Comparison of Availability Figures of Distributed Systems Using Multiple Redundancy Methods

Abstract

Both hardware and software architectures have to meet the strict requirements for high availability. High availability requires considerations already in the design phase of the products. There are practical rules in system design, such as avoiding single points of failures. In general, the application of redundancy methods can increase overall availability of systems. In this paper we gather the most common redundancy solutions for parallel systems: systems or subsystems that consist of a number of identical units performing the same tasks. This paper lists the different redundancy concepts and gives several options for the possibilities of modelling. Therefore this paper investigates the difference between totally independent and load sharing parallel models as well. In addition to parallel redundancy we can introduce the concept of adding spare units to parallel systems. The general modelling and solution is presented in this paper as well with an introduced mathematical model. The model can be extended to the case of parallelly redundant subsystems in chain. For special cases exact formulae for the availability is presented. However, the interest is to show a general calculation method using Markov chain model. A possible method for the creation of the Markov chain’s transition-rate matrix in general cases is presented. The steady state solution of the master equation is shown using the defined transition-rate matrices. The comparison of different redundancy and sparing principles is presented with the possible calculation methods.