<html xmlns:v="urn:schemas-microsoft-com:vml" xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:w="urn:schemas-microsoft-com:office:word" xmlns:m="http://schemas.microsoft.com/office/2004/12/omml" xmlns="http://www.w3.org/TR/REC-html40">
<head>
<meta http-equiv=Content-Type content="text/html; charset=iso-8859-1">
<meta name=Generator content="Microsoft Word 12 (filtered medium)">
<style>
<!--
/* Font Definitions */
@font-face
{font-family:"Cambria Math";
panose-1:2 4 5 3 5 4 6 3 2 4;}
@font-face
{font-family:Calibri;
panose-1:2 15 5 2 2 2 4 3 2 4;}
@font-face
{font-family:Tahoma;
panose-1:2 11 6 4 3 5 4 4 2 4;}
@font-face
{font-family:Consolas;
panose-1:2 11 6 9 2 2 4 3 2 4;}
/* Style Definitions */
p.MsoNormal, li.MsoNormal, div.MsoNormal
{margin:0cm;
margin-bottom:.0001pt;
font-size:12.0pt;
font-family:"Times New Roman","serif";
color:black;}
a:link, span.MsoHyperlink
{mso-style-priority:99;
color:blue;
text-decoration:underline;}
a:visited, span.MsoHyperlinkFollowed
{mso-style-priority:99;
color:purple;
text-decoration:underline;}
pre
{mso-style-priority:99;
mso-style-link:"HTML Preformatted Char";
margin:0cm;
margin-bottom:.0001pt;
font-size:10.0pt;
font-family:"Courier New";
color:black;}
span.HTMLPreformattedChar
{mso-style-name:"HTML Preformatted Char";
mso-style-priority:99;
mso-style-link:"HTML Preformatted";
font-family:Consolas;
color:black;}
span.EmailStyle19
{mso-style-type:personal-reply;
font-family:"Calibri","sans-serif";
color:#1F497D;}
.MsoChpDefault
{mso-style-type:export-only;
font-size:10.0pt;}
@page Section1
{size:612.0pt 792.0pt;
margin:72.0pt 72.0pt 72.0pt 72.0pt;}
div.Section1
{page:Section1;}
-->
</style>
<!--[if gte mso 9]><xml>
<o:shapedefaults v:ext="edit" spidmax="1026" />
</xml><![endif]--><!--[if gte mso 9]><xml>
<o:shapelayout v:ext="edit">
<o:idmap v:ext="edit" data="1" />
</o:shapelayout></xml><![endif]-->
</head>
<body bgcolor=white lang=EN-GB link=blue vlink=purple>
<div class=Section1>
<p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri","sans-serif";
color:#1F497D'>Hi Igor,<o:p></o:p></span></p>
<p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri","sans-serif";
color:#1F497D'><o:p> </o:p></span></p>
<p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri","sans-serif";
color:#1F497D'><o:p> </o:p></span></p>
<p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri","sans-serif";
color:#1F497D'>SDP performance can be measured with the Qperf utility included
with the OFED drivers/utilities. I was testing with single port 3<sup>rd</sup>
Gen (Same series as the ones you link) Mellanox DDR (20/16Gbit) cards on
Opterons 2 years ago and was getting 11Gbit/sec. The newer 4<sup>th</sup> gen
ConnectX mellanox cards will achieve 13-14Gbit/sec in DDR so if you are buying
new get ConnectX.<o:p></o:p></span></p>
<p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri","sans-serif";
color:#1F497D'><o:p> </o:p></span></p>
<p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri","sans-serif";
color:#1F497D'>For comparison in datagram mode I was getting 3.2Gbit/sec and in
connected mode with a 32Kb MTU in connected mode I was getting 7Gbit/sec. This
was 2 years ago so new version of OFED may well have improved performance.<o:p></o:p></span></p>
<p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri","sans-serif";
color:#1F497D'><o:p> </o:p></span></p>
<p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri","sans-serif";
color:#1F497D'><o:p> </o:p></span></p>
<p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri","sans-serif";
color:#1F497D'>Hope this helps,<o:p></o:p></span></p>
<p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri","sans-serif";
color:#1F497D'><o:p> </o:p></span></p>
<p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri","sans-serif";
color:#1F497D'><o:p> </o:p></span></p>
<p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri","sans-serif";
color:#1F497D'>Rob<o:p></o:p></span></p>
<p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri","sans-serif";
color:#1F497D'><o:p> </o:p></span></p>
<div>
<div style='border:none;border-top:solid #B5C4DF 1.0pt;padding:3.0pt 0cm 0cm 0cm'>
<p class=MsoNormal><b><span lang=EN-US style='font-size:10.0pt;font-family:
"Tahoma","sans-serif";color:windowtext'>From:</span></b><span lang=EN-US
style='font-size:10.0pt;font-family:"Tahoma","sans-serif";color:windowtext'>
drbd-user-bounces@lists.linbit.com [mailto:drbd-user-bounces@lists.linbit.com] <b>On
Behalf Of </b>Igor Neves<br>
<b>Sent:</b> 31 May 2010 14:33<br>
<b>To:</b> Michael Iverson<br>
<b>Cc:</b> drbd-user@lists.linbit.com<br>
<b>Subject:</b> Re: [DRBD-user] Infiniband card support and help<o:p></o:p></span></p>
</div>
</div>
<p class=MsoNormal><o:p> </o:p></p>
<p class=MsoNormal>Hi,<br>
<br>
<br>
<br>
On 05/31/2010 12:45 PM, Michael Iverson wrote: <o:p></o:p></p>
<div>
<p class=MsoNormal>Igor, <o:p></o:p></p>
</div>
<div>
<p class=MsoNormal><o:p> </o:p></p>
</div>
<div>
<p class=MsoNormal>I'm basically doing the same thing, only with MHEA28-XTC
cards. I wouldn't think you'll have any problems creating a similar setup with
the MHES cards.<o:p></o:p></p>
</div>
<div>
<p class=MsoNormal><o:p> </o:p></p>
</div>
<div>
<p class=MsoNormal>I've not attempted to use infiniband sdr, just ipoib. I am
running opensm on one of the nodes. I'm getting throughput numbers like this:<o:p></o:p></p>
</div>
<div>
<p class=MsoNormal><o:p> </o:p></p>
</div>
<p class=MsoNormal><br>
Would be very nice if you could test your setup with drbd infiniband sdr
support, probably you will not need to re-sync anything.<br>
<br>
<br>
<o:p></o:p></p>
<div>
<div>
<p class=MsoNormal>cirrus:~$ netperf -H stratus-ib<o:p></o:p></p>
</div>
<div>
<p class=MsoNormal>TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to <a
href="http://stratus-ib.focus1.com">stratus-ib.focus1.com</a> (172.16.24.1)
port 0 AF_INET : demo<o:p></o:p></p>
</div>
<div>
<p class=MsoNormal>Recv Send Send
<o:p></o:p></p>
</div>
<div>
<p class=MsoNormal>Socket Socket Message Elapsed
<o:p></o:p></p>
</div>
<div>
<p class=MsoNormal>Size Size Size Time
Throughput <o:p></o:p></p>
</div>
<div>
<p class=MsoNormal>bytes bytes bytes secs.
10^6bits/sec <o:p></o:p></p>
</div>
<div>
<p class=MsoNormal><o:p> </o:p></p>
</div>
<div>
<p class=MsoNormal> 87380 16384 16384 10.00
7861.61 <o:p></o:p></p>
</div>
</div>
<p class=MsoNormal><br>
Nice.<br>
<br>
<br>
<o:p></o:p></p>
<div>
<p class=MsoNormal><o:p> </o:p></p>
</div>
<div>
<p class=MsoNormal>A couple of things to watch out for:<o:p></o:p></p>
</div>
<div>
<p class=MsoNormal><o:p> </o:p></p>
</div>
<div>
<p class=MsoNormal>1. Upgrade the firmware on the cards to the latest and
greatest version. I saw about a 25% increase in throughput as a result. The
firmware updater was a pain to compile, but that was mostly due to Ubuntu's
fairly rigid default compiler flags.<o:p></o:p></p>
</div>
<p class=MsoNormal><br>
Will watch that!<br>
<br>
<br>
<o:p></o:p></p>
<div>
<p class=MsoNormal>2. Run the cards in connected mode, rather than datagram
mode, and put the MTU at the max value of 65520. My performance benchmarks of
drbd show that this is the best setup. <o:p></o:p></p>
</div>
<p class=MsoNormal><br>
If I use infiniband sdr support from drbd, should I care about MTU?<br>
<br>
<br>
<o:p></o:p></p>
<div>
<p class=MsoNormal>The replication rate on my setup is completely limited by
the bandwidth of my disk subsystem, which is about 200 MB/s for writes. I can
share some performance comparisons between this and bonded gigabit ethernet, if
you would like. However, I won't be able to provide it until tomorrow, as it is
a holiday in the US today, and I don't have ready access to the data.<o:p></o:p></p>
</div>
<p class=MsoNormal><br>
We have a couple of setups that have I/O performances greater than 500MB/sec,
so we really need 10Gbit trunks. <br>
<br>
Thanks for the help, but I don't need performance results from Gbit setup's, we
have a couple, and we know the problems! :) Anyway if you want to paste it
here, I guess no one will complain.<br>
<br>
<br>
<o:p></o:p></p>
<div>
<p class=MsoNormal>On Mon, May 31, 2010 at 6:17 AM, Igor Neves <<a
href="mailto:igor@3gnt.net">igor@3gnt.net</a>> wrote:<o:p></o:p></p>
<p class=MsoNormal>Hi,<br>
<br>
I'm looking for a 10Gbit backend for storage drbd replication. I'm expecting to
setup infiniband solution connected back to back, this means both nodes will be
connected together without a switch.<br>
<br>
I wonder if I bought two of this cards MHES14-xtc and a cable, I will be able
to produce such setup?<br>
<br>
Link to the cards: <a
href="http://www.mellanox.com/content/pages.php?pg=products_dyn&product_family=19&menu_section=41"
target="_blank">http://www.mellanox.com/content/pages.php?pg=products_dyn&product_family=19&menu_section=41</a><br>
<br>
Another question, I intend to use this with infiniband sdr support added to
drbd in 8.3.3, and I found this on the spec's of the card.<br>
<br>
"In addition, the card includes internal Subnet Management Agent (SMA) and
General Service Agents, eliminating the requirement for an external management
agent CPU."<br>
<br>
This means I don't need to run openSM in any nodes? I will just need to setup
two cards, a cable, connect them, and setup IPoIB to start replicating in
10Gbit?<br>
<br>
Thanks very much,<o:p></o:p></p>
</div>
<p class=MsoNormal><br>
Thanks, once again.<br>
<br>
<o:p></o:p></p>
<pre>-- <o:p></o:p></pre><pre>Igor Neves <a href="mailto:igor.neves@3gnt.net"><igor.neves@3gnt.net></a><o:p></o:p></pre><pre>3GNTW - Tecnologias de Informação, Lda<o:p></o:p></pre><pre> <o:p></o:p></pre><pre> SIP: <a
href="mailto:igor@3gnt.net">igor@3gnt.net</a><o:p></o:p></pre><pre> MSN: <a
href="mailto:igor@3gnt.net">igor@3gnt.net</a><o:p></o:p></pre><pre> JID: <a
href="mailto:igor@3gnt.net">igor@3gnt.net</a><o:p></o:p></pre><pre> PSTN: 00351 252377120<o:p></o:p></pre><pre><o:p> </o:p></pre></div>
</body>
</html>