Application performance is directly related to the response time of the system. Clients issue requests to get information. Such requests consume differing amounts of central processing unit (CPU) processing time and bandwidth, and generate different traffic profiles. Usually, services are designed based on peak traffic and request rates. There are different architectures to provide the information to users. In this chapter, the authors describe the Web caching approach, and propose the grid caching architecture.

The authors consider the case of applications that require a reduced response time, because this is the case with most commercial applications. In this chapter, they propose using the grid caching architecture to minimize this time. The authors analyze different parameters that influence the response time in the Web caching approach: CPU cycles for data encryption, key generation, creation of data on the fly, and so on. Moreover, the authors analyze the Web caching architecture, which is strongly dependent on the consistency model imposed in the system, and which requires a load balancing scheme to distribute requests among the servers where documents are replicated.

The authors present a grid caching architecture where a workload modeler and predictor is a key component. This element is used to detect overload conditions, and to enable the servers on the grid. The chapter also presents design ideas for the proposal.

The chapter introduces an interesting vision of the grid infrastructure, one that allows it to be applied to applications other than the typical.