On the Optimal Provider Selection for Repair in Distributed Storage System with Network Coding

TL;DR

This paper addresses the problem of selecting optimal providers and routing data to minimize regeneration time in distributed storage systems using network coding, proposing an efficient near-optimal algorithm validated by simulations.

Contribution

It formulates the provider selection and data routing problem as an integer linear programming model and develops a near-optimal solution based on linear programming relaxation.

Findings

The proposed algorithm effectively reduces regeneration time.

Simulation results confirm the efficiency of the near-optimal approach.

The method outperforms baseline strategies in large-scale DSS.

Abstract

In large-scale distributed storage systems (DSS), reliability is provided by redundancy spread over storage servers across the Internet. Network coding (NC) has been widely studied in DSS because it can improve the reliability with low repair time. To maintain reliability, an unavailable storage server should be firstly replaced by a new server, named new comer. Then, multiple storage servers, called providers, should be selected from surviving servers and send their coded data through the Internet to the new comer for regenerating the lost data. Therefore, in a large-scale DSS, provider selection and data routing during the regeneration phase have great impact on the performance of regeneration time. In this paper, we investigate a problem of optimal provider selection and data routing for minimizing the regeneration time in the DSS with NC. Specifically, we first define the problem in the DSS with NC. For the case that the providers are given, we model the problem as a mathematical programming. Based on the mathematical programming, we then formulate the optimal provider selection and data routing problem as an integer linear programming problem and develop an efficient near-optimal algorithm based on linear programming relaxation (BLP). Finally, extensive simulation experiments have been conducted, and the results show the effectiveness of the proposed algorithm.