Web Service Selection based on Parallel Cluster Partitioning and Representative Skyline

摘要

Optimizing the composition of web services is a multi-criteria optimization problem that consists in selecting the best web services candidates from a set of services having the same functionalities but with different Quality of Service (QoS). In a large scale context, the huge number of web services leads to a great challenge: how to find the optimal web services composition while satisfying all the constraints within a reasonable execution time. Most of the solutions dealing with large scale systems propose a parallel Skyline phase performed on a partitioned data space to preselect the best web services candidates. The Global Skyline is computed after the consolidation of all the Local Skylines and, eventually the optimization algorithm is applied. However, these partitioning approaches are only based on pure geometric rules and do not classify the web services according to their real contribution to the optimal or sub-optimal solution search area. We will propose in this paper an intelligent partitioning approach using a cluster based algorithm combined with the representative Skyline.