Partial Replication-Enabled Cooperative Transmission for Distributed Edge Computing

With the proliferation of delay-sensitive and computation-intensive mobile applications, recent years have witnessed a transition towards more advanced distributed edge computing. However, the diverse computation results among different edge nodes (ENs) in conventional distributed edge computing present challenges to cooperative transmission and may result in prolonged downlink latency. In this regard, replicated edge computing has emerged as a promising solution to perform computations repeatedly across multiple ENs, thereby facilitating cooperative transmission in the downlink. Nonetheless, this approach inevitably leads to additional computation cost. To encompass the advantages of both distributed edge computing and replicated edge computing, a novel partial replication based distributed edge computing is proposed in this work. Specifically, by partitioning the original task into a common part computed at all ENs and multiple private parts, with each computed by one EN, cooperative downlink transmission can be enabled without incurring excessive computation cost. To minimize the overall processing delay, a task partition and downlink transmit power allocation algorithm is developed based on alternating optimization. Simulation results are presented to validate the superiority of the proposed scheme.