Résumé en anglais

Failures during the execution of Transactional Composite Web Services (TCWSs) can be repaired by forward or back–ward recovery processes, according to the component WSs transactional properties. In previous works, we presented TCWS fault tolerant execution approaches relying on WSs replacement, on a compensation protocol, and on unrolling processes of Colored Petri-Nets (CPNs) to support forward and backward recovery. We represent a TCWS and its corresponding backward recovery process by CPNs. Even though these recovery processes ensure system consistency, backward recovery means that users do not get the desired answer to their queries and forward recovery could imply long waiting time for users to finally get the desired response. In this paper, we present an alternative fault tolerant approach in which, in case of failures, the unrolling process of the CPN controlling the execution of a TCWS is check–pointed and the execution flow goes on as much as it is possible. In this way, users can have partial responses as soon as they are received and can re-submit the checkpointed CPN to re-start its execution from an advanced point of execution (checkpoint). We present the checkpointing algorithm integrated to our previous work.