Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 13 Nov 2002 23:48:31 -0500
From:      Alex Newman <dolemite@psiu.dyndns.org>
To:        freebsd-hackers@freebsd.org
Subject:   tcp randomness
Message-ID:  <20021114044831.GA17762@ultraalex.psiu.dyndns.org>

next in thread | raw e-mail | index | archive | help
So i am doing an experiment which adds  a random delay to tcp flows to
achieve alot of the same stuff you get with RED and was planning on
testing it with a divert socket. The problem is sometimes this involves
adding a delay of say 2 ms for instance. Is this even possible on intel
hardware. I know that thier is almost no way to get ethernet to be that
accurate (I don't know about device polling maybe that will do it) but i
think adding this delay to outgoing packets before it hits the card
should add the entropy i am looking for.
I have a couple of questions:

1) How much delay will i encour by passing the packet out through ipfw
to a divert socket?

2) Does changing the HZ rate in the kernel actually cause ints to happen
faster? Will this help me?

-- 
Alex Newman			Passant, comme toi j'ai passe`. 		
dolemite@wuli.nu		Le fleuve est revenu se perdre dans sa source.
www.wuli.nu/users/dolemite	Fais silence; assieds-toi sur ce marbre brise`.
				Pose un instant le poids qui fatigue ta course;
				J'eus de me^me un fardeau qu'ici j'ai de`pose`. 

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20021114044831.GA17762>