This is a very interesting article on how you folks at Netflix have gone about identifying the culprit API/Package resulting in high CPU utilization. The initial part of the article is very informative but the crux of the article which deals with identifying the package is a little vague.
Could you explain how :
- You reprocessed the stack traces to collapse the repetitive framework calls
- How did you identify the “potential expensive” package?
- Filtered out the frames in each call stack until you reached a method calling this package.