Pinning based Energy Aware Computation Offloading for Mobile Cloud Computing

Abstract

Nowadays mobile devices are becoming in each other's pocket in part of human life. Most of us preferred the mobile as a good platform for their computation. But Mobile devices are still experienced with frontier resources like CPU performance and battery life when executing computation intensive applications. The exponential growth in this field necessitates the Technology Acceptance model. To overcome these limitations, the computation intensive jobs are migrated to rich resourceful server. This popular approach in mobile cloud computing is called as computational offloading. To find the computationally intensive jobs in an application, static and dynamic energy profiling is performed. Applications are represented as Task integration graph (TIG), the tasks in TIG can be executed on the mobile device or migrated to nearby resourceful server. In this paper we proposed P-ECOM algorithm for cloudlet-based image processing over quad-core machine to measure the energy profile of an image compression step. Towards the goal, each of the three image file sizes 276MB, 546MB, 715.6MB are partitioned into 1 to 4 splits. Compression process are executed in two phases, with pinned with specific core and without pinning. Measure the energy consumption for compressing each partition of image by pinning each process with specific CPU core. And also measure energy for without pinning the process with specific core. This research work gives an important insight into energy consumption pattern which concludes that if all the available cores in the offloaded devices are evenly loaded and pinned with specific core then energy consumption is minimum than without pinning the process with specific core. As per the proposed Technology Acceptance model, these energy profile details are integrated into partitioning engine. This will dynamically partition the applications based on the number of cores available and also pinned with specific core in target offloading devices to minimize the energy consumption.