Distributed systems for efficient Web-based services

The overwhelming and rapid success of the World Wide Web is placing dramatic pressure on the network and server infrastructures. The augmented service demand, coupled with the growing importance of Web-based services, has exasperated the need for highly scalable and reliable Web infrastructures. The performance problems of the systems providing Web-based services are even worsening because of the need of client authentication and system security, the increased complexity of middleware and application software, and the high availability requirements of corporate data centers and e-commerce sites. Performance problems may arise in many network and server points during an interaction with a Web-based system, but in this chapter we consider the server-side component of this infrastructure. Increasing server capacity and providing scalable Web-based services through multiple standard machines is the currently dominating choice for Web providers because of scalability, availability, and cost reasons. We describe Web system architectures consisting of multiple nodes that are distributed over local and, mainly, geographic network locations. We focus on the internal dispatching mechanisms and algorithms for assigning or redirecting client requests. We study the patterns and predictability of end-to-end performance of Web-based services that make necessary to pass from locally to geographically distributed Web architectures.