摘要: Applicative systems are promising candidates for achieving high performance computing through aggregation of processors. This paper studies the fault recovery problems in a class applicative systems. The concept functional checkpointing is proposed as nucleus distributed mechanism. entails incrementally building resilient structure evaluation an program proceeds. A simple rollback algorithm suggested to regenerate corrupted by redoing most effective checkpoints. Another algorithm, which attempts recover intermediate results, also presented. parent faulty task reproduces twin failed task. regen~rated inherits all offspring so that partial results can be salvaged.