Research On Energy-efficient Real-time Task Scheduling Algorithm Based On Multi-core Processors

With the continuous development of computer technology, the real-time systems arewidely used. The functional requirement to real-time systems of people is graduallyincreasing and the high-performance multi-core processors have become the mainstream ofreal-time system applications microprocessor. Increasingly serious global warming problemmakes the low-power design as an important means of green computing, witch has caused thehighly concerned of many experts and scholars. Besides, plenty of literature studies haveshown that reduced frequency and energy-saving resulted in the increasing transient error andlower system reliability and negative impacted on the high reliability real-time systems likemaritime or aerospace. This paper aims to study the task scheduling problem of multi-corereal-time system and obtain an ideal task scheduling algorithms. In the premise of meeting thespecific requirements of multi-core real-time systems, the algorithm in this paper gives fullplay to parallelism of multi-core processors and rational use of free time which devote to theminimum energy consumption of the system.According to whether the reliability requirement is needed or not, the real-time systemscan be divided into ordinary real-time systems and highly reliable real-time systems. Thispaper proposed a dynamic optimization energy-efficient scheduling algorithm whichincluding retiming optimization, static task scheduling identify and the use of dynamic idletime for the ordinary real-time systems. Compared with the existing scheduling algorithms.Retiming technology based on software pipeline eliminate dependency constraints betweentasks within the iteration and maximum the degree of parallelism of task set. HybridIntelligent optimization algorithm was used to get the better static energy-efficient real-timetask scheduling, and further reduce the operating frequency of the task execution by idle time.Although the DEERTS algorithm can reduce energy consumption, it can not maintain thesystem’s original reliability. To conquer the short coming, design a dynamic optimization ofreliability-aware energy-efficient scheduling algorithm DDRAPM for high reliable real-timesystems based on algorithm DEERTS. The determine stage of static task scheduling reservesshared recovery module for maintain the original reliability of system and uses of the twodynamic idle time generated by uncertainties of task execution time and transient error.At the last, a lot of simulation experiments were done used E3S test platform and TGFFtest platform. To compare with the existing algorithms, the experimental results show that:DEERTS algorithm and DDRAPM algorithms have better task parallelism and reduced the energy consumption of the system, and DDRAPM algorithm is able to guarantee thereliability of the system.