Research Of Energy-efficient Scheduling Algorithm Based On DVS For Hard Real-time Tasks

With the rapid development of semiconductor and chip technique latest years, more and more portable devices, such as MP3,PDA,mobile phone and laptop and so on, play a very important role in people's daily lives. However, the large extent increase of chip density and clock frequency leads to the power consumption of SLSI system bigger and bigger at the same time, the energy consumption problem is especially obvious in the portable devices which support by battery. The sharply increase of energy consumption results in not only higher costs of chip packaging and cooling, but also the increase of missing chip efficiency, and contributes to the reduction of stability and reliability of real-time system. As a result, the energy consumption problem has become a critical matter which restricts the development of portable devices, and modern system designers focus on the research of energy optimization and energy-efficient techniques.Most of real-time energy-efficient scheduling algorithms have good efficiency in saving energy which on the basis of dynamic voltage scaling (DVS) technique. DVS technique has become a rather promising critical technique in saving energy, which reduces system energy by adjusting CPU voltage/frequency dynamically under the condition of guaranteeing system's real-time performance.This paper summarizes and analyses research actuality of hard real-time energy-efficient scheduling algorithms at first, discussing correlative techniques and several classical algorithms of current hard real-time energy-efficient scheduling algorithms. It also makes research on energy-efficient scheduling about hard real-time period tasks and mixed tasks based on the DVS technique:(1) In the aspect of energy-efficient scheduling about hard real-time period tasks, aiming at the disadvantage of traditional open loop scheduling algorithm, this paper proposes a new energy-efficient scheduling algorithm MSF-DVS which introduces the average execution time into the traditional tasks scheduling model. On the one hand, we compute static slack time which every task may has offline, and sort the tasks based on maximal slack time first strategy; on the other hand, we combine the feedback control and dynamic voltage scaling techniques to adjust CPU speed dynamically. The simulation results show that this algorithm has good control performance and can reduce the whole energy consumption efficiently.(2) In the aspect of energy-efficient scheduling about hard real-time mixed task, we analyses the differences between mixed task and period task for hard real-time energy-efficient scheduling firstly, and then analyses a rather terse energy-efficientscheduling of hard real-time mixed task------OLDVS algorithm deeply, pointing outthe reason of this algorithm shortages. Then we address some methods to improve the effect of saving energy according the restricted factors. At last, we modify OLDVS algorithm adopting task splitting technique, and the theory analytical results suggest that the improve algorithm can save more energy.