Dependability of the system based on structured service collaboration model

The collaboration of web services is often required to realize larger and more complex distributed applications. One of the problems that should be dealt with in the service collaboration is unpredictable failures of individual service. We present in our previous work a structure-based service collaboration model, in which several basic structures are abstracted, including star-structure, ring-structure, treestructure and graph-structure. Seen from the characteristic of each structure, both ring-structured and tree-structured systems are partial fault-tolerance. In order to promote the dependability of the system, we proposed a neighborhoodbased system maintenance mechanism, in which such additional information as neighbor set is maintained by individual service node. We also give further discussions on various approaches to failure detection, to structure reconstruction and to state stabilization, which are useful in the system design or reconfiguration.