Date: Fri, 30 Jan 2004 15:43:40 -0800 From: Mike Hunter <mhunter@ack.Berkeley.EDU> To: Eli Dart <dart@nersc.gov> Cc: freebsd-net@freebsd.org Subject: Re: Assymetric results from iperf across gigabit link (long) Message-ID: <20040130234340.GA3288@ack.Berkeley.EDU> In-Reply-To: <20040130220000.81179F8EB@gemini.nersc.gov> References: <20040130213935.GB25194@ack.Berkeley.EDU> <20040130220000.81179F8EB@gemini.nersc.gov>
next in thread | previous in thread | raw e-mail | index | archive | help
On Jan 30, "Eli Dart" wrote: > In reply to Mike Hunter <mhunter@ack.Berkeley.EDU> : > > I switched the two pieces of hardware, and the photons still prefer going > > uphill, so maybe there's a problem with the fiber after all. I'd still > > appreciate any hints on what to ask freebsd to help me figure it out. > > (Yes, we have real fiber test gear, this is more of an experiment.) > > Have you looked at interface error counts? (netstat -inbd) If you > have dirty fiber, you could just be missing/corrupting the odd packet > here and there, which would be enough to impact a high-bandwidth TCP > flow. % netstat -inbd bge0 Name Mtu Network Address Ipkts Ierrs Ibytes Opkts Oerrs Obytes Coll Drop bge0 1500 <Link#2> 00:30:1b:b1:08:b6 1321030 0 153151700 1971658 0 2929764081 0 0 bge0 1500 fe80:2::230:1 fe80:2::230:1bff: 0 - 0 3 - 216 - - bge0 1500 169.229.254.1 169.229.254.140 1320970 - 134650330 1971653 - 2902160837 - - [snip] > the other thing to check is if a UDP cannon has the same bandwidth > asymmetry, since the UDP iperf tests don't have TCP's timidity (if > memory serves me).... I'll give that a shot...thanks. Although my flood pings should have served the same purpose (given that I fixed the icmplimit problem), right? Thanks, Mike
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040130234340.GA3288>