QoS-based Dynamic Clustering Deployment Strategy in MEC
Yangyang Zha, Qibo Sun · 2022
In Mobile Edge Computing (MEC), User Equipment (UE) can access edge servers with low latency. However, resources in edge servers are limited and the mobility of UE leads to high variability. How to optimize service deployment is crucial. In this paper, the above problem is expressed as an optimization problem for the global average delay and a Dynamic Clustering Deployment Strategy (DCDS) is proposed. According to the multi-dimensional constraints of geographic location, bandwidth, computing resources and reliability, servers are divided into several clusters by the Bisecting K-Means clustering algorithm. In addition, focusing on resource constraints, services are preferentially deployed to servers with idle resources on the cluster to complete static service deployment to ensure system load balancing capabilities. Considering the dynamism of user equipment, this paper proposes the variation of user requests and adopts the dynamic service deployment strategy to update the deployment of the original services. The feasibility of the strategy is verified by experiments on real datasets. Based on the experimental results, compared with the benchmark algorithm, the dynamic clustering deployment strategy can reduce the average interaction delay and achieve stable load balance.