Bandwidth aware application partitioning for computation offloading on mobile devices

Computation offloading is a promising method for reducing power consumption of mobile devices by offloading computation to remote servers. For computation offloading, application partitioning is a key component. However, making a good application partitioning is challenging, as it needs to carefully consider the tradeoffs between the communication cost and computational benifits. Most of previous work makes application partitioning by using a static bandwidth to measure the communication cost and thus cannot adapt to scenarios with dynamic bandwidth. To address this problem, in this paper, we propose a Bandwidth Aware Application Partitioning Scheme (BAAP). BAAP models the bandwidth as a random variable and formulate the application partition as a 0-1 Integer Programming with Probability (IPP) problem. Then BAAP adopts Branch and Bound algorithm to solve the problem. Experimental results show that BAAP can greatly reduce energy consumption while satisfying the cost and time constraints with guaranteed confidence probabilities regardless of different network bandwidth.