Parallelization Framework of Root Cause Analysis Based on Global Cache and Global Lock

The end-to-end resolution time cost of system failure can be split into the time of root cause analysis (RCA) and the problem resolution. In complex enterprise applications and large-scale cluster fault diagnosis activities, the RCA has a long duration due to the scale and complexity of the business, directly causing miss SLAs. To enhance the efficiency of problem discovery, an asynchronous parallel framework is presented in this paper, which establishes concurrent detection or orchestration jobs based on the dependency graph of various types of resources and parallel analysis jobs based on the parallel access capability of a specific resource bearer. Since there is one or two root causes of most anomalies, the framework's message notification mechanism was also designed as a pattern that gives reports immediately after discovery and as a full-path coverage pattern to achieve early reporting of root cause analysis and early intervention by human. Thus, the root cause analysis duration is reduced and the end-to-end solution time of system failure is shortened.