Date: Wed, 06 May 1998 05:14:52 -0700 From: David Greenman <dg@root.com> To: Reinier Bezuidenhout <rbezuide@oskar.nanoteq.co.za> Cc: freebsd-hackers@FreeBSD.ORG Subject: Re: Weird UDP T/P on fxp0/de0 - 2.2.6-RELEASE Message-ID: <199805061214.FAA13851@implode.root.com> In-Reply-To: Your message of "Sat, 06 May 1998 12:16:04 %2B0200." <199805061017.MAA01011@oskar.nanoteq.co.za>
next in thread | previous in thread | raw e-mail | index | archive | help
>I am using ttcp on both sides to test data transfers with >diffirent packet sizes. I ran into some really STRANGE numbers >during these test. I have basically two setups on each side I haven't looked at newer versions of ttcp, but the older version, on the transmitter side, would do a backoff (sleep) whenever it got ENOBUFS from the kernel. The sleep is tuned for 10Mbps ethernet, so the result is that the transmitter runs with wild swings filling and draining the kernel's output queue. One solution is to reduce the timeout substantially or eliminate it. This will cause ttcp to waste lots of CPU while it is running (constantly getting ENOBUFS from the kernel), but has the desired positive effect of keeping the output queue full. When you make this change, you get results like this: [implode:dg] ttcp -u -p9 -n8192 -t core ttcp-t: buflen=8192, nbuf=8192, align=16384/+0, port=9 udp -> core ttcp-t: socket ttcp-t: 67108864 bytes in 5.75 real seconds = 11393.10 KB/sec +++ ttcp-t: 15087 I/O calls, msec/call = 0.39, calls/sec = 2622.80 ttcp-t: 0.0user 5.3sys 0:05real 94% 141i+315d 168maxrss 0+2pf 0+59csw 0.054u 5.386s 0:05.75 94.4% 141+315k 0+0io 0pf+0w ...that's on a Pentium/133, so the results may be slightly higher on a faster machine. In any case, FreeBSD w/fxp driver has no trouble keeping the fast ethernet saturated. -DG David Greenman Co-founder/Principal Architect, The FreeBSD Project 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?199805061214.FAA13851>