The Research On Energy Saving Scheduling Strategy Base On Reliability Constraint About Multi-core Platform

In recent years,multi-core technology has been widely used in many fields.Multi-core processors will not only enhance system performance,they are certain to bring higher power consumption.As is known to all,high power consumption will certainly cause the increase of system temperature and lead to decreasing of system reliability in a large part.Thus,how to minimize the system energy consumption is one of the main issues that multi-core system should concern about.Besides,the demand of reliability property must be put into the applications of multi-core platform.So proper task assigning approaches and task scheduling approaches must be designed to lower energy consumption while being sure to satisfy system reliability as well as applications' QoS.However,there is a mutual promotion and restriction relationship between the optimization research of these two aspects.Therefore,how to reduce the system energy consumption under the precondition of ensuring certain reliability is an urgent problem to be solved in the research of multi-core processor system task scheduling technology,which has important and practical significance for various applications in the computer field.Targeted to the frame tasks and hard real-time periodic tasks respectively,the paper works on studying the multi-core energy-saving scheduling strategy based on reliability constraints.The main contents are as follows:Aiming at reducing energy consumption and ensuring original reliability for frame tasks,this paper analyses the problems in the reliability-based dynamic power management strategy and putting forward a single-core task scheduling strategy TSDR which is based upon task sequencing and dynamic recovery.This strategy guarantees reliability by arranging a certain number of recovery blocks for tasks as backup tasks.Based on task sequencing,TSDR combines the idea of hybrid idle recycling and dynamically adjusts the frequency of tasks when recovering blocks.Experiments show that this strategy saves 14.5% energy consumption compared with the competitor when there is no fault occurs,and saves about 19.7% energy compared with the competitor when there is a fault occurs.Then,on the multi-core processor platform,TSDR strategy is extended according to the worst-fit scheduling method and a multi-core energy-saving scheduling strategy WFD-TSDR is proposed.Experiments show that it has better energy-saving effect when tasks can be allocated to more processor cores.Hereafter,when the task model is adjusted to hard real-time periodic tasks,in order to save energy,satisfy original reliability and tolerate a single permanent fault,the standby-sparing strategy for fixed priority is improved and a task scheduling strategy MTDP based on hybrid tasks and dynamic priority is proposed.Because hard real-time periodic tasks are different from frame tasks in number and deadline,the strategy assigns backup tasks to each primary task to ensure the goal of system reliability.Moreover,the strategy can also tolerate a single permanent fault for dual-core processors because of task mixing allocation.In the scheduling process,besides assigning the primary and secondary tasks mixedly to the processor,the strategy also combines the idea of dynamic priority to enhance the schedulability,and reduces the overlap of the primary and secondary tasks by dynamically calculating the reserved time to reduce energy consumption.Experiments show that in the best case,MTDP strategy can save up to 17% energy compared with the comparative strategy when no fault occurs.And in the worst case,the strategy can also save about 20% energy compared with the comparative strategy when all the primary tasks lose effectiveness.In order to further verify the effectiveness of MTDP strategy,MTDP strategy is applied to multi-core processor platform based on clustering idea and a multi-core energy-saving scheduling strategy CDU-MTDP is proposed.The experimental results show that the CDU-MTDP strategy is superior to the comparative strategy in energy saving,and it can save up to 25% of energy consumption.