C. Hernández-López and F. Pérez-Costoya (Spain)
Multithread, Object Oriented, Checkpointing, TransientSoftware Failures.
Checkpointing and rollback recovery have been largely used as a solution against transient software failures, how ever most research efforts on this topic have been focused on single threaded processes and little work has been done in the area of checkpointing for multithreaded applications. Besides, current checkpointingtools do not take in conside ration whether applications have an object oriented nature or not, failing to take advantage of some important charac teristics of this programmingmodel like object serialization and reflection. In this paper we propose a model to analize the execution of multithreaded object oriented programs, show how checkpointing can be applied to this model and used for the recovery of transient software failures; we also present a recovery algorithm which ensures to reach a con sistent program state after a transient failure independently of the checkpointing policy used.
Important Links:
Go Back
