Re: [RFC] Scheduler queue implementation ...

[Alan Cox]

> So we can have simply two queues ( per CPU ), one that stores I/O bound (
> counter > K for example ) and RT tasks that is walked entirely searching
> for the better tasks, the other queue will store CPU bound tasks that are
> executed in a FIFO policy.

Oh as an aside btw - there are many real world workloads where we have a lot
of non cpu hog processes running. A lot of messaging systems have high task
switch rates but very few cpu hogs. So you still need to handle the non hogs
carefully to avoid degenerating back into Linus scheduler.