A task processing efficiency improvement scheme based on Cloud-Edge architecture in computationally intensive scenarios

Abstract

As Cloud-Edge architecture combines low latency and high-performance computing, it has a wider range of application scenarios. Considering that, how to reasonably allocate tasks to the edge cloud or the center cloud and how to select appropriate virtual machines (VMs) for tasks are two key aspects affecting the efficiency of Cloud-Edge architecture. For the former, the ARDT algorithm is proposed, which dynamically adjusts the optimal threshold by predicting the number of real-time end devices. For the latter, the DBMR algorithm is designed based on the Max-Fit algorithm, in which a structure of the blocking table is put forward to reduce the time complexity of the original algorithm from O(n2) to O(n(n−m)). Experimental results show that the ARDT-DBMR algorithm optimizes the average service time and task failure rate by 30%-40% compared with the baseline algorithms and other advanced algorithms and significantly ameliorates the problem of abnormal fluctuations in VM utilization in computationally intensive scenarios.