A survey of methods for system-level fault diagnosis

摘要: With the increasing need for efficient means of automatic fault diagnosis in large distributed computing systems, system-level has been a fertile research area last few years. There are two types methods: classical and adaptive. The methods select set tests, find results all these then process to identify faulty units. adaptive first just one fault-free unit use it Each diagnostic can assume so called symmetric or asymmetric test invalidation. former states that tests performed by good units always give correct results, while produce any results. latter fails test, even if influence result faulty. We survey number each under both invalidation assumptions. is considered context certain model (such as, e.g., Boolean n-cube where processors represented nodes links edges graph). Finally, comparison shows faster (require fewer steps diagnosis) but less (may misdiagnose more as faulty) than methods.