Following scenario: 1. Leverage application dependency data (assuming w/ a plug-able interface so it could deal with different external source like opentracing etc) 2. Correlation alerts in the dependency call graph for upstream/downstream applications. Correlation could be done with time window, etc. 3. From a upstream application alert, trace down to find list of possible responsible root causes (alert), and score them with confidence level. 4. Serve as an API.
Following scenario: