Graph-Based Application Partitioning Approach for Computational Offloading in Mobile Cloud Computing

Background: Computational offloading is emerging as a popular field in Mobile Cloud Computing (MCC). Modern applications are power and compute-intensive which leads to the energy, storage and processing issues in mobile devices. Using the offloading concept, a mobile device can offload its computation to the cloud servers and receives back the results on the device. Objective: The main objective of the work is to provide a solution of an important question that arises in the offloading scenario is that which part of the application needs to be offloaded remotely and which part would run locally. Methods: In order to identify remote and local code, the application needs to be partitioned. In this paper, the graph partitioning approach is considered which is based upon the spectral graph partitioning with the Kernighan Lin algorithm. An application is assumed to be a graph and each node of the graph is assumed as a method. Results: Experimental results show that the proposed hybrid approach performs optimally in partitioning the application. The results indicate that considering the combination of spectral approach with the Kernighan Lin algorithm performs optimally as compared to random and multilevel partitioning in a mobile cloud scenario. Conclusion: The proposed technique gave better results than the existing techniques in terms of edge cut which is less, concluding minimum communication cost among components and thus save energy of the mobile device.