Dynamic invocation of replicated web services

This paper presents a framework for dynamic invocation of geographically replicated Web services. The framework implements a transparent invocation process, based on different server selection policies, which aims at reducing service response time at the client side. Preliminary experimental results, obtained in a real-world Internet setting, show that, in addition to the individual capacity of each server, the service invocation process is affected mainly by client characteristics such as connection bandwidth and workload distribution throughout the day.