Burst load scheduling latency optimization through collaborative content caching in edge-cloud computing

Burst load scheduling latency optimization through collaborative content caching in edge-cloud computing

연구 분야: Networking

학회: Cluster Computing

To meet the ever-increasing requirements of the applications, edge-cloud computing is introduced to shunt the services to the edge networks for execution and storage to achieve low-latency task processing. Due to the heterogeneity of edge networks, the traffic in the network will endure nonuniform and dynamic burst load all the time. To this end, this paper proposes a burst load scheduling method that aims to optimize the burst load scheduling latency through collaborative edge content caching. In our method, burst load scheduling is defined to be performed on a three-tier resource scheduling framework of user terminals, edge servers, and cloud data center. The task scheduling policy schedules the tasks to the appropriate edge server for execution in real time, while the collaborative content caching policy with the shortest latency is selected in the edge-cloud computing to schedule the data contents requested by the application during the execution of the tasks. Comparative experiments in a simulated edge-cloud environment show that our method outperforms NOSF, EVMS, and PSO in terms of data transfer, cache hit rate, task completion rate, and server load. The data transfer is stabilized at 99-102Mbps, the cache hit rate is stabilized at 82–84%, the task completion rate is improved by 1–4 percentage points, and the server load remains at a low level of about 44–47 tasks per second. That is, the method achieves significant improvements in reducing the overall processing latency of burst load evacuation through collaborative content caching.