A fault tolerant platform of web services based on service composition

The fault tolerance in distributed systems is strongly related to the replication notion, which can be either passive or active. To ensure the recovery of an application when a fault occurs, fault analysis and prevention techniques are used, so the faulty service is replaced by another one. Fault tolerance is ensuring the correct behaviour of the application. We suggest in this paper to design a fault tolerant framework along the invocation of a service. This framework is based on replacing the faulty service by another equivalent service, or a composition of services. The approach consists on collecting, on a particular node called "supervisor", all the features required to protect the execution of a distributed application when a fault happens in one of its running services. The faults that we try to tolerate in our platform are mainly hardware faults, link breakdowns and faults due to the random mobility of wireless ad hoc network nodes.