Service Selection for Composition with QoS Correlations

TLDR

QoS is a key criterion in service selection, yet most existing methods ignore correlations between services, which can degrade performance. This study investigates how to select candidate services for composite services while accounting for QoS correlations to achieve optimal QoS. We introduce the correlation‑aware service pruning (CASP) method, which considers all potential services in a composite and prunes those that cannot be optimal. Experiments demonstrate that CASP effectively handles complex service correlations and markedly improves the QoS of the resulting composite services.

Abstract

QoS as an important criterion has attracted more and more attention in the service selection process. Various QoS-aware service selection methods have been proposed in recent years. However, few of them take into account of the QoS correlations between services, causing several performance issues. QoS correlations can be defined as that some QoS attributes of a service are not only dependent on the service itself but are also correlated to other services. Since such correlations will affect QoS values, it is important to study how to select appropriate candidate services while taking into account of QoS correlations when generating composite services with optimal QoS values. To this end, we propose a novel method of service selection, called the correlation-aware service pruning (CASP) method. It manages QoS correlations by accounting for all services that may be integrated into optimal composite services and prunes services that are not the optimal candidate services. Our experiments show that this method can manage complicated correlations between services and significantly improve the QoS values of the generated composite services.