Date of Publication :31st May 2023

Abstract: In the applications of Internet of Things (IoT), the computing and power resources of the end devices limit the deployment in real life especially for the real time services. Although the cloud server provides huge computing and storage resources, the transmission latency and its uncertainty due to the network congestion cause the difficulty in the provision of services. The edge server is then a possible way toward this problem. However, the time delay tolerance if some computing bound tasks may still not be satisfied due to the computing power of the offloaded server. Then it needs more than one server to execute its subtasks in parallel if the tasks can be partitioned into several independent subtasks. This paper studies the subtasks parallel offloading for local user equipment (UE), neighbor device to device (D2D) UE, and mobile edge computing (MEC) server by using the concept of 0/1 knapsack model. The main objective of the proposed subtask offloading for parallel processing (SuOPP) scheme is proposed to increase the acceptance ratio of task offload requests under the sustained delay. As the costs of different processing units may be different according to their processing speed, the proposed scheme also takes the cost into consideration. The simulation results illustrate that the proposed SuOPP scheme achieves higher acceptance ratio and lower cost when comparing to the other scheme. Although the average waiting time of the proposed scheme is a little higher than the compared scheme, it still satisfies the desired delay constraint.

Reference :