Comparative study of genetic and blackboard algorithms for solving QoS-aware service selection problems

In distributed, heterogeneous systems, where several deployments of a specific service exist, it is a crucial task to select and combine concrete deployments to build a service chain that can be an arbitrary workflow or a query execution plan. In order to decide between deployments with identical functionality, non-functional properties, also called Quality of Service (QoS) properties, are taken into account. Service selection processes are designed to reach certain objectives while meeting defined global constraints. These optimization problems are known as QoS-aware service selection problems and are NP-hard in the strong sense. Several heuristic approaches have been proposed to address these QoS-related issues. In this paper, two heuristics a genetic and a blackboard algorithm - motivated by specific application scenarios are described and compared in terms of performance, scalability and applicability. Based on the application characteristics, the advantages of one over the other algorithm are discussed, and a common setup is carried out to perform a competitive comparison. The results indicate that under a certain threshold the blackboard outperforms the genetic algorithm, whereas the latter one shows a better scalability with increased number of deployments.