A response time-driven server selection substrate for application replica hosting systems

This paper presents a server selection approach that relies on a response time estimator and a scalable monitoring infrastructure that provides the utilization measures for replica server hosts. The big challenge rises from the difficulty to capture, both accurately and efficiently, the dynamic state of the hosts and network resources in a large-scale replication context. The monitoring infrastructure can be tuned such as to minimize the network traffic while maintaining the estimation accuracy high