Comparison of Load Sharing and Unit Sparing redundancy models for Cloud Network Elements

Abstract

In a network of computing units (i.e. computer clusters and cloud architecture) a common practice is to use multiple units for the same task and operate them in parallel. Our study gives a short introduction to the availability of such distributed systems discussing redundancy methods. Markov chain modeling is an applicable tool for availability investigation. In this work we first introduce a simple Markov chain model then we continue with the general one. The general model is an extension of the simple one, it takes more factors into account, consequently it is more complicated but also more realistic. After creating the general model, comparison of two special cases is presented. As a main result it is shown that in the general case load sharing performs better than using spare units.