A novel two-phase approach for QoS-aware service composition based on history records

Abstract

Existing service composition approaches usually assume that the quality of service does not change over time. Actually, the performance of a service may fluctuate due to the dynamic environments, which may violate the end-to-end QoS constraints or degrade the QoS of a composite service. History records (or QoS records) can exhibit QoS fluctuations over the past period and reflect the actual performance of a service. Thus, a novel Two-Phase Approach for Service Composition (TPASC) based on QoS records is presented to address the problem. The proposed approach has the two phases: one is preliminary filtering, where a heuristic QoS decomposition algorithm is proposed to decompose the end-to-end QoS constraints into local constraints, and then some promising services are selected in terms of the probability that they can meet the local constraints; the other is MIP (Mixed Integer Programming)-based service selection, where QoS records associated with each promising service are reduced by K-means clustering method, and then a MIP model is proposed to rank and select services in terms of the reduced QoS records. The experimental results show TPASC can conduct service composition with high efficiency and significant cost savings in dynamic environments.