Dependency-Aware Parallel Offloading and Computation in MEC-Enabled Networks

This letter proposes an effective dependency-aware subtask offloading and computation scheme in mobile edge computing (MEC) enabled networks, where we construct sequential execution windows of the MEC server by incorporating the hybrid dependencies among subtasks of an application, and the subtasks within the same subtask execution window could be offloaded and computed in parallel. By doing so, the completion delay and energy consumption of the application can be reduced. To make the proposed scheme achieve the minimum completion delay and energy consumption, we further jointly optimize the transmission rate and start execution time of subtasks within each execution window. The formulated problem is non-convex and difficult to solve. To make it tractable, we design a bisection search and successive convex approximation based iterative algorithm. Simulation results validate that compared with other existing schemes, our proposed scheme could reduce the summation of the completion delay and energy consumption by 10.36%.