Mobile Computation Bursting: An application partitioning and offloading decision engine

Abstract

Most of the today's Smartphones use multithreading and execute several application jobs in parallel. The Mobile Computation Bursting (MCB) exploits this nature of the Smartphones and aims at partitioning the jobs of a mobile application into different clusters consisting of high computation jobs from that which requires less computation, based on their frequency requirement to compute a task. The nature of a job, i.e., the frequency requirement is identified by Probability Distribution Function (PDF), that represents the number of cycles required for each job to complete the task. The novel algorithm proposed in this paper classifies these jobs using the density-based clustering algorithm using KL divergence. The offloading algorithm proposed in this paper decides whether to execute the cluster in the device or offload to the Cloud. The interaction and transportation of code and data between the mobile device and Cloud get communicated via a mobile agent, thus providing service to mobile users, even when the device moves away from the vicinity of the wireless network. The Mobile Computation Bursting technique is compared with the traditional offloading algorithms, and the results reveal that MCB proves to be more efficient and beneficial to offload the computation to Cloud.