Energy-Aware Online Task Offloading and Resource Allocation for Mobile Edge Computing

Abstract

Mobile edge computing with the near-data processing paradigm can support applications requiring low latency and high computing capability, where energy cost is a significant part of the expenditure. This paper formulates and studies the problem of online joint task offloading and resource allocation for latency minimization subjecting to a time average energy cost constraint in mobile edge computing systems. The formulated problem has four time-variant system states, i.e., data lengths, task sizes, channel conditions, and electricity prices, which are modeled based on real-world data. At the beginning of each time slot, the system has to make five online decisions jointly: base station selection, server selection for task offloading, communication bandwidth allocation, computing resource allocation, and frequency scaling. We prove the offline version of the formulated problem is NP-hard. We design an online algorithm with a provable approximation ratio and low computational complexity for the proposed problem. In particular, it balances energy cost and latency based on the drift-plus-penalty algorithm and makes server and base station selection decisions using a game theoretic-based algorithm. We conduct extensive real-world data-driven simulations to evaluate the proposed algorithm. Simulation results show that the proposed approach outperforms popular baselines.