Analysis of revenue improvements with runtime adaptation of service composition based on conditional request retries

In this paper we consider the runtime service adaptation mechanism for service compositions that is based on conditional retries. A single retry may be issued while a concrete service within composition is executed. This retry could either invoke the same concrete service or a functionally equivalent service implementing the same task. We determine the optimal moments to terminate the current request and replicate it. The calculation of these moments for each task within the workflow is based on different QoS parameters from Service Level Agreements, like services’ response–time distributions and cost–relating parameters. The calculations are performed taking into account the remaining actual time–to–deadline, and the benefit of conditional retry mechanism is illustrated by simulations. We further discuss the impact of costs and response–time distributions’ parameters to the solution at hand.