Date: Tue, 02 Nov 2004 11:48:19 -0800 From: Justin Bennett <justin@z-axis.com> To: FreeBSD Hackers <freebsd-hackers@freebsd.org> Subject: Re: ipfw pipes: theoretical speed vs. reality Message-ID: <4187E483.8050008@z-axis.com> In-Reply-To: <4187E38F.2070809@elischer.org> References: <4187E23C.7000900@z-axis.com> <4187E38F.2070809@elischer.org>
next in thread | previous in thread | raw e-mail | index | archive | help
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 | | | Justin Bennett wrote: | | All, | | I have recently been setting up pipes to shape bandwidth on our local | net. | | However, unless I am missing something, the TCP overhead seems quite | large. | | If I configure the following pipe: | | $IPFW pipe 1 config bw 64Kbit/s | $IPFW add 31 pipe 1 ip from 192.168.0.1/24 to any | $IPFW add 32 pipe 1 ip from any to 192.168.0.1/24 | | The run traffic from my machine (192.168.0.2) through it, I get less | than half the expected bandwidth (3.9KB/s). I tried another test with a | 512Kbit/s pipe, and got around 30KB/s. | | I know on most ATM/FR lines, you can expect about 10% overhead, but 50% | seems high. | | When I remove the pipe, my T1 comes back to life, and I can pull the | same data at 160+KB/s. | | Am I missing something? | | | |> yes you are queuing data in both directions on the same queue, thus |> serialising it.. | |> use 2 seprarate pipes. one for each direction. Thanks. I knew it was simple, just couldn't wrap my head around it. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (MingW32) iD8DBQFBh+SDlNUG+Ne1CZMRAl4WAJ9DEaiyU9N/SPKGi6WlqdHe2qkhmwCghv7m So0IGU5OchsjXYHduNPm4C8= =ZjYx -----END PGP SIGNATURE-----
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4187E483.8050008>