Came across this article and out of curiousity ended up reading this . Note the dates, the original proposal or Proof of concept code and patch set was started 3 years ago.. Phew… i guess given the complexity involved it’s necessary, but kinda puts some perspective on the blind Windows bashing i used to do until a couple of years ago.. Very humbling..
Anyway, this part of the article triggers some more thoughts/ideas from my memory..
enable the no-tick mode whenever the workload is right for it, regardless of a request (or the lack thereof) from any specific process
Years ago, i had applied for the M.Sc in management sciences program at Indian institute of Science. Bangalore. By some fluke, i cleared their entrance test and was called for the interview. One form i had to fill out before it was to make a project/thesis proposal. I was naive, optimistic, had just read Eli Goldratt’s Flow (management book) and had just learnt a bit about linux kernel scheduling.. So ofcourse, i wrote out a proposal to replace the kernel scheduler with flow-based management. i.e: to say based on the workload requested by the currently running processes, we can schedule them.
I didn’t know any better to understand the complexities and details involved, but was mostly adviced that i should have applied to the CS department.. Anyway,now this comment tells me i need to look at what is currently being done, and motivates me to see if i can contribute, even if only by commenting and adding to the discussions.