Root Cause Analysis Using Sequence Alignment and Latent Semantic Indexing

Automatic identification of software faults has enormous practical significance. This requires characterizing program execution behavior. Equally important is the aspect of diagnosing (finding root-cause of) faults encountered. In this article, we address the problem of identifying the root cause of failure from the test sequences that caused failure. Taking analogies from biological sequence alignment and information retrieval domains we propose two approaches for finding the root cause of failure. The first approach is to align all the test sequences pertaining to a fault and identifying the common pattern among these sequences. The other approach is based on an information retrieval technique viz., the latent semantic indexing (LSI). Our experiments and analysis showed that the sequence alignment based approach has the potential to aid significantly in identifying the root cause of failure. The LSI based approach automatically clusters the test sequences based on their functionality, which assists in determining the different manifestations of a fault.