[Bioclusters] resources on administering clusters

Jeff Layton bioclusters@bioinformatics.org
Mon, 25 Mar 2002 14:46:59 -0500


chris dagdigian wrote:

> Jeff Layton wrote:
>
> ><snip>
> >
> >First, I take issue with the idea that because I'm interested in
> >pure speed, that I don't take the time or care to develop or find or
> >buy the tools I need to admin the clusters effectively and efficiently.
> >I probably do the same things you do, but instead of coming up with
> >tools that ease my admin burden, I come up with tools that ease
> >my admin BUT don't compromise speed. I don't now if people
> >document what they do in making these tools, but if you start searching
> >the web, you'll find them (at least I did).
> >
> ><snip>
> >
> Hey Jeff--
>
> Good to see a critical mass building up on the list. I didn't mean to
> insult the speed demons out there- it just that I've seen lots of groups
> (particulary biologists in academic settings) do really hair-raising
> stuff like wire capacitors into their motherboards to they can be
> overclocked safely. I have no problem doing that stuff at home but I'd
> be worried if that hardware had to live in a datacenter and be cared for
> by the IT group. This reflects my bias of course as someone who came up
> on the industry/biotech side of the fence.

No problem. I sometimes get a little sensitive :) I think the next "wave"
of clusters is in the Bio-computing field. We're already seeing it and it
looks like it's picking up momentum. The interesting thing to me in
the Bio-cluster field is that the clusters are really compute farms but
the IO requirements are potentially very large. This shows my lack of
knowledge of the field. I think it would be just great if someone could
present a non-Bio perspective of how Bio apps work and where potential
bottlenecks are. I would really LOVE to see something like that!

I also come from the industry side where we carefully consider all
of the issues you mentioned.

I've never heard of anybody mod-ing the motherboard like that!
That's scary for "production" servers. When Beowulfs first made
an appearance a few years ago, there were many small companies
offering clusters with over-clocked CPUs. Not a good thing to put
your work (and possibly other people's lives) in the hands of a CPU
that's out of spec.



>
>
> I will say though that I still believe that people put far too much
> emphasis on hardware price/performance without looking at releated stuff
> like physical footprint, electrical consumption, heat output &
> management overhead. Those dual-AMD systems are really fast but if you
> don't realize that they come with 550-watt power supplies your
> cooling/HVAC people people are going to be pretty upset.

Oh yeah. Watch electrical, cooling (cooling is just not enough,
you also need airflow), weight can be an issue, dimensions
(imagine a doorway that's too low - it's happened), and lockable
racks (parts have a tendency to disappear). There are lots of things
that need to be consider.

Thanks!

Jeff



>
>
> -Chris
>
> _______________________________________________
> Bioclusters maillist  -  Bioclusters@bioinformatics.org
> http://bioinformatics.org/mailman/listinfo/bioclusters