Presence-pattern aware service selection and composition in a smart space

Service composition provides supports for automatic construction of required services on the fly from component services provided by different providers. In a smart space, as new providers may come in and existing ones may leave from time to time, the collection of available component services may change dynamically, resulting in broken compositions. Automatic reselection or recomposition mechanisms can be applied to fix broken compositions. However, they may introduce extra efforts and time and there are applications requiring (composed)services to be continuously available (at least) during a period of time, otherwise the applications would break down. Both the situations affect users' experience. If we could know how long each component service would be continuously available, we could optimize the composition process by reducing the frequency of broken compositions. In this paper we propose a scheme to achieve the goal. It utilizes presence-patterns of providers to estimate how long a component service may continuously available and it always selects the most continuously available composition that matches requirements. Simulations are carried out to show how and how well the presence-patterns based scheme work.