Availability analysis of a distributed system with homogeneity in client and server under four different maintenance options

Abstract

System availability is a major performance in distributed system design and analysis. For reliability evaluation and system performance, this study analyzed a distributed system consisting of two hosts (clients) connected to two servers via a load balancer. The system is configured as series-parallel system consisting of three subsystems A, B and C. Subsystem A consists of two clients working in parallel with each other, subsystem B has two active parallel servers, while subsystem C consists of a load balancer. Clients, Servers and LD balancer failure and repair time are to be exponentially distributed. The system is under four different maintenance models: Model 1 is the system with perfect repair and replacement at total failure, Model 2 is repairable system with minimal repair at failure, Model 3 is the system with replacement at total failure and online preventive maintenance (PM) at partial failure and Model 4 is fault-tolerant system with replacement at total failure. The system is analyzed using first-order differential difference equations to derive the expressions for availability. The model is implemented using Java programming language. To make the computation and analysis of systems availability more efficient, numerical examples are presented to illustrate the performance of the model. Impact of both clients and server's failure and repair rates on availability is determined and presented in figures and graphs. The analysis of the results obtained with respect to the availability indicates that systems which are fault tolerant perform best and have the highest availability.