A Fluctuation-Aware Approach for Predictive Web Service Composition
Xiaoning Sun, Jiangchuan Chen, Yunni Xia, Qiang He, Yuandou Wang, Xin Luo, Rongqing Zhang, Wuhong Han, Quanwang Wu · 2018
Service-Oriented paradigm enables the composition of loosely coupled services provided with varying nonfunctional properties in terms of Quality-of-Service (QoS). Given a composition template, finding the set of appropriate services that guarantees users' functional requirements under given QoS constraints has been widely acknowledged to be a challenge. The problem of optimal service selection and composition is usually addressed by considering timeinvariant, stochastic, or bounded QoS of candidate services in their QoS analysis models and scheduling algorithms. Our proposed work, instead, considers time-varying QoS of services and leverages its run-time fluctuations for generating dynamic and predictive service composition schedules, by using an ARIMA-based time-series prediction model and genetic algorithms. Extensive case studies based on multiple service composition templates and real-world QoS datasets clearly suggest that our proposed method outperform traditional ones in terms of response time and throughput.