Root-cause analysis of SAN performance problems: an I/O path affine search approach

摘要

We present a novel algorithm, called IPASS, for root cause analysis of performance problems in storage area networks (SANs). The algorithm uses configuration information available in a typical SAN to construct I/O paths, that connect between consumers and providers of the storage resources. When a performance problem is reported for a storage consumer in the SAN, IPASS uses the configuration information in an on-line manner to construct an I/O path for this consumer. As the path construction advances, IPASS performs an informed search for the root cause of the problem. The underlying rationale is that if the performance problem registered at the storage consumer is indeed related to the SAN itself, the root causes of the problem are more likely to be found on the relevant I/O paths within the SAN. We evaluate the performance of IPASS analytically and empirically, comparing it to known, informed and uninformed search algorithms. Our simulations suggest that IPASS scales 7 to 10 times better than the reference algorithms. Although our primary target domain is SAN, IPASS is a generic algorithm. Therefore, we believe that IPASS can be efficiently used as a building block for performance management solutions in other contexts as well.