Optimizing Service Systems Based on Application-Level QoS

• Published in: IEEE transactions on services computing Vol. 2; no. 2; pp. 108 - 121
• Main Authors: Liang, Qianhui, Wu, Xindong, Lau, Hoong Chuin
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01-04-2009; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; composite services; Data mining; Decision support systems; Dynamical systems; Dynamics; Internet service providers; Mathematical models; Maximization; negotiation; Negotiations; Optimization; Optimization of services systems; Probability density function; Quality of service; quality of services; robust; Robustness; service selection; Studies; system utility; Utilities; Web services
• ISSN: 1939-1374; 1939-1374; 2372-0204
• DOI: 10.1109/TSC.2009.13

Making software systems service-oriented is becoming the practice, and an increasingly large number of service systems play important roles in today's business and industry. Currently, not enough attention has been paid to the issue of optimization of service systems. In this paper, we argue that the key elements to be considered in optimizing service systems are robustness, system orientation, and being dynamic and transparent. We present our solution to optimizing service systems based on application-level QoS management. Our solution incorporates three capabilities, i.e., 1) the ability to cater to the varying rigidities on Web service QoS in distinct application domains and of various users in a robust and heuristic manner, 2) the ability to formulate the overall system utility of a service system perceived by a particular system end user and to suggest its maximization using a utility model incorporated into a three-dimensional weighting scheme, and 3) the ability to dynamically achieve a higher perceived system utility of a service system via transparent negotiations. The calculation of the system utility encompasses a negotiation algorithm and a robust search algorithm for selecting heuristically best Web services. The effectiveness of the proposed algorithms and our solution is demonstrated by simulation experiments and our demo deployment, SSO.