Date: Thu, 17 Apr 2014 10:09:46 +0100 From: Karl Pielorz <kpielorz_lst@tdx.co.uk> To: =?UTF-8?Q?Stig_Inge_Lea_Bj=C3=B8rnsen?= <stiginge@pvv.org>, freebsd-infiniband@freebsd.org Subject: Re: Mellanox MT25418 performance IPOIB? Message-ID: <830453C43410411669F120CC@study64.tdx.co.uk> In-Reply-To: <lilt1l$15v$1@ger.gmane.org> References: <DF61C9D7E8748BF2824DDFED@study64.tdx.co.uk> <lilt1l$15v$1@ger.gmane.org>
next in thread | previous in thread | raw e-mail | index | archive | help
--On 16 April 2014 14:28:37 +0200 Stig Inge Lea Bj=C3=B8rnsen=20 <stiginge@pvv.org> wrote: > You could try to use Infiniband in datagram mode, recompile without the > IPOIB_CM option. Ok, same boxes - I just removed IPOIB_CM out of the kernel config and=20 recompiled / installed it. iperf now consistently returns: [ 3] local 10.10.10.2 port 37715 connected with 10.10.10.1 port 5001 [ ID] Interval Transfer Bandwidth [ 3] 0.0-10.0 sec 8.71 GBytes 7.46 Gbits/sec > The maximum MTU will then be 4092, but it might default to 2044. I run > OpenSM on both machines and set the MTU explicitly: Yes, mtu defaulted to 2044 - raising it to 4092 actually causes a loss of=20 .5-1Gbit with iperf (obviously not 'real world') - but at leas I can now=20 try and use the stuff with real world (now it's returning more than a few=20 hundred kilobytes a second!). > The bandwidth reported by iperf is ~7.5 Gbits/sec in both directions. Same as here then. Thanks for your email! Regards, -Karl
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?830453C43410411669F120CC>