Int. J. of Industrial and Systems Engineering   »   2015 Vol.19, No.1

 

 

Title: Availability analysis of software rejuvenation in active/standby cluster system

 

Authors: Madhu Jain; Preeti

 

Addresses:
Department of Mathematics, Indian Institute of Technology Roorkee, Roorkee-247667, India
Faculty of Humanity, Physical and Mathematical Sciences, Shobhit University, Meerut-250110, India

 

Abstract: Due to continuous running of software applications for a long time, the computer system may be subjected to degradation or crash failure. This phenomenon is known as software ageing. Software rejuvenation is a proactive method to tackle and overcome from these types of problems. In this study, an active/standby cluster system in which personal computers are connected in loosely coupled fashion for providing an improved availability is studied. In cluster system configuration with active/standby components, there may be a primary server which can be treated as active server and other servers work as standby units of the cluster system. The concept of software rejuvenation has been incorporated to enhance and improve the performance of the embedded cluster system. A three level software rejuvenation policy is taken into consideration to develop a Markov model. The steady state probabilities of the model are obtained which are further used to obtain the availability and downtime cost of the active/standby cluster system. The results obtained are also compared by computing the same using adaptive neuro-fuzzy inference system (ANFIS) approach. A numerical illustration has been provided to validate the model.

 

Keywords: active/standby cluster systems; software ageing; software rejuvenation; rebooting; downtime costs; availability analysis; cluster configuration; Markov model; adaptive neuro-fuzzy inference system; ANFIS; modelling.

 

DOI: 10.1504/IJISE.2015.065948

 

Int. J. of Industrial and Systems Engineering, 2015 Vol.19, No.1, pp.75 - 93

 

Available online: 11 Nov 2014

 

 

Editors Full text accessAccess for SubscribersPurchase this articleComment on this article