Model-driven methodology for building QoS-optimised Web service compositions

As the number of available Web services increases there is a growing demand to realise complex business processes by combining and reusing available basic Web services. For each of the needed basic Web services there may be many candidate services available from different vendors and with different Quality of Service (QoS) values. This paper proposes a model-driven methodology for building new Web service compositions that are QoS-optimised. We investigate if UML is suitable for modelling the QoS aspects and we explain how transformations can be used to automate most parts of the methodology. Some of these transformations are already implemented in previous work. As part of the methodology we present a control flow pattern approach that optimises the QoS values of the composition given user-defined requirements and preferences. The pattern approach is compared towards the local and global approaches identified by other authors. Experiments are carried out within our simulation tool that simulates the selection of services and aggregation of QoS in a given composition. The test results are used to identify recommendations of when to use the different approaches based on measurements of computation time of the optimisation component and achieved QoS values for the new composition. The methodology is explained by relating it to a gas dispersion emergency case.