed according to the idle time calculated according to Equation 1 can be calculated using Equation 2 below.

P = W / D = W / W + I

Where P is the performance level, W is the workload (full speed equivalent work) of the task to be performed, D is the deadline of the task to be performed, and I is the idle time calculated according to Equation 1 above.

6. In addition, when the task is preemption, the performance level when the task is re-executed can be calculated using the following Equations 3 to 4.

W = W '-W' '

Here, W 'denotes a workload of a task to be re-executed, and W' 'denotes a workload already performed.

I '= TTI-t' _elapse -tp ' _task -tap' _task -t _ISRa '

Where I 'is the idle time upon _rerun , t' _elapse : the time spent in the TTI, tp ' _task is the estimated time of the tasks to be performed, tap' _task is the predicted execution time of the aperiodic task, and t _ISRa 'is the ISR of the ISR. Represents the time to be reset in consideration of the execution time.

6. If the task exceeds the TTI boundary point, set the panic workload to reflect the following Equation 5 when setting the performance level in the next TTI.

는 다음 TTI에서의 수정 workload, W는 수행될 task의 workload (full speed equivalent work), W_p는 panic workload, deadline시까지 수행완료 되지 못한 task의 workload, 안정성을 위하여 수행되지 못하고 남은 workload 보다 많은 값이나, 고정값을 적용할 수 있다.here,

Is the modified workload at the next TTI, W is the workload (full speed equivalent work) of the task to be performed, W_p is the panic workload, the workload of the task that has not been completed until deadline, or more than the remaining workload that has not been performed for stability. A fixed value can be applied.

I = TTI-telapse-tptask-taptask-tISR (Equation 1)

7. In another embodiment, when the idle time calculation equation is used, the following equation (6) may be calculated to improve the tendency of the performance level to be set higher in the second half of the TTI.

I = [TTI-t _elapse -tp _task -tap _task -t _ISR ] * '[a + (1-a) * (telapse / TTI)]

Here, a represents an arbitrary value between 0 and 1 (the optimal value is selected depending on the experimental results or the application to be applied).

8. Also, considering the stability of the system, hook the triggering of non-periodic tasks whose deadline is not constant at the OS level to set the performance level to the maximum, and when the execution of the non-periodic task is finished. How to keep up to you can apply.

FIG. 2 is a diagram illustrating an embodiment of a method of setting a performance level using a load and an idle time of a task in a data processing system according to the present invention.

Referring to FIG. 2, first, when system call_1 associated with a task occurs, a TWF according to the associated task is extracted during system call processing. In the example applied to FIG. 2, it is assumed that system call_1 is associated with both task A and B, and the associated TWF is referred to as TWF_A and TWF_B, respectively.

Workload of Task A sets the workload according to TWF_A and TWF_B stored in the policy stack using TWF_A and TWF_B extracted during system call_1 processing. The sum is the workload (W) to be used for setting the performance level. If there is no history of the workload for TWF_A before, set the performance level to maximum. After Task A is executed, save the policy stack stored in TWF_A and use it for future task A workload estimation according to TWF_A.

Idle time (I) of Tasks A and B is calculated by Equation 1 as described above. That is, t _elapse , the time elapsed before Task A is performed in the total time of TTI, _tptask , which is the execution time of tasks expected to be performed in the system call, and tap ', which is the execution time for the interrupt that is expected to be performed in TTI. Refers to the rest of time except _task . Using the calculated W and I, the performance level is set using Equation 2.

3 is an exemplary diagram illustrating a method of setting a performance level of a task when a job is preemulated.

Referring to FIG. 3, it is assumed that system call_2 is associated with task C, and the associated TWFs are referred to as TWF_C, respectively. It is also assumed that task C has a higher priority than task B, causing task preemption. In this case, Workload of Task C uses TWF_C extracted during system call_2 processing to set the workload according to TWF_C stored in the policy stack. Task B is executed, and the remaining workload and task C's workload sum are the workloads (W) to be used for setting the performance level. Likewise, if there is no history of the workload for TWF_C before, set the performance level to maximum.

The subsequent process, that is, the process of calculating the idle time and setting the performance level using the same is the same as the process in the description of FIG.

4 is another exemplary embodiment of a method for setting a performance level of a task when a job is preemulated.

Referring to FIG. 4, it is assumed that system call_2 is associated with task C, and the associated TWF is called TWF_C, respectively. In addition, assuming task C has a higher priority than task B, and task preemption occurs, if task B is re-executed after task C has been preemptiond and executed, the remaining workload will be executed. The workload (W) used to set the performance level. The process of setting the performance level using this is the same as the process in the description of FIG.

FIG. 5 is another exemplary diagram of resetting a performance level even if a task is not pre-empted in a case of a predetermined system call.

Referring to FIG. 5, it is assumed that system call_3 is associated with task D, and the associated TWFs are referred to as TWF_D, respectively. In this case, task D has a lower priority than task B so that task preemption does not occur. In this case, since the task preemption does not occur and the workload prediction for task C may not be reflected, the performance level can be reset even if there is no task scheduling for a specific system call.

That is, when system call_3 occurs, the priority of the task D and the task B being executed is checked, and if the task preemption does not occur, the performance level can be reset.

In this case, Equation 3 is applied to set the sum of the remaining workloads performed in the predicted task D and the predicted workload of Task B as the workload (W) used to set the performance level.

Likewise, subsequent processes are the same as those in the above description of FIG. 2.

FIG. 6 illustrates another example of setting a panic workload when a task execution time exceeds a Pseudo_TTI (Transmission Time Interval Boundary) to reflect when setting a performance level in the next TTI. .

Referring to FIG. 6, Task D starts at the previous TTI and exceeds the TTI boundary, and system call_1 occurs at the boundary of the new TTI.

It is assumed that system call_1 is associated with both task A and B. The associated TWF is assumed to be TWF_A and TWF_B, respectively. Task D has a higher priority than task A or task B. In this case, the sum of the remaining workloads performed among the predicted workloads of Tasks A, B, and Task D is used as the workload (W) used to set the performance level. In this case, the calculation of the workload is performed using Equation (5).

The subsequent process is the same as the process in the description of FIG.

Also, if the task exceeds the TTI boundary for the stability of the system, it is also applicable to set the performance level to the maximum until the task is completed. For example, when tasks A and B both have a higher priority than task D, they may be processed as described in FIG. 4.

1 is a configuration diagram of a power management apparatus of a data processing system according to the present invention;

2 is a view showing an embodiment of a method for setting a performance level using a load and an idle time of a task in a data processing system according to the present invention;

3 is a diagram for describing a method of setting a performance level of a task when a job is preemulated;

4 is another embodiment of a method of setting a performance level of a task when a job is preemulated;

FIG. 5 illustrates another embodiment of resetting a performance level even if a task is not pre-empted for a case of a predetermined system call;

FIG. 6 illustrates another example of setting a panic workload when a task execution time exceeds a Pseudo_TTI (Transmission Time Interval Boundary) to reflect when setting a performance level in the next TTI. .

Claims (2)

In a method for reducing power consumption in a mobile communication system, Measuring the workload of a task for each task workload factor (TWF), Determining the number of tasks to be performed and the TWF when a system call or task event occurs; Acquiring a workload corresponding to the determined TWF; Determining an idle time for determining a performance level of a task to be performed, Determining the performance level of the task to be performed. An apparatus for reducing power consumption in a mobile communication system, Measure the workload of a task by TWF, determine the number of tasks and TWF to be performed when a system call or task event occurs, obtain the workload corresponding to the determined TWF, and determine the idle time to determine the performance level of the task to be performed. Advance Power Management (APM), which determines the performance level of the task to be performed, And an operating system kernel for generating the system call or task event.

Images