A Mapping-based Dynamic Semi-Online Task Scheduling Method for Minimizing Energy in Edge Computing.

Semi-online task scheduling in the edge computing platform refers to a scheduling scenario where there exists unknown performance computing nodes in the system. Existing online and offline task scheduling methods may lead to long makespan or transmission time due to the influence of unknown nodes, and existing semi-online methods are not fully applicable to the scenario discussed in this paper, which aggravates the problem of high energy consumption in edge computing plat-form. To solve this problem, this paper proposes a Mapping-based Dynamic Semi-online Scheduling(MDSS) strategy for edge computing platform. Firstly, we consider three main factors that affect the energy consumption in edge computing platform, that is, the energy consumption of task execution, task transmission and idle state. Then an energy consumption optimization problem is established from the perspectives of the processing speed, routing delay and queuing delay of the edge nodes, which correspond to those three factors. Secondly, for the unknown node, this paper introduces a mapping mechanism, which maps the performance of unknown nodes to a certain given node. Then the mapping relations are dynamically adjusted by continuously checking the task queue lengths of the mapped two-sides. Therefore, the prior knowledge of known nodes can be fully used. Thirdly, an im-proved Power-of- D algorithm based on the mapping mechanism is proposed to solving the formulated problem. It utilizes the prior knowledge of known node performance, which reduces task processing time and transmission time, so that the energy consumption can be reduced. Finally, some experiments are conducted to evaluate MDSS. Results show that the dynamic mapping mechanism is effective in the semi-online scenario and MDSS can greatly reduce the energy consumption in the edge computing platform.