Back to Search Start Over

A Hybrid Heuristic Algorithm Using Artificial Agents for Data Replication Problem in Distributed Systems

Authors :
Bahman Arasteh
Seyed Salar Sefati
Simona Halunga
Octavian Fratu
Tofigh Allahviranloo
Source :
Symmetry, Vol 15, Iss 2, p 487 (2023)
Publication Year :
2023
Publisher :
MDPI AG, 2023.

Abstract

One of the key issues with large distributed systems, such as IoT platforms, is gaining timely access to data objects. As a result, decreasing the operation time of reading and writing data in distributed communication systems become essential demands for asymmetric system. A common method is to replicate the data objects across multiple servers. Replica placement, which can be performed statically or dynamically, is critical to the effectiveness of distributed systems in general. Replication and placing them on the best available data servers in an optimal manner is an NP-complete optimization problem. As a result, several heuristic strategies for replica placement in distributed systems have been presented. The primary goals of this research are to reduce the cost of data access time, reduce the number of replicas, and increase the reliability of the algorithms for placing replicas. In this paper, a discretized heuristic algorithm with artificial individuals and a hybrid imitation method were developed. In the proposed method, particle and gray-wolf-based individuals use a local memory and velocity to search for optimal solutions. The proposed method includes symmetry in both local and global searches. Another contribution of this research is the development of the proposed optimization algorithm for solving the data object replication problem in distributed systems. Regarding the results of simulations on the standard benchmark, the suggested method gives a 35% reduction in data access time with about six replicates. Furthermore, the standard deviation among the results obtained by the proposed method is about 0.015 which is lower than the other methods in the same experiments; hence, the method is more stable than the previous methods during different executions.

Details

Language :
English
ISSN :
15020487 and 20738994
Volume :
15
Issue :
2
Database :
Directory of Open Access Journals
Journal :
Symmetry
Publication Type :
Academic Journal
Accession number :
edsdoj.5ef316c32f1c46e482f2346e92a8bbca
Document Type :
article
Full Text :
https://doi.org/10.3390/sym15020487