Energy-Efficient Real-Time Scheduling For Periodic Tasks On Multiprocessor System

How to effectively reduce energy consumption has become an important issue in multi-processor system designing.Many existing algorithms are applied dynamic voltage and frequency scaling(DVFS)techniques to adjust each processor voltage or frequency,but this only considers the dynamic power consumption of processor execution while ignoring the static power consumption in multi-processor systems.With the development of processor technology,static energy consumption's share in energy consumption is increasing.Considering the dynamic and static power consumption of the processor,this article presents a leakage-aware scheduling algorithm,which can select a minimum number of processors and CPU frequency.After that,we not only proved feasibility and correctness of the algorithm,but also shown the practicability and validity of the algorithm.The main contribution of this article as follows:In this paper,we build a leakage-aware scheduling algorithm called LASER This algorithm extends the DVFS technique by addressing the static energy consumption in multiple processor systems running periodic tasks.LASER consists two principal components.The first one determines the minimum number of processors that can guarantee timing constraints of a set of periodic tasks.The second one is in charge of assigning dividable tasks to multiple processors.LASER also judiciously reduces energy-saving overhead by preventing voltages from frequently changing.We implement LASER in a simulated multiprocessor system equipped with three real-world processors-the Intel PXA270 and PXA255 processor as well as the Crusoe TM5800 processor.Our experimental results show that LASER reduces the energy of DVFS-enabled multiprocessor systems by up to 13% with an average of 10%.LASER exhibits better energy efficiency on PXA270 and TM5800 than on PXA255;however,task sets require more PXA270 processors than PXA255 to complete the tasks before their deadlines.