LI-RMLO: Loop Iteration based Runtime Method Level Offloading for Mobile Applications

Application partitioning is of particular importance to address the long communication delay issue which occurs during offloading from mobile device to cloud. With the launch of mobile edge framework, the connection delay tends to get reduced to a certain extent and paves the way for utilizing neighbourhood competent mobile devices for taking care of intensive computations of fragile devices. While Mobile Edge Computing (MEC) transforms the Mobile cloud computing in one way, the improvisation in shortening the granularity of code to be offloaded is raising on the other side. Method Level Offloading (MLO) is one such solution of code granularity reduction which is discussed in many research works. Our proposed model, Loop Iteration based Runtime Method Level Offloading (LI-RMLO) adopts MLO to offload applications having iterative and independent methods that behaves diverse to smaller and larger inputs. In such cases, the method which is labeled as offloadable never needs to be executed in the remote machine for all runs. Instead our model analyzes the iteration limit of the method at runtime and decides to offload only when the local execution time of the method exceeds its remote execution time. By this adaptation, we achieve an optimal offloading performance and ward off unwanted offloading to the proximate edge nodes.