Who runs large capability jobs

[Greg Lindahl]

> One thing of interest to me on that list is that one can
> take 124 Power3II's and rank 95th on the list. So you're
> talking about the performance of the 94 machines above that
> and I'm talking about the performance of the 406 systems at
> or below that.

Right. Whatever. As I said, I don't care about that issue.

> Wait a minute. This is contradictory. First you say the little
> guys go home and then the general researcher gets access to
> the big systems. Sounds to me like only certain groups will get
> access and thus many researchers will be left to fend for themselves
> to get processing capacity, probably building their own small
> Beowulfs. Which is it?

What's the question? Yes, not everyone gets huge amounts of time; other
people are expected to use local machines, be they their own clusters, or
resources at their University.

> > But I assure you that the protein folding guys are headed to
> > 1,000+ cpu runs. And they have hundreds of millions of dollars
> to spend. On
> > computers.
>
> Question I have is how many companies do this vs other types
> of cluster processing?

As I said before, I don't know or care.

> The origins of this digression were your comments on SGI
> not being scalable systems. It may well make sense for some
> situations but it is unlikely to be true for many others.

Calling a system "scalable" because it works for the average problem is
senseless. The average overall computer user has problems that need 1 CPU.
Does that make a desktop "scalable"? No, it makes it "sufficiently scalable
for the problem that user wants to run". That's entirely different.

> Also, the highest ranking SGI Origin is number 3 on the top500
> list so for some problems, SGI's obviously scale.

No. That is not a single SGI Origin, which is what they were claiming was
scalable. It's a cluster of them, and it doesn't work very well. The % of
peak it gets on Linpack, for example, is quite low, lower than expected. It
was quite a coup that ASCI Red is still #1 on Linpack, despite #2 and #3
having substantially higher machoflops (theoretical peak speed).

> One thing I learned when I worked for IBM is to never badmouth
> another vendor's equipment. Let the numbers/vendor speak for
> themselves.

I made a specific statement about scaling to 1,000 processors. A single SGI
machine does not scale that big; the biggest is 512, and the O3000 doesn't
raise that. The numbers speak for themselves. I don't know why we've had
this long thread about it. I hope someone got something useful out of it,
but it's been mostly people talking past each other.

> How about back to Beowulfs now.

Sure.

-- g