Profiling Node Conditions of Distributed System with Sequential Pattern Mining

Recently, with wide-spread of distributed systems, distributed monitoring systems are needed to mange such systems. However, since monitoring architecture of distributed system faces a huge amount of log data which come from local computing nodes, information aggregation is fundamental scheme for monitoring distributed system. In this paper, we preset a novel approach for extracting computing node-condition profiles by using sequential pattern mining, which is one of data mining techniques. Extracted computing node condition profiles represent node condition patterns which are occurred in many computing nodes frequently. Thus, extracted profiles enable summarized distributed system conditions to be small sized and easy-understandable information.