Caching techniques for parallel I/O servicing

Parallel and distributed systems architectures support parallel I/O components. Caching has been applied to distributed I/O subsystems as a standard solution to the problems of fastening data accessibility and increasing data reliability. Cache consistency mechanisms have been implemented in order to influence the cache usefulness in a positive way. This paper presents a new caching technique based on the genetic algorithm idea and examines the effect of this technique on the parallel I/O cache consistency and updating process. Cached data blocks on parallel disks are considered as a population evolving over simulated time and are updated at regular intervals towards an improved cache content. The proposed cache update scheme is compared with the LRU caching scheme which has been widely adopted. The proposed technique shows improved performance compared to conventional caching under simulation runs for various workloads.