Mobile Crowd Computing with Work Stealing

Abstract

By pooling together the processing power of mobile devices within a crowd to form a 'mobile cloud', these devices be efficiently utilized to help realize the full potential of mobile computing. However, the dynamic nature of mobile computing makes sharing and coordinating work non-trivial. Although never been used before in the mobile computing domain, the concept of work stealing possesses useful traits such as self adaptiveness, and decentralized nature that can help with these issues. Here we explore this concept of 'work stealing' for crowd computing on an opportunistic network of mobile devices, for both machine and human computation. We also present experimental data and discuss the findings.