A QoS – based framework for Efficient Web Servers

This research work contributes to research aiming to improve the QoS of commercial web servers. The QoS is defined in terms of server throughput and service time as perceived by users. The work is a two-part approach. The first part is concerned with load balancing of web traffic, while the second tackles admission control for the server itself. The contribution is as follows: we elaborated the session-based admission control by introducing the server cluster utilization as an internal control parameter for the number of accepted sessions, besides the traffic burstiness as an external parameter and studying in detail the variability of the number of the completed sessions with system parameters. We could prove the usefulness of the emerging trend of combining load balancing with admission control for commercial web servers. For load balancing we used content-aware load dispatching policy: Distributed Workload-Aware Request Distribution (DWARD) and for admission control we have used adaptive time slot scheduling predictive session based admission control, where the length of the execution intervals of the algorithm depend on the burstiness arriving to the system. Via a simulation model, we could show that a web server cluster augmented with load balancing and session-based admission control performs better in terms of QoS than a server without these controls. We could also show how the throughput and service times of the server cluster are tradeoffSupervisors Prof. Salma Ghoniem ; Dr. Sahar Haga