Michael Huntingdon wrote:
At 04:13 PM 8/15/2006, Mike Davis wrote:
I'm not 100% sure about that Mark. I care about big-A administration.
I care about showing departments what resources are actually
available. I care about what is the most efficient use of limited
University resources. When I meet with researchers they often say that
they had no idea that there were 500+ processors dedicated to research
here.
I know that other people have the same issues. Another is the funding
model issue. Which is best overhead, direct, or central budget? Or how
about knowing what resources we each provide our users. Does a given
organization focus on hardware support, software support or both?
Those are some of the Big-A issues. Here is one that is both Big-A and
small-A.
Running one of the new Sun x4100's with both dualcore processors at
100% uses <270 watts (as determined by kill-a-watt. That is Big-A
because it means that we can be more efficient in our use of AC and
power. It is small-a for the same reasons. For example spinning up a
v20 uses 250 watts for both processors at full power. I can't discuss
some of my application specific performance due to license
constraints, but I can say that I like the 4100 in general for
Computational Physics and Chemistry.
I often scratch my head wondering how certain decisions are make at the
"central IT" level, so a perspective from the campus that involves both
performance and up time (plug in the wall) costs is refreshing. We far
to often see a complete disconnect between the two, which very often
means that none of the invested parties (at either the
NSF/NIH/state/federal level) ever really enjoy the value of each dollar
they invest.
My HPC efforts are now involving both cases of 'A' as we're trying to
change the HPC paradigm on campus from almost solely SMP to a
combination of memory paradigms to better serve the research community.
And the 'we' is not directly in the Administration chain but is from a
couple of on-campus players who are frustrated with the status quo. I
look at infrastructure costs as well as per-node costs, maintenance and
potential down-time vice spares. Also, I now have to consider
networking costs, both on campus and in our commodity, Internet2 and NLR
dealings. Bandwidth is no longer free, although a lot of researchers
still think it is.
I appreciate that Sun may be suggesting (these days) that their systems
are more environmentally friendly; however, given the
price/performance/environmental/support...and really crazy extended down
time associted with engineering issues, logic at least for some, makes
distancing significant IT investment with Sun a decision that follows
very few conversations.
Interesting: I've started buying a lot more Sun as their costs are
comparable to the blade cluster prices I've gotten from other vendors,
we have seen as good, or better up-times with v2100 through v4200 series
hardware, and time-to-repair has been stellar, with the on-site Sun
support. When we had problems flashing a BIOS upgrade and couldn't
recover it ourselves, our last real Sun downtime, they had a new
motherboard in within 6 hours and the system back up an hour later. I
haven't suffered from the engineering issues you seem to have encountered.
My point comes honestly from your comments, which we hold dear.....the
growing number of research system/cpu's on campus affect each and every
one of us on a daily basis. Having spent this week at the LSS event at
Stanford, I am ever more convinced, how diverse the needs...and the
number of possible solutions. So that must be a big-A approach with a
huge tilt in a not so big-A direction.
Another that is both is what submission systems we are using and Why?
Same questions, that affect both administration and Administration.
Mike davis
Mark Hahn wrote:
beowulf traffic itself is "noise"? If you are thinking of a "list for
university deans" or members of research support offices or
departmental
...
administerable and accountable should they get audited) -- then yeah, I
think a new list or other venue would be very useful.
yes. the overlap is minimal, I believe - I'd say the two approaches
are even inimical. someone who is primarily interested in big-A
Administration will have values opposed to mine as a technologist.
as a random pot-shot, big-a people tend to have great faith in
negotiating special purchasing relationships with a vendor, or
believe that integration
is the high-road to success (or an end in itself). I know, OTOH,
that a vendor who makes a good desktop may make the worlds worst compute
nodes, and that, for instance, the service requirements are nearly
opposite.
here's my general conclusion about central-IT efforts: if the idea
(centralized storage, whatever) is so good,
people will beg to use it. if you have to force people to use it,
you are simply wrong in some way (perhaps subtly).
Just to touch on this, I'm in general agreement, although our big-a have
just negotiated a big-iron acquisition for a decent price. It'll go
into the central IT core for shared computing resources... I'll see how
well it works out. The central IT HPC folks were the proximate cause of
me building my first cluster...
gerry
--
Gerry Creager -- [EMAIL PROTECTED]
Texas Mesonet -- AATLT, Texas A&M University
Cell: 979.229.5301 Office: 979.458.4020 FAX: 979.862.3983
Office: 1700 Research Parkway Ste 160, TAMU, College Station, TX 77843
_______________________________________________
Beowulf mailing list, Beowulf@beowulf.org
To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf
