A PERFORMANCE MODEL OF HIGHLY AVAILABLE MULTICOMPUTER SYSTEMS

S. Misbahuddin, N. Al-Holou, and S.T. Khan

Keywords

Availability, modelling, Markov models, fault tolerant computing,multicomputer systems, performance modelling

Abstract

Multicomputers are distributed memory systems in which independent computers are connected with each other by a network. Multicomputers are widely used in business and scientific applications. Due to their flexibility, scalability, and cost performance, multicomputer systems are considered preferable to multiprocessor systems. They are also gaining attraction in the area of Internetbased computing, in which delays or interruptions in client–server interactions may be equally costly to service providers and customers. Multicomputer systems attributed with high available (HA) solutions may be used to withstand service unavailability due to the web servers’ failures. In this paper we investigate a Markov model-based approach to estimate the availability of multicomputer systems with and without HA. The availability is computed as a function of server failure rate and the number of surviving servers. This model can estimate the number of nodes needed for a required availability measure.

Important Links:



Go Back