Very good. I am considering Infiniband too, based on the work that is mentioned by Kamal. Anybody has any idea of the cost involved in Infiniband infrastructure? <br><br>saji<br><br><div class="gmail_quote">2011/4/15 Ólafur Rögnvaldsson <span dir="ltr"><<a href="mailto:or@belgingur.is">or@belgingur.is</a>></span><br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">I agree with Aaron, it's the low latency of the InfiniBand that makes it a more suitable inter-node fabric than 1/10G Ethernet.<br>
<br>
Kind regards, Ólafur.<br>
<div><div></div><div class="h5"><br>
> Mike,<br>
><br>
> Its been my experience that latency is the critical issue when running<br>
> wrf in parallel, not raw speed. I would spec out the latency of 10GigE,<br>
> which I think is about the same as GigE, and compare it to Infiniband.<br>
> I think youll find infiniband is better in this case.<br>
><br>
> Aaron<br>
><br>
> Zulauf, Michael wrote:<br>
>> Hi all - quick question. Does anyone have data or experience comparing<br>
>> the performance and scaling of WRF with cluster interconnects utilizing<br>
>> 10 gigabit ethernet (10GigE) vs Infiniband?<br>
>><br>
>> We're looking to expand and update our computing resources, and we'd<br>
>> originally spec'd it with Infiniband. The nodes will have either dual<br>
>> quad- or hex-core Nehalem type processors. There's been some IT<br>
>> pushback, suggesting that we should go with 10GigE. I don't have any<br>
>> direct experience with 10GigE, but my experience with 1GigE shows that<br>
>> Infiniband scales far better.<br>
>><br>
>> I've seen what I'd consider marketing material on the web that suggests<br>
>> that 10GigE is comparable to Infiniband, but they don't specifically<br>
>> mention WRF. On the other hand, I've seen other sites that suggest an<br>
>> Infiniband interconnect is far superior. Again, these don't<br>
>> specifically mention WRF. I know the particular application in use is<br>
>> critical when deciding these things, and that WRF is a pretty demanding<br>
>> application when it comes to the interconnect.<br>
>><br>
>> My suspicion is that Infiniband is still significantly superior, but if<br>
>> I'm going to be able to make any headway with IT, then I'll probably<br>
>> need some type of numbers to back up my arguments.<br>
>><br>
>> Can anyone help?<br>
>><br>
>> Thanks,<br>
>> Mike<br>
>><br>
>><br>
> _______________________________________________<br>
> Wrf-users mailing list<br>
> <a href="mailto:Wrf-users@ucar.edu">Wrf-users@ucar.edu</a><br>
> <a href="http://mailman.ucar.edu/mailman/listinfo/wrf-users" target="_blank">http://mailman.ucar.edu/mailman/listinfo/wrf-users</a><br>
<br>
</div></div><font color="#888888">--<br>
Kveðja/regards,<br>
Ólafur Rögnvaldsson<br>
Reiknistofa í veðurfræði - Belgingur<br>
<a href="http://www.riv.is" target="_blank">www.riv.is</a> - <a href="http://www.belgingur.is" target="_blank">www.belgingur.is</a><br>
</font><div><div></div><div class="h5"><br>
<br>
_______________________________________________<br>
Wrf-users mailing list<br>
<a href="mailto:Wrf-users@ucar.edu">Wrf-users@ucar.edu</a><br>
<a href="http://mailman.ucar.edu/mailman/listinfo/wrf-users" target="_blank">http://mailman.ucar.edu/mailman/listinfo/wrf-users</a><br>
</div></div></blockquote></div><br><br clear="all"><br>-- <br><br>Saji N Hameed,<br>ARC-ENV, Center for Advanced Information Science and Technology,<br>University of Aizu, Tsuruga, Ikki-machi,<br>Aizuwakamatsu-shi, Fukushima 965-8580,<br>
Japan<br> <br>Tel: +81242 37-2736<br> email: <a href="mailto:saji@u-aizu.ac.jp">saji@u-aizu.ac.jp</a><br> url: <a href="http://www.u-aizu.ac.jp">http://www.u-aizu.ac.jp</a><br> bib: <a href="http://www.researcherid.com/rid/B-9188-2009">http://www.researcherid.com/rid/B-9188-2009</a><br>