scheduling algorithms in 2.1.90

Raul Miller (rdm@test.legislate.com)
Mon, 23 Mar 1998 12:54:38 -0500


I was very pleased to see all the scheduling algorithms in net/sched/,
as I've been wanting to try out some of the scheduling algorithms I've
been hearing about. However, I am a little bemused by this comment from
Documentation/Configure.help:

###
### what user level programs are needed to administrate these packet
### schedulers?
###

Indeed.

Looking at sch_generic.c, I see a routine named dev_set_scheduler
but unless I've missed something, there's nothing in the kernel source
to pass a Qdisc into this thing.

Has anyone been thinking about the interface for this system? I presume
there's going to be a /proc/sys/net/*/conf/*/sched which holds a token
identifying one of these algorithms? And if the token is cbq, then there
will be a /proc/sys/net/*/conf/*/cbq/ with a directory for each class?
I don't know what would uniquely identify the class -- perhaps just the
kernel's address for the thing?

I can live without cbq, and I can live with default parameters on
red and tbf, but I'd like to be able to switch between scheduling
algorithms relatively quickly (hopefully, without having to boot up with
a specially patched kernel for each scheduler), at least to test these
things.

Is anyone working on this? Is there some cannonical location for
such work?

Thanks,

-- 
Raul

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu