S4: Scalable Statistical Service Selection

Description

In service computing, services should discover, bind and communicate with one another. Finding good partner services is curial. Selecting such services can be divided into three tasks, namely applying selection criteria, service matchmaking, and service ranking. Many existing techniques assume that selection criteria applicable to atomic services are also applicable to composite services. However, applying selection criteria to composite service may require to compose the service beforehand, which implies that the service composition problem should have been solved. That poses a dilemma. Furthermore, services may evolve, making the ranking task tedious and expensive to evaluate and re-evaluate the quality of component services so that organizations may reorganize their service compositions. Large-scale evaluations from different service consumers further put huge burdens on individual service providers. Scalable service selection approach is curial to be developed. In this project, the researchers propose to explore the service selection problem functionally, and gradually integrate with the non-functional metrics. They propose to attack the problem from a statistical perspective, and evaluate those techniques through controlled experiments.