Hacker News new | past | comments | ask | show | jobs | submit login

I am not very familiar with OSes and things at the kernel level, can anyone answer the question in the comments of the post?

"How did you measure the time spent in each section (HW, kernel, app)? how did you get such granularity?"




From the Arrakis paper: "To analyze the sources of overhead, we record timestamps at various stages of kernel and user-space processing." You can probably implement this with something like perf dynamic tracing: http://www.brendangregg.com/perf.html#DynamicTracing




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: