A replica distribution technique for cluster server system

Abstract

In recent years, the rapid growth in the net traffics and on-line processing load has been significant, particularly in the area of database servers. Many factors are responsible for this growth, such as increasing demand to make information more available or accessible as well as increased information storage both in terms of size and volume. One solution is to resort to a cluster-based server. This however could indirectly increase the use of online applications and thus the demand for non-stop cluster servers. Hence, in designing a clustcr server, the main focus should be on providing highly reliable and good quality service servers. This is important since both hardware and software are failureconstraint. Thus, in preparing a reliable service, the cluster server must be able to maintain its reliability on some data replicas in addition to ensuring that all replicas have one permanent copy of the replica. This thesis proposed a replication technique known as "Neighbor Replica Distributed Technique", which is introduced to achieve a highly reliable service for a cluster server under four-node system.