Maximization of Value of Service for Mobile Collaborative Computing Through Situation-Aware Task Offloading

Mobile collaborative computing (MCC) is an emerging platform for effectively improving the quality of mobile service by exploiting the idling computational resources in distributed mobile devices (MDs) through peer-to-peer task offloading. Recently, diverse MCC applications have been developed to provide multiple functional benefits and individualized value to users. In this paper, we propose to use a new concept of value of service (VoS) to represent the total value of all tasks and devices with respect to their performance including latency and energy consumption. To improve service provisioning under fast-varying conditions, a situation-aware offloading scheme is proposed to maximize VoS by opportunistically leveraging the changing resource availability conditions. Specifically, we consider a collaborative computing system where a user can offload input data of computation to other available MDs. VoS maximization for two popular offloading scenarios, i.e., binary and partial offloading, are formulated separately. Decision making of binary offloading is an NP-hard problem and solved by a novel heuristic algorithm which achieves suboptimal solution in polynomial time. Partial offloading is formulated as a non-convex problem involving task partition decision. By exploiting the unique characteristics of the problem, we propose an adapted barrier method (ABM) which achieves significant improvements in convergence efficiency.

Automated summary

Mobile collaborative computing (MCC) is a new platform that effectively improves the quality of mobile service by utilizing idle computational resources in distributed mobile devices through peer-to-peer task offloading. This paper proposes a concept of value of service (VoS) to represent the total value of tasks and devices based on their performance. A situation-aware offloading scheme is proposed to maximize VoS by leveraging changing resource availability conditions. The paper also presents solutions to maximize VoS for binary and partial offloading scenarios.