GBTM: Graph based troubleshooting method for handling customer cases using storage system log

Abstract

Present day computing environments consist of different bits of hardware and software that are associated with each other in a complex way. Hence, in case of failures of such system, it is very difficult to detect the exact module which has caused the problem. In such a situation, an automated technique which can pin down to (at least) a set of modules that may be responsible for the failure would be very useful for support engineers. This paper makes an important step towards that direction. We propose a graph based troubleshooting methodology exploring storage system logs (EMS) of around 4500 customer cases to troubleshoot customer problems. We provide a ranked list of modules to the support engineers which can significantly narrow down the troubleshooting process for around 95% cases with only 10% false positive rate whereas the competing baseline MonitorRank covers only 74% cases with 23% false positive rate.