The effect of clustering in client-caching architectures

Computer systems deployed in contemporary industrial and production environments typically involve a large number of workers who work simultaneously to draft new components and/or trace provided services. Efficient database support for such systems is crucial. The handling of high volume data among various sites of a network based infrastructure poses new challenges as it can be only carried out by aggregates of data servers and powerful workstations/PCs (clients). Operating in such environments, client/server databases (C/S DBSs) (M. Carey et al., 1991) have shown reduced response times for client transactions over their centralized counterparts. Previous work has indicated, however, that when the number of clients attached per server becomes large, C/S DBSs fail to guarantee satisfactory performance rates. This is known as the scalability problem. Client object caching is used as a mechanism to improve system scalability by easing resource contention at the server sites. Design and manufacturing databases display different operational characteristics from their conventional counterparts as they routinely support multiple and mostly independent projects. In such settings, a group of cooperating designers or workers achieve a complex task by closely interacting among themselves and dynamically sharing design data. We exploit the aforementioned feature of design and manufacturing databases, and propose a cluster based caching configuration. The cluster based configuration groups sites according to the similarity of their access patterns. This results in improved client performance rates and offers better scalability