new SGI Origin & Onyx 3k?

[Michael Huntingdon]

>Date: Wed, 26 Jul 2000 22:39:48 -0500
>From: "W Bauske" <wsb at paralleldata.com>
>Organization: PDS Inc.
>X-Mailer: Mozilla 4.7 [en] (X11; U; Linux 2.2.5-15 i686)
>X-Accept-Language: en
>To: beowulf <beowulf at beowulf.org>
>Subject: Re: new SGI Origin & Onyx 3k?
>Sender: beowulf-admin at beowulf.org
>X-Mailman-Version: 1.1
>List-Id: Discussion of topics related to Beowulf clusters
<beowulf.beowulf.org>
>X-BeenThere: beowulf at beowulf.org
>
>Greg Lindahl wrote:
>> 
>> > > For the customers I'd like to have, frequently.
>> >
>> > Note the "Like to have". Also, you don't address the first
>> > question so I'll assume you agree people don't do that unless
>> > it's for benchmarks from your statement.
>> 
>> Wes, I have no interest in getting into a dicksize flamewar. All I
wanted to
>> do was make the pithy comment that SGI shouldn't call it scalable if it
only
>> goes small. If you think I'm unqualified to make an off-the-cuff remark,
>> please feel free to ignore it.
>
>No flame war intended. Sorry if it seemed that way.
>
>> 
>> There are people who DO use 1,000+ nodes for non-benchmark purposes. I DO
>> work with them already. I don't think my experience is very relevant to
>> making comments about scalability; anyone can look at the Top500 list and
>> other sources and get an idea where supercomputing is going.
>> 
>
>I agree there are very large set of machines out there.
>What I said was that most jobs run in parallel are not on
>those size systems. Practically no one but the government
>own that size of system so by default the rest of us are
>running on smaller systems. The large systems are an exception,
>not the rule. Smaller systems are the norm. That was my main
>point. Your using a large system is outside the norm and
>I do believe you run benchmarks, not daily/weekly/monthly
>production runs producing some results with an application. 
>If I don't understand what you do, please correct me.

I almost hate to enter the conversation at this point as this seems to be
progressing quite nicely; however, I believe the premise is becoming a bit
skewed and perhaps short sighted. The "envelope pushing" expertise
associated with this group may or may not be specific to environments of
128 to 1,000+ nodes or CPU's. I'd be willing to bet that many of us are
working with groups developing code that can take advantage of ten thousand
or more CPU's.

The premise here should not be limited to what x-vendor can accomplish
today..rather what they can be driven to through the brain-power of groups
like this. As mentioned by Greg, 1000 CPU's is an easily (ok..maybe not so
easy but) obtainable goal. It's not what vendors can do now, but what is
needed to support the needs of science and research. Drive the numbers guys!! 

And where is Robert when he's really needed...vacation?

cheers
~m


>_______________________________________________
>Beowulf mailing list
>Beowulf at beowulf.org
>http://www.beowulf.org/mailman/listinfo/beowulf
>