Title: Performance evaluation and reliability analysis of predictive hardware failure models in cloud platform using ReliaCloud-NS

Authors: Rohit Sharma; Raghuraj Singh

Addresses: Department of Computer Science and Engineering, Dr. Ambedkar Institute of Technology for Handicapped U.P., 208024, Kanpur, India ' Department of Computer Science and Engineering, Harcourt Butler Technical University, 208002, Kanpur, India

Abstract: Cloud computing systems at the present time established as a promising trend in providing the platform for coordinating large number of heterogeneous tasks and aims at delivering highly reliable cloud computing services. It is most necessary to consider the reliability of cloud services and timely prediction of failing hardware in cloud data centres so that it ensures correct identification of the overall time required before resuming the service after the failure. In this paper reliability of two recently introduced predictive hardware failure models has been analysed. The first model is on the basis of two open data sources, i.e., self-monitoring, analysis and reporting technology (SMART), windows performance counters and the second model is based on FailureSim which is a neural networks-based system for predicting hardware failures in data centres is done over our carefully designed two test cloud simulations of 144 VMs and 236 VMs. The results are thoroughly compared and analysed with the help of ReliaCloud-NS that allow researchers to design a CCS and compute its reliability.

Keywords: cloud computing system; CCS; virtual machines; VM; Monte Carlo simulation; MCS; neural networks; annual failure rate; AFR; self-monitoring, analysis and reporting technology; SMART.

DOI: 10.1504/IJCC.2021.118015

International Journal of Cloud Computing, 2021 Vol.10 No.3, pp.207 - 224

Received: 12 Dec 2018
Accepted: 13 Sep 2019

Published online: 24 Sep 2021 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article