Performance Evaluation of Dynamic Web Service Selection Strategies in Service Oriented Architecture

In this paper, we explore the performance potential of dynamic (runtime) web service selection within the scope of Service Oriented Architecture (SOA). The web service selection is executed by a service orchestrator (request dispatcher) which is responsible to deliver composite web services at desired quality levels for the orchestrator's clients. We investigate service response times for the case where SOA state-of-the-art static web service composition is used and for the case where dynamic web service selection is applied. Modelling request scheduling at individual web services as Processor Sharing queueing systems, simulation results are presented for different runtime selection strategies in scenarios ranging from the "ideal" situation (up-to-date state information, no background traffic) to more realistic scenarios in which state information is stale and/or background traffic is present. In particular, we show the effectiveness of a selection strategy based upon the "synthesis" of Join the Shortest Queue and Round Robin strategies. For some specific scenarios we derive and validate insightful approximate formulae for the resulting response times. Our investigations quantify the performance gains that can be achieved by dynamic service selection compared to static (a-priori) service selection currently used.