L.A. Sanabria, B. Soh, T.S. Dillon, and L. Chang (Australia)
Web Reliability, Genetic Algorithm
This paper presents a method to optimise the reliability design of a Web server. The server is modelled as a fault tolerant system made up of series/parallel elements. The problem is to find the most reliable configuration ie. to find unit performance and redundancy allocation for each stage of the server subject to a cost constraint. A central characteristic of the problem is that the failure rate of the server components is modelled by random variables in order to capture uncertainties in their values and allow design engineers to consider a wide range of processing units. The problem then, becomes one of constrained stochastic optimisation. Such problem is difficult to solve using conventional optimisation methods, a relatively knew technique is optimisation by Genetic Algorithms. This technique can tackle the problem in a simple yet effective way and obtain acceptable results in a reasonable time.
Important Links:
Go Back
