An Optimal Tree-Structured Repair Scheme of Multiple Failure Nodes for Distributed Storage Systems

To reduce recovery cost of repairing multiple failed nodes, many repair schemes have been proposed for erasure codes based distributed storage systems. However, most of the existing researches ignore the network topology of storage devices. Motivated by such considerations, we combine delay repair s...

Full description

Bibliographic Details
Main Authors: Anan Zhou, Benshun Yi, Yusheng Liu, Laigan Luo
Format: Article
Language:English
Published: IEEE 2021-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/9336728/
id doaj-859d1e7f009a49b1aa83965d5eb0f2e5
record_format Article
spelling doaj-859d1e7f009a49b1aa83965d5eb0f2e52021-05-19T23:03:00ZengIEEEIEEE Access2169-35362021-01-019218432185810.1109/ACCESS.2021.30549549336728An Optimal Tree-Structured Repair Scheme of Multiple Failure Nodes for Distributed Storage SystemsAnan Zhou0https://orcid.org/0000-0003-3800-8254Benshun Yi1https://orcid.org/0000-0002-2818-9357Yusheng Liu2https://orcid.org/0000-0002-4286-4670Laigan Luo3https://orcid.org/0000-0001-6960-5006School of Electronic Information, Wuhan University, Wuhan, ChinaSchool of Electronic Information, Wuhan University, Wuhan, ChinaSchool of Electronic Information, Wuhan University, Wuhan, ChinaSchool of Electronic Information, Wuhan University, Wuhan, ChinaTo reduce recovery cost of repairing multiple failed nodes, many repair schemes have been proposed for erasure codes based distributed storage systems. However, most of the existing researches ignore the network topology of storage devices. Motivated by such considerations, we combine delay repair schemes with network topology and propose a tree-structured model based on fountain codes with large value of (n, k, r) to improve the repair efficiency. More precisely, with the consideration of network topology, a new target named data recovery cost is defined to measure the efficiency of coded fragment download and source file reconstruction, and then the optimal recovery threshold is derived to minimize the average data recovery cost of general tree-structured model. Moreover, we analyze and compare the average data recovery cost of general tree-structure with different systematic parameters. To further improve the data transmission efficiency, an optimal tree-structured scheme based on improved tabu search algorithm (ITSAORT) is proposed. Compared with other algorithms, the ITSA-ORT scheme uses Prim algorithm to generate the initial solution and then uses special method to obtain the corresponding neighborhood structure. The experimental results show that the proposed scheme can find a globally optimal solution and obtain lower cost of data recovery. In addition, the ITSA-ORT scheme has lower computational complexity than the optimal tree-structured scheme based on particle swarm optimization algorithm (PSO-ORT) and the optimal tree-structured scheme based on firefly algorithm (FA-ORT).https://ieeexplore.ieee.org/document/9336728/Centralized communication modelfountain codesdelay repairnetwork topologytree-structured modeloptimization algorithms
collection DOAJ
language English
format Article
sources DOAJ
author Anan Zhou
Benshun Yi
Yusheng Liu
Laigan Luo
spellingShingle Anan Zhou
Benshun Yi
Yusheng Liu
Laigan Luo
An Optimal Tree-Structured Repair Scheme of Multiple Failure Nodes for Distributed Storage Systems
IEEE Access
Centralized communication model
fountain codes
delay repair
network topology
tree-structured model
optimization algorithms
author_facet Anan Zhou
Benshun Yi
Yusheng Liu
Laigan Luo
author_sort Anan Zhou
title An Optimal Tree-Structured Repair Scheme of Multiple Failure Nodes for Distributed Storage Systems
title_short An Optimal Tree-Structured Repair Scheme of Multiple Failure Nodes for Distributed Storage Systems
title_full An Optimal Tree-Structured Repair Scheme of Multiple Failure Nodes for Distributed Storage Systems
title_fullStr An Optimal Tree-Structured Repair Scheme of Multiple Failure Nodes for Distributed Storage Systems
title_full_unstemmed An Optimal Tree-Structured Repair Scheme of Multiple Failure Nodes for Distributed Storage Systems
title_sort optimal tree-structured repair scheme of multiple failure nodes for distributed storage systems
publisher IEEE
series IEEE Access
issn 2169-3536
publishDate 2021-01-01
description To reduce recovery cost of repairing multiple failed nodes, many repair schemes have been proposed for erasure codes based distributed storage systems. However, most of the existing researches ignore the network topology of storage devices. Motivated by such considerations, we combine delay repair schemes with network topology and propose a tree-structured model based on fountain codes with large value of (n, k, r) to improve the repair efficiency. More precisely, with the consideration of network topology, a new target named data recovery cost is defined to measure the efficiency of coded fragment download and source file reconstruction, and then the optimal recovery threshold is derived to minimize the average data recovery cost of general tree-structured model. Moreover, we analyze and compare the average data recovery cost of general tree-structure with different systematic parameters. To further improve the data transmission efficiency, an optimal tree-structured scheme based on improved tabu search algorithm (ITSAORT) is proposed. Compared with other algorithms, the ITSA-ORT scheme uses Prim algorithm to generate the initial solution and then uses special method to obtain the corresponding neighborhood structure. The experimental results show that the proposed scheme can find a globally optimal solution and obtain lower cost of data recovery. In addition, the ITSA-ORT scheme has lower computational complexity than the optimal tree-structured scheme based on particle swarm optimization algorithm (PSO-ORT) and the optimal tree-structured scheme based on firefly algorithm (FA-ORT).
topic Centralized communication model
fountain codes
delay repair
network topology
tree-structured model
optimization algorithms
url https://ieeexplore.ieee.org/document/9336728/
work_keys_str_mv AT ananzhou anoptimaltreestructuredrepairschemeofmultiplefailurenodesfordistributedstoragesystems
AT benshunyi anoptimaltreestructuredrepairschemeofmultiplefailurenodesfordistributedstoragesystems
AT yushengliu anoptimaltreestructuredrepairschemeofmultiplefailurenodesfordistributedstoragesystems
AT laiganluo anoptimaltreestructuredrepairschemeofmultiplefailurenodesfordistributedstoragesystems
AT ananzhou optimaltreestructuredrepairschemeofmultiplefailurenodesfordistributedstoragesystems
AT benshunyi optimaltreestructuredrepairschemeofmultiplefailurenodesfordistributedstoragesystems
AT yushengliu optimaltreestructuredrepairschemeofmultiplefailurenodesfordistributedstoragesystems
AT laiganluo optimaltreestructuredrepairschemeofmultiplefailurenodesfordistributedstoragesystems
_version_ 1721436238217478144