Research On Service Deployment Problem In Cloud-Edge Collaborative Environment

Since 2022,China has implemented the "East Data,West Computing" strategy,which aims to systematically guide data processing demands from the eastern to the western regions,to address the issue of imbalanced data processing demand and supply between the two regions.Edge computing technology is an essential component of this strategy,which can provide users with lower task processing latency and stronger privacy protection.In the future,the central and western regions will become the cloud computing centers,connecting edge computing nodes across the country,and forming a new digital infrastructure pattern.However,the limited resources of edge computing devices pose two significant challenges in a production environment: 1)how to design service deployment strategies that make efficient use of the processing capabilities of both cloud and edge computing,providing higher-quality services to users;and 2)how to optimize the resource utilization of edge clusters,given the limited resources of individual edge devices and the small resource scale of edge device clusters,to maximize the processing capabilities of edge computing.This paper focuses on the research of service deployment strategies in the cloudedge collaborative environment,and the main contributions are summarized as follows:(1)In order to reduce the response delay of tasks,this article proposes a selfadaptive service deployment algorithm based on speedup ratio weights(SWD-AD)for cloud-edge collaborative service clusters.Firstly,by comparing the execution time of tasks in the cloud and at the edge,the speedup ratio is calculated,and a service deployment algorithm that comprehensively considers speedup ratio weights and resource consumption weights is designed.Then,during the operation of the cluster,the task processing information of each service is collected,and the cumulative speedup ratio weight is calculated.A dynamic service adjustment strategy based on the cumulative speedup ratio weight is used to migrate services between the cloud and the edge.Performance evaluation experiments show that the service adjustment strategy can significantly reduce the average response time of tasks,reducing it by 29.38% and25.86% compared to Swarm algorithm and K8 s algorithm,respectively.In addition,the effectiveness of the service adjustment strategy is affected by the degree of balance of task types and is suitable for scenarios with unbalanced task types.(2)In order to improve the resource utilization of edge clusters,this paper proposes a workload co-location strategy(CWD)suitable for edge clusters.Firstly,tasks are classified into two categories: resource-intensive and latency-sensitive,and a taskbased service deployment strategy is proposed.By monitoring the resource utilization of edge nodes,services are assigned to deployment nodes when resource-intensive tasks arrive.Then,to prevent resource-intensive tasks from affecting the execution of latency-sensitive tasks,a process scheduler Batch is designed specifically for resourceintensive tasks.Evaluation experiments on physical clusters show that compared to directly deploying resource-intensive tasks,the CWD algorithm reduces the average latency by 9.71% and increases CPU resource utilization by 20.42%.Single-machine performance tests show that the Batch scheduler has excellent isolation performance under different load intensities.