Hybrid based QoS-aware selection of web services compositions

Abstract

The composition of web services performs a complicated service using its interacting single services, each with a few QoS (quality of service) features. Many compositions exist that perform the same complex service but with different QoS features whose near-optimal ones should be selected. Before the selection, the compositions’ QoSs based on summarizing the composition patterns should be calculated. Many methods have been presented for the compositions’ QoSs calculation and near-optimal composition selection. However, in summarizing they considered: (1) not all kinds of patterns in compositions, (2) not formal-based pattern definitions, and (3) not high-performance method. To address these issues and to present an effective method over related studies for composition selection, we suggest a Pattern-based and multi-objective architecture (PBMOA). We applied our approach to four real-world web-based service-oriented systems using a dataset of 2507 candidates and evalu ated results. We then compared the solutions derived from eight related studies with the solutions derived from our architecture in terms of features and a few performance metrics. Additionally, we demonstrated the erality of the results in terms of features by utilizing statistical tests. Our solutions (near-optimal compositions) outperform related studies’ solutions by an average of 79 percent, according to the coverage ratio performance indicator.