Research On Energy Efficient Scheduling For Multi-core Embedded Real-Time System Base On Dynamic Voltage Scaling

Power management is of primary importance in embedded system, which can be attributed to longer battery life, reliability and packaging costs. And real-time tasks scheduling is another important problem, because embedded system often carry on real-time jobs, such as sensor networks, unmaned vehicles,satellites, which can be called embedded real-time systems as well. It is necessary to research scheduling algorithms which can get lower energy consumption and meanwhile guarantee the deadlines of real-time tasks.Meanwhile, as the multi-core processors being widly used in embedded real-time systems, it soon became a hot research direction. This paper tackles the problem of reducing energy consumption for multi-core embedded real-time system, using well-known techniques, such as Dynamic voltage and frequency scaling (DVS).We propose two novel scheduling algorihtm use Earliest Deadline First (EDF) to guarantee meeting the deadlines of all real time task sets for each core and to make DVS more efficiency:DVS-DA and DVS-DS.Consider that the static partitioned scheduling algrithms, such as DVS-CC, which apply Cycle conserving EDF base on DVS scheduling algorithm in multicore embedded real-time system, only allow the every execution of a particular task to take place in the assigned core, even there is a idle core. It may cause unnecessary energy consumption. So DVS-DA algorithm make a dynamic partition, when there is arrivaling task, it will be assign to the core who have the minimum workload. This mehthod will lead to a more balanced workload between all cores, so the DVS technique will have a better effiency.Moreover, the DVS-DS algorithms optimize the DVS-DA by turn the unnecessary core to sleep state, which save the Leakage energy consumption.The two algorithms is proved to be energy efficiency by the Evaluatio.