Elastic Offloading of Multitasking Applications to Mobile Edge Computing

Abstract

This paper focuses on offloading multitasking applications to a Mobile Edge Computing (MEC) environment. We address this issue through a mobile user's perspective that is seeking to obtain the execution result of a resource-hungry multitasking application, possibly through offloading some tasks to a mobile multiple edge servers' environment. The completion time of the application is constrained by a predefined strict deadline and the offloading decision aims to minimize the terminal's energy consumption. Moreover, the multitasking application is modeled by a weighted Directed Acyclic Graph (DAG) which characterizes the dependencies between tasks. To tackle this issue, we first model the multitasking offloading problem in a mobile multiple edge servers' environment as a Zero-one Integer Programming problem. Then, we propose an efficient adaptive offloading algorithm, named eTOMEC (Elastic Tasks graph Offloading for MEC), which decides which task must be offloaded and accordingly selects the edge server that will perform the task execution. Compared to many state-of-the-art offloading approaches, our proposal allows parallel offloading of tasks to several edge servers. Assessment results show that our proposal achieves better performances in terms of completion time for the application and energy consumption of the user's terminal.