Ceaser: Exploring the Monitoring Rules in the Time Dimension

Abstract

In cloud environment, most failures would not strike at a sudden moment but evolve from a minor anomaly. During the evolving procedure, it may generate a series of events. But users can't specify exactly the event sequences of all the failures due the lack of time or knowledge. Automatically exploring the monitoring rules in time dimension can help users refine their rules and timely discover the failure. Based on these facts, we introduce our prototype: Ceaser. It analyzes the raw rules that user specified based on history event sequence, and try to refine them in the time dimension. Unlike pattern mining or event summarization, the rule refinement process is handled by a rule generator Ceaser-S and an extensible CEP engine Ceaser-E. The cooperation of these two can not only discover the event sequence pattern which one event happens after another, they can also explore the monitoring rules including the negative pattern or Kleeneclosure pattern and other user defined pattern.