Performance-Energy Aggregate metric based Scheduler (PEAS) for Smartphones

Since the introduction of Cyber Foraging, there has been a need for a scheme that partitions computation among client and server for optimal performance without compromising energy consumption. In previous schemes, scheduler was installed on server resulting in low adaptability to changing network and local computational resources. MARS, the first adaptive, online and lightweight scheduler, achieved better performance as well as energy savings over these schemes by enhancing adaptability due to its installation on mobile device. We introduce PEAS (Performance-Energy Aggregate metric based Scheduler), an improved version of MARS, which is even more lightweight hence more adaptable. PEAS operates on the principle: keep both server and the client fully utilized at all times such that energy constraints are inherently taken care of. A thorough evaluation of PEAS over wide range of applications and network conditions is presented. We propose modifications in our scheduler to tackle high energy consumption for poor networks. Converging trend in performance of different schemes with network improvement is also demonstrated. Keywords— Pervasive computing, Cyber foraging, Clientserver, Adaptive, RPC, Scheduler