A dynamic decentralized strategy of replica placement on edge computing

Abstract

Smart phone and its applications are used more and more extensively in our daily life. Short delay of arriving data is important to these applications, especially to some time-sensitive ones. To reduce transmission latency and improve user experience, a dynamic decentralized data replica placement and management strategy which works in edge nodes is proposed in this article. It studies the location, access frequency, latency improvement, and cost spent on placing replicas on edge nodes to seek a balance between cost spent for storage and reduced latency. Specifically, dynamic and decentralized replica placement strategy algorithm has load guarantee for edge nodes to avoid overload; it dynamically create or delete data replicas on edge nodes according to the request frequency. Dynamic and decentralized replica placement strategy is decentralized to relieve transmission cost. Experiment results show that dynamic and decentralized replica placement strategy algorithm in edge computing environments can greatly reduce transmission latency, balance edge nodes load, and improve system performance. Dynamic and decentralized replica placement strategy also considers the cost spent for storage, and it pursues a balance between many factors.

Integration of Iot With Cloud, Fog, and Edge Computing: A Review

“…Despite its prevalent use, edge computing, like other data distribution and storage systems, faces challenges related to the growing volumes of data in today's digital era. To address concerns about latency-aware data distribution at the edge of a network, researchers [36] have proposed a distributed information dissemination strategy. This strategy relies on the dynamic creation, replacement, and removal of data replicas through continuous analysis of data requests received from edge network nodes.…”

“…This strategy relies on the dynamic creation, replacement, and removal of data replicas through continuous analysis of data requests received from edge network nodes. The process [36] comprises two versions, namely the source and edge versions. The source version operates solely in the node housing the central storage, producing copies of data items in that proximate node.…”

A practical analysis of balancing policies for rearranging data replicas in HDFS clusters

“…In [Yin and Deng 2022], the authors investigated methods for placing replicas in edge-cloud environments. Based on the analyses, a strategy was proposed in which the replicas are placed on edge nodes based in their load and on the cost-effective value.…”