In distributed, service-oriented systems, in which several concrete service instances need to be composed in order to respond to a request, it is important to select service deployments in an optimal and effcient way. Quality of Service attributes of deployments and network links are taken into account to decide between work ows that are identical in terms of their functionality. Several heuristic approaches have been proposed to solve the resulting QoS-aware service selection problem, known to be NP-hard. In our previous work, motivated by two concrete application scenarios, we proposed a blackboard and a genetic algorithm and com- pared them in terms of solution quality, performance and scalability. In order to seamlessly run and evaluate further approaches and parallel versions of the current algorithms in a distributed environment, a general framework for service selection optimization has been implemented using Cloud Computing resources. A performance study on sequential and parallel blackboard and genetic algorithms for solving service selection problems has been carried out in the Cloud.

Top