Research On Energy-Aware Scheduling Strategy Based On Multi-core Platform

In recent years,with the rapid development of the computer industry,multi-core system has been applied in various fields? Although the application of multi-core processor improves the performance of the system,the system energy consumption is also increasing.In the multi-core computing system,the unbalanced allocation of tasks and unreasonable clock frequency will cause the waste of resources.The high energy consumption of the system not only does not conform to the concept of green computing,but also can cause the common radiation problem to affect the performance of the system.How to reduce the energy consumption based on the multi-core computing system has been the key research direction in the field of energy saving scheduling.The main research content of this thesis is as follows:1)This paper studies the energy-efficient scheduling of the DAG tasks with constraints,and analyses the problems of Task Duplication Scheduling based on Dependency Degree,and proposes an improved Task Duplication Scheduling based on Dependency Degree for the shortcomings of the algorithm.The improved algorithm uses DVFS to recycle the free time without affecting the makespans of the tasks,and the optimum frequency selection method is adopted to confirm the clock frequency to reduce the energy consumption.2)In this paper,the method of optimal frequency selection is applied to the scheduling of independent periodic tasks,and the independent periods task is divided into some segment according the deadline of the tasks.At each segment,the frequency is determined according to the optimal frequency selection method.By employing the optimal frequency selection method,the frequency is choose based on the degree of parallelism of the task comparison with the method chosing a single frequency the energy consumption is lower.3)In order to solve the problem that the schedulability of DULB partitioning strategy is poor when the system load is high and the load balancing is also not well when the system load is low,this paper proposes the DULB based on Average Compensation strategy.The strategy divides the task set into two parts,large task and small task.The large task is used to divide and the stage and the small task is used to compensation.In the process of partition if the load of the processor is more than the average of the partition set of tasks after accepting a new task,then remove the task and chose the task whose size is closest the difference between the load of the processor and the average of the partition set of tasks.If the load of the processor is more than the average of the partition set of tasks,the processor can not accept tasks any more.In the task compensation phase,if the system load is more than 85%,the improved strategy is aim for improving scheduliability,and the task is divided to the processor with the largest load which can accept the task.If the system load is less than 85%,the task will be divided the same as partition phase.