Beowulf in a Box (fwd)
Robert G. Brown
rgb@phy.duke.edu
Tue, 29 Sep 1998 14:38:32 -0400
On Tue, 29 Sep 1998, Kragen wrote:
> On Tue, 29 Sep 1998, Robert G. Brown wrote:
> > I wouldn't be surprised if an Intel
> > Human or two listens in on the beowulf list, but for obvious reasons
> > (if one thinks about it) they need to be mousey-quiet.
>
> Well, I've thought about it. Why do they need to be mousey-quiet?
Because they do quite a lot of business with Microsoft, on the one hand,
and the government, on the other. beowulfs are linux, and I don't think
Intel has wanted to appear particularly receptive to linux, at least in
public, until it was ready (Intel, not linux, which has been ready for
years now). You may think that this is absurd and paranoid, and it may
be -- however, I've watched the conjoined fortunes of Intel and
Microsoft soar ever higher for 16 years now, and I think that they take
the business relationship very seriously. Seriously enough to passively
ignore possible Microsoft competitors, anyway, at least until they have
proved themselves too strong a force to be ignored.
The government (IMHO) is actually the more touchy problem. You see,
Intel boxes alone are still a wee bit away from the level of power that
the U.S. government considers a weapon (for better or worse or stupider)
and controls as an export. (I'm not too sure about 4 processor Xeons --
they may be there already). Ten or twenty Intel boxes put together,
however, clearly violate export controls. Somebody might use them to
decrypt our secrets or calculate missile trajectories or something.
As long as Intel hasn't/doesn't get into the beowulf business itself, I
think that they are relatively "safe" from the government from an export
control point of view -- if a customer like, say, Israel or India wants
to buy ten high end Intel boxes they certainly can (and do!). If they
want to buy a T3E or Origin or SP2, however, they have to literally make
a diplomatic issue of it, which is expensive, cumbersome, and negatively
impacts sales. On the other hand, if that customer hooks the 32 dual
PII's together into a beowulf capable of enough FLOPS to blow an SP2 out
of the water on plenty of "restricted" tasks like decryption (which
parallelizes very nicely) or whatever, hence clearly violating export
controls, it's not Intel's fault (and is a possibility that the bean
counters are blind to -- they can read single system specs off a chart
but don't think about adding the systems and their capabilities because
neither Intel nor anybody else officially acknowledges the possibility.
As soon as Intel endorses beowulfs or starts openly building beowulfs,
possibly for remarket, the whole parallel/component supercomputer issue
will come to a political head with the same folks that brought you the
V-chip and are still trying to quash RSA technology in charge. It may
be that they are ready to come out of the closet; Moore's Law is going
to make the export control levels pretty absurd in the next year or two
regardless. I think high end DEC alphas will start to violate them this
year or next year, I cannot remember, but it was discussed on one of the
linux lists a year or two ago.
I'm curious -- Doug (Eadline), do you guys have to pass your turnkey
beowulfs through any kind of customs thing? Have you run up against
export controls, or do you ship your systems as "components" and avoid
the problem? Or do you export your systems at all?
> Red Hat is funding GNOME, and some time ago, they issued a public
> statement about why they will no longer be including KDE on their CDs.
> Also, they announced last week that they would (a) no longer ship CDE
> because of security problems; (b) continue to support copies of CDE
> they'd already sold; and (c) give people a refund if they'd bought
> CDE.
OK, makes sense, didn't know that. I do know that KDE has some
copyright issues tied up in it -- we've been giving it a try locally.
> The source for almost every figure on the number of Linux users is Bob
> Young's white paper, "Sizing the Linux Market", which was updated early
> this year to the figure "5 to 10 million". It also mentions that the
> market doubles yearly, so we're probably well over 10 million by now.
Very cool.
rgb
Robert G. Brown http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567 Fax: 919-660-2525 email:rgb@phy.duke.edu