摘要： 目前主流的边缘存储策略通过协同或非协同的方式来提高存储资源的请求命中率，从而降低请求延迟以满足时间敏感型业务的需求，然而这些策略并未考虑存储节点的副本数量过多所带来的一致性开销问题。提出一种基于虚拟传播树（VST）的边缘存储算法，针对边缘存储中的一致性开销问题，设计VST生成算法和节点淘汰算法，从而在副本数量高、一致性需求大的场景下实现可控低延迟服务。实验结果表明，该算法可以在一致性场景下提供低延迟服务，在非一致性场景下同样具有稳定的性能表现，请求时延和存储开销低于CV和NCV算法。

关键词: 边缘存储, 请求延迟, 一致性场景, 副本数量, 虚拟传播树

Abstract: The current mainstream edge storage strategies use cooperative or non-cooperative methods to increase the hit rate of storage resource requests,and thereby reduce request latency to meet the requirements of time-sensitive services.However,these strategies do not take into account the consistency overhead caused by the large number of copies of the storage node.To address the consistency problem,this paper presents the design of an edge storage algorithm based on Virtual Spread Tree(VST).The VST generation algorithm and node elimination algorithm are designed to implement controllable low-latency services in the scenarios with a large number of copies and high consistency requirements.Experimental results show that the proposed algorithm can provide low-latency services in consistency scenarios,and also provides stable performance in non-consistency scenarios.At the same time,it outperforms CV and NCV algorithms in request delay and storage overhead.

Key words: edge storage, request latency, consistency scenarios, number of copies, Virtual Spread Tree(VST)

中图分类号: 

• TP393