Quantifying the adaptability of workflow-based service compositions

Abstract

Service composition, which is the integration of heterogeneous Web-accessible services, has introduced an open phenomenon for creating software systems that are deployed in an unpredictable execution context, where changes in the requirements, user preferences, and component services frequently occur. The consideration of adaptability in designing such systems is necessary in this unstable and volatile environment. There have been intensive efforts on solutions to make service compositions adaptable, but little attention is given to the evaluation of such adaptability. This paper proposes a set of metrics to quantify the adaptability of service compositions that are specified from the adaptive WS-BPEL-based frameworks. The metrics are based on two adaptability dimensions: the structure and binding variabilities. Structure variability, enables runtime changes to the composition workflow while binding variability, allows dynamic binding to concrete partner services. We evaluate the metrics through a case study, utilising variants of a travel booking process specified from different adaptive WS-BPEL-based frameworks found in the literature. The metrics are applicable to a wide spectrum of adaptive service compositions and their frameworks. A support tool is also developed to automate metrics computation. Through the metrics, adaptability of service compositions and their frameworks can be assessed and compared. This would facilitate design decisions in building flexible and robust services where adaptability is a first class concern.