Peer Offloading With Delayed Feedback in Fog Networks

Comparing to cloud computing, fog computing performs computation and services at the edge of networks, thus relieving burden data center reducing task latency end devices. Computation is a crucial performance metric in especially for real-time applications. In this article, we study peer offloading problem network with unknown dynamics. scenario, each node (FN) can offload its tasks neighboring FNs time slot manner. The latency, however, could not be fed back dispatcher instantaneously due uncertainty processing FNs. Besides, competition occurs when different one FN same time. To tackle above difficulties, model as sequential selection delayed information feedback. Using adversarial multiarm bandit framework, construct an online learning policy deal Different contention resolution approaches are considered resolve competition. Performance analysis shows that regret proposed algorithm, or loss suboptimal selections, achieves sublinear order, suggesting optimal policy. prove strategy result Nash equilibrium (NE) all playing Simulation results validate effectiveness