<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html><head><title></title>
<META http-equiv=Content-Type content="text/html; charset=iso-8859-1">
<meta http-equiv="Content-Style-Type" content="text/css">
<style type="text/css"><!--
body {
margin: 5px 5px 5px 5px;
background-color: #ffffff;
}
/* ========== Text Styles ========== */
hr { color: #000000}
body, table /* Normal text */
{
font-size: 9pt;
font-family: 'Courier New';
font-style: normal;
font-weight: normal;
color: #000000;
text-decoration: none;
}
span.rvts1 /* Heading */
{
font-size: 10pt;
font-family: 'Arial';
font-weight: bold;
color: #0000ff;
}
span.rvts2 /* Subheading */
{
font-size: 10pt;
font-family: 'Arial';
font-weight: bold;
color: #000080;
}
span.rvts3 /* Keywords */
{
font-size: 10pt;
font-family: 'Arial';
font-style: italic;
color: #800000;
}
a.rvts4, span.rvts4 /* Jump 1 */
{
font-size: 10pt;
font-family: 'Arial';
color: #008000;
text-decoration: underline;
}
a.rvts5, span.rvts5 /* Jump 2 */
{
font-size: 10pt;
font-family: 'Arial';
color: #008000;
text-decoration: underline;
}
span.rvts6
{
font-weight: bold;
color: #800000;
}
/* ========== Para Styles ========== */
p,ul,ol /* Paragraph Style */
{
text-align: left;
text-indent: 0px;
padding: 0px 0px 0px 0px;
margin: 0px 0px 0px 0px;
}
.rvps1 /* Centered */
{
text-align: center;
}
--></style>
</head>
<body>
<p>Hallo Bogdan,</p>
<p><br></p>
<p>Freitag, 18. April 2008, meintest Du:</p>
<p><br></p>
<p><br></p>
<p><span class=rvts6>BC> Sorry to divert a bit the thread towards its initial subject and away </span></p>
<p><span class=rvts6>BC> from the security issues currently discussed...</span></p>
<p><br></p>
<p><span class=rvts6>BC> I've just seen a presentation from a University (which shall remain </span></p>
<p><span class=rvts6>BC> unnamed) which partnered with Microsoft for... well, HPC. The reasons </span></p>
<p><span class=rvts6>BC> for using Windows were more or less the same that have been mentioned </span></p>
<p><span class=rvts6>BC> in this thread, so I won't repeat them. To note is that they weren't </span></p>
<p><span class=rvts6>BC> using Windows exclusively, but only on a part of the cluster, the rest</span></p>
<p><span class=rvts6>BC> running Linux.</span></p>
<p><br></p>
<p>Since i think i heard the same presentation i have to add some thoughts here...</p>
<p><br></p>
<p><span class=rvts6>BC> Towards the beginning of the presentation there was a mention of a MPI</span></p>
<p><span class=rvts6>BC> latency benchmark showing 2.something microseconds over their IB </span></p>
<p><span class=rvts6>BC> (unknown make and speed) in mainboards using latest generation Intel </span></p>
<p><span class=rvts6>BC> CPUs with Microsoft's MPI libs, which seemed like a decent performance</span></p>
<p><span class=rvts6>BC> and got me pretty excited. </span></p>
<p><span class=rvts6><br></span></p>
<p>But as far as i saw it they didn't state the type of IB. 2.7 is great for Infinihost III (normal SDR/DDR cards) - but pretty bad for Qlogic Infinipath or ConnectX. Since the cluster is pretty new one should ask which interconnect they are using.</p>
<p><br></p>
<p>When i saw the first performance numbers for CCS 2003 i was a little shocked - and the recent numbers seem to be a huge improvement. </p>
<p><br></p>
<p><span class=rvts6>BC> But then I changed my mind when I started </span></p>
<p><span class=rvts6>BC> to hear what a great feature it is to have several nodes booting and </span></p>
<p><span class=rvts6>BC> installing the OS in the same 50 minutes (yes, minutes!) that a single</span></p>
<p><span class=rvts6>BC> node takes, due to a wonderful feature called multicast. </span></p>
<p><span class=rvts6><br></span></p>
<p>50 minutes for a single node is of course unacceptable. 50 Minutes for 256 nodes is okay i think. But i doubt that it scales that well. Even Multicast packages get lost - needs retransmission etc. </p>
<p><br></p>
<p>Does Rocks (for example) in the latest release use Multicast to install nodes?
<br><br></p>
<p><br></p>
<p><span class=rvts6>BC> And then </span></p>
<p><span class=rvts6>BC> things turned really strange after a statement saying that in Linux it</span></p>
<p><span class=rvts6>BC> takes several minutes to start a parallel job while in Windows only </span></p>
<p><span class=rvts6>BC> about 10 seconds. Then I started wondering: were those 2.something </span></p>
<p><span class=rvts6>BC> microseconds a measure of the same latency that I know of ?</span></p>
<p><br></p>
<p>This was really strange. I never took the time of a startup so could please somebody tell me a number of 2048 processes starting on a linux cluster? Would be interesting to know. </p>
<p><br></p>
<p><br></p>
<p><span class=rvts6>BC> I can't say for sure that this was part of some Microsoft strategy and</span></p>
<p><span class=rvts6>BC> not a PR effort gone bad, but I'm strongly enclined towards the first </span></p>
<p><span class=rvts6>BC> which leads me to believe that the answer to the question in the </span></p>
<p><span class=rvts6>BC> subject is: by disinforming people. #</span></p>
<p><br></p>
<p>All presentations i saw about CCS or the new 2008 product where pretty honest about that they don't want to beat linux but want to make it easier for people that are no linux/unix experts. This presentation yesterday was pretty interesting because it did a real comparison with linux - and one has to see if the results are reproduceable. </p>
<p><br></p>
<p><span class=rvts6>BC> Yes, there are probably many CEOs </span></p>
<p><span class=rvts6>BC> of SMBs, who don't know/care much about technical details and don't </span></p>
<p><span class=rvts6>BC> have a clue about Linux HPC, who are going to be impressed by such </span></p>
<p><span class=rvts6>BC> statements. And when you can run HPL from Excel by modifying in a cell</span></p>
<p><span class=rvts6>BC> one of the parameters and getting the results back in that table, </span></p>
<p><span class=rvts6>BC> results from which you can quickly generate a graphic and say "whew, </span></p>
<p><span class=rvts6>BC> I should be in Top500", who can say that clustering is hard and </span></p>
<p><span class=rvts6>BC> user-unfriendly ?</span></p>
<p><br></p>
<p>Hey... the excel-sheet was pretty neat! ;-)</p>
<p><br></p>
<p><br></p>
<p><span class=rvts6>BC> I'm all for healthy competition in this area, especially as I think </span></p>
<p><span class=rvts6>BC> that HPC didn't evolve significantly in the past few years. But such </span></p>
<p><span class=rvts6>BC> aproaches are far from healthy... well, at least for my definition of </span></p>
<p><span class=rvts6>BC> healthy competition. ;-)</span></p>
<p><br></p>
<p>My oppinion is that many people in the HPC world just don't like Windows (including me). And a strange thing (that i experienced myself): even after years of using and deploying linux-clusters you are a newbie to the Microsoft-CCS world. That feels really strange because you're no expert anymore. Maybe that keeps people from liking MS CCS (or HPS 2008).</p>
<p><br></p>
<p>Cheers,</p>
<p>Jan</p>
<p><br></p>
<p><br></p>
<p>Bye Jan </p>
</body></html>