earticle

영어

Different with traditional approaches that offload computation to a single remote server, the performance of an application can be further enhanced by simultaneously distributing its computing tasks to multiple computing resources around the mobile terminal. How to effectively multiply partition application components is critical for parallel computation offloading. In this paper, an application multi-partitioning scheme is proposed, which optimally offloads application's components to surrounding resources. A graph mapping model is converted and set up to represent components and resources as undirected graphs, and the A* algorithm is employed to efficiently search the optimal mapping from component graph to resource graph, which minimizes computing costs and inter-resource communication costs. Simulation results demonstrate that the performance can be efficiently promoted by our algorithm, which outperforms the traditional approaches to a large degree.