Dynamic resource allocation strategy for latency-critical and computation-intensive applications in cloud–edge environment

Abstract

Edge computing is more and more popular due to its low latency and bandwidth-efficient services. Edge computing is mainly applied to the latency-critical and computation-intensive application. However, there are several challenges to the improvement on the quality of service in edge computing environment. For instance, the reduction of server latency, the network transmission efficiency, etc. In this paper, we propose the dynamic resource allocation algorithm for cloud–edge environment. The dynamic resource allocation algorithm consists of the resource scheduling algorithm and the resource matching algorithm. In the resource scheduling algorithm, a resource scheduling problem can be obtained according to the stored penalty of scheduling contents, the value of scheduling contents and the transmission cost of scheduling contents. Then, tabu search algorithm is applied to find the optimal solution to the resource scheduling problem. Furthermore, the resources are scheduled into the edge servers from cloud datacenter with the optimal solution. In the resource matching algorithm, an optimization problem of the resource matching is built with respect to the resource location, the task priorities and the network transmission cost. For addressing this problem, the optimal problem is converted to an optimal matching problem of the weighted bipartite graph. Moreover, an optimal matching problem of the weighted complete bipartite graph is created by adding the spurious containers. Then, the optimal strategy of the resource matching for tasks on the edge servers is achieved. Finally, the performance of the proposed algorithms and some typical resource allocation algorithms is evaluated via extensive experiments. The results indicate that proposed algorithms can effectively reduce network delay and enhance QoS.