MapReduce Based Skyline Services Selection for QoS-aware Composition

Abstract

Service selection is an important issue of Service-oriented computing (SOC), which is a fundamental step to the composition of complex and large-grained services from single-function components. Skyline operation is recently adopted to select candidate services for composition, as skyline services have better QoS. However, the fast increasing web services, multiple quality attributes to be considered, and dynamic service environment pose a big challenge to skyline service selection. In this paper, we present a parallel skyline service selection method to improve the efficiency by upgrading the MapReduce paradigm. In particular, an angle-based data space partitioning approach is employed in our MapReduce based skyline service selection. To handle the dynamic nature of service environment, we employ Paper-Tape (PT) Model which is used to rapidly locate varying services, and present a dynamic skyline service selection algorithm based on PT model. By experimenting over 10,000 web services along 10 quality attributes, we demonstrate the efficiency of our proposed methods.