Date: 30 May 2002 17:24:07 -0700 From: Jon Noack <noackjr@compgeek.com> To: freebsd-ipfw@freebsd.org Subject: Update: peer-to-peer asymmetric simulation Message-ID: <20020531002407.20594.cpmta@c015.snv.cp.net>
next in thread | raw e-mail | index | archive | help
I applied the patch from my original email and seem to have gotten it to work (after setting net.inet.ip.fw.one_pass=0). Example rules:
For each <X> from 101 to 108
pipe N config bw 32Kbit/s delay 70ms queue 4Kbytes
add pipe N ip from 192.168.1.<X> to any
pipe N config bw 48Kbit/s queue 4Kbytes
add pipe N ip from any to 192.168.1.<X>
There are 2 pipes per host for a total of 16 pipes.
Limiting bandwidth down to 32Kbit/s seems to introduce an innate 20ms delay while 48Kbit/s introduces a delay of around 10ms (stayed constant at HZ=1000 and HZ=10000). The above rules give me a consistent 200ms ping time between hosts {(70ms [+ 20ms innate] [+ 10ms innate]) * 2}. Limiting bandwith to 128Kbit/s up and 1Mbit/s down introduces an innate delay of 10ms (for a full ping) about 2/3 of the time (average 7ms added over 60 pings). This is due to queuing delay, correct (or is this uncertain due to the hack to bridge.c)? Is there anything I can do to reduce this?
Finally (I ask again), why is this not enabled by default? What am I risking with this?
Thanks again,
Jon Noack
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-ipfw" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020531002407.20594.cpmta>