Subject: interesting project: HFSC
To: None <>
From: Miles Nordin <carton@Ivy.NET>
List: current-users
Date: 10/15/1999 20:40:41
Has anyone read about the packet scheduling theory work Hui Zhang's been
doing?  His web page mentions a NetBSD implementation they wrote for
proof-of-concept.  Actually he mentions it _before_ the FreeBSD port.
presumably we are meant to understand that Real Scientists use NetBSD.

I wonder if this thing would make a good CPU scheduler? :'

Evangelism aside, pulling this up to -current, playing with it, optimizing
it, u.s.w., might be an interesting and fun project for someone interested
in theoretical CS, as well as extremely useful to the hopefully-increasing
numbers of people who use NetBSD for their routing (what with the
excellent built-in IPv6 support and all).

At the very least, it's ``Research Done Using NetBSD,'' and TNF might want
to brag about it on the web site.  anyway, have a look:

FWIW, I read a lot of these papers when I saw the fair queueing stuff show
up in Linux.  H-PFQ (the prequel to HFSC) seemed to make all the earlier
algorithms look rather silly.  Linux (as of 2.2.1) claims fair queueing
support, but only offers other, inferior algorithms, and does not support
H-PFQ.  Consequently it cannot both enforce link sharing goals and make
RSVP guarantees--you have to choose between CSZ and CBQ.

Miles Nordin / v:1-888-857-2723 fax:+1 530 579-8680
555 Bryant Street PMB 182 / Palo Alto, CA 94301-1700 / US