Re: new scheduler

[vhandel@chem.vu.nl (Ramon van Handel)]

>Hi,
>
>I am currently planning a new scheduler for Linux
>and I was wondering what features from the following
>list are most important to you (the implementation
>has to remain leight-weight, so we will have to
>choose).
>
>- better realtime scheduling

-->  yay !

>- better throughput (on SMP systems) by less memory contention
>  and less unneeded context switches

-->  yay !

>- better interactive response under load

-->  yay !

>- load-isolation between users (if user A and B run
>  processes and A has 10 and B only 1, it would be
>  fairer if both users each got 50% of the CPU)

Hmmm....  I prefer it the way it is.  Say I am doing a few big calculations
(yes, I do that once in a while ;)) and some other user is playing xbill, I
don't see why his xbill and my calculations would each get 50% CPU.

>- simplicity, less overhead

Well, that's always a good thing.

>- larger difference between niced and non-niced CPU
>  usage

Hmmm... I thought the current system worked OK, but I could be wrong.

Anyway, you didn't put one thing on your list that I'd like to see:  could
you document the code a bit more extensively than it's done now ?  That's
always a good thing, and linux hackers rather tend to neglect it.

>For the technically minded: the scheduler will be a
>variation on the scheduler mechanism we've got right
>now, as it's quite leightweight and functional and I
>don't see any problems that can't be fixed within
>the framework of the current system

Could you detail on your planned changes ?

Ramon


---
Ramon van Handel   <vhandel@chem.vu.nl>
Chemistry Student, OS Programmer and all-round Weirdo
The ant has made himself illustrious / Through constant industry industrious.
So what? Would you be calm and placid / If you were full of formic acid?
(Ogden Nash)


-
Linux-future: thinking about the future of the Linux kernel
Archive:      http://humbolt.nl.linux.org/lists/
Wish list:    http://users.ox.ac.uk/~mert0236/linux-future.html