QoS-driven global optimization of services selection supporting services flow re-planning

Abstract

In most cases, QoS-driven global optimization is a non-linear 0-1 programming. Genetic Algorithms are superior in non-linear 0-1 programming and multi-objective optimization. However, encoding methods that many Genetic Algorithms (GAs) adopts are too complex or simple to apply to services selection. A novel Tree-coding Gas (TGA) is presented for QoS-driven service selection in services composition. Tree-coding schema carries the messages of static model of service workflow, which qualifies TGA to encode and decode chromosomes automatically, and keeps the medial results for fitness computing. Tree-coding can also support the services composition flow replanning at runtime effectively. The experiment results show that TGA run faster than one-dimensional Genetic Algorithms when the optimal result is the same, furthermore the algorithm with Tree-coding is effective for re-planning. © Springer-Verlag Berlin Heidelberg 2007.