Dr.-Ing.

Optimising Quality-of-Service for the Composition of Electronic Services

December 2006, Technische Universität Berlin
Faculty IV - Electrical Engineering and Computer Science
URN: urn:nbn🇩🇪kobv:83-opus-14724
Online Edition

Abstract

Electronic services and their composition gain a growing interest from businesses that intend to implement their processes with software systems. The general characteristics of electronic services resemble the idea of process-orientation as proposed by the business process re-engineering initiative introduced in the 90s. Thus, the software industry promotes developing service compositions in order to efficiently implement business processes.

The development of service compositions involves service brokers. These brokers implement a trading functionality in order to identify the suitable services based on requirement descriptions. The trading functionality can also consider different optimisation criteria in order to optimise the resulting composition. In the proposed application scenario, numerical Quality-of-Service (QoS) characteristics usually serve as optimisation criteria. When multiple criteria have to be considered at once, an optimisation problem arises that can result in an unfeasible computational effort.

A novel approach for this problem is to apply heuristic algorithms. This approach requires a discussion, because heuristic algorithms do not guarantee to find the optimal solution. The question is how well the approximations compare with the optimal solution referring to the resulting QoS of the composition.

Based on the characteristics of the application scenario a method is developed for computing the QoS of compositions based on the QoS statements of the involved services. A QoS-based selection must use such a method in order to determine the QoS of the entire composition when selecting individual services. Based on this method, a model for the problem of QoS-based selection is defined. The model enables the understanding about the problem and it also serves as the reference for the discussion about the relations to other combinatorial problems. Moreover, the problem model is used for the explanation of the heuristic algorithms applied to the selection problem.

Based on the problem model and the relevant QoS concepts, the implementation of a simulation provides the evaluation of the performance of the heuristic algorithms. The simulation presents measures based on the resulting QoS of the composition and the computation time. The results from conducted simulation runs allow the comparison among the algorithms and with a method that always finds the optimal solution.