A Group-based Fault Tolerant Mechanism for Heterogeneous Mobile Clouds

Recent developments of the enhanced hardware on mobile devices such as quad-core CPUs and various sensors have made it possible to build a powerful heterogeneous mobile cloud o„oading service that consists of a mobile ad-hoc cloud, nearby servers and public cloud services. However, the availability and mobility management of mobile devices in the network can signi€cantly hinder the performance of mobile cloud systems due to the frequent system faults caused by dynamic changes, and prevent applications from o„oading to mobile ad-hoc networks. In order to improve the mobile cloud service reliability, we propose a group based fault tolerant mechanism GFT-mCloud that classi€es mobile devices into groups based on its processing capacity, mobility, and reliability. Di‚erent fault tolerance techniques are then devised adaptively based on the task o„oading schedules and the speci€c group of machines it’s o„oaded. GFT-mCloud is designed as a standalone module that can work with existing mobile cloud code o„oading systems. Extensive experiments have been conducted to evaluate the proposed mechanism. Œe results show that our fault tolerant mechanism is able to outperform conventional fault tolerant algorithms in the mobile cloud o„oading environment.