graph database for trace analysis
Fri Sep 17 15:29:56 CEST 2010
What do you folks think of using a graph database for trace analysis?
Each trace converts into several graph nodes with attributes like source process, destination process, timestamp, etc.
The benefit I see is being able to construct queries that time requests as they go through the system. Each request consists of several messages sent between different processes and results in a response.
It seems near impossible to do this kind of analysis using available profiling tools. Interactive querying and analysis is definitely impossible at the moment.
More information about the erlang-questions