[Bioclusters] Workstation Selection for Bioinformatics Research

[Hunter Matthews]

On Thu, 2003-04-10 at 14:31, Duzlevski, Ognen wrote:
> Hi Dr. Wicks,
> 
> can you elaborate on the price tradeoffs? How many more PCs would you get
> for a number of Macs? I am not anti-Mac, just curious to know what is an
> actual benefit in even thinking about Macs in a cluster.
> 

Maybe 1.5:1 or even 2:1. 

And the macs are definately MUCH slower in most situations for
computation programs. (The exceptions are notable however- HMMER with
the altivec turned on runs rings around most anything).

[And yes, I have run benchmarks with our actual data]

The macs are nice in that they provide a pretty interface AND most of
the stability/network admin abilities of the unix machines. They're
grossly annoying because there are still things they either can't do or
seem to not want you to do.

Right now, we're suggesting the Dell Optiplex GX260's for most everyone,
in the small form factor. They are very small, power efficient units
that look good (always a factor with faculty) on a desk. Thermal is
"good" - better than any athlon, not as good as a larger unit with
better internal space.

"But But -- I can't put 8 hard drives and 3 cdrws in those!" No, no you
can't. 95% of our users don't want to. For the others, we just get the
larger cased GX260's. (and pretty much watch them not use the extra
spaces)

Dual booting proved problematical here - we do nightly patch runs in
both the unix machines and now the windows machines, so whichever OS was
booted, it was the other OS that needed the critical patch. 

VMware is more attractive for "dual OS" type situations, and we have one
faculty member who swears by cygwin (letting him develop in a "unix
like" enviroment under XP). 

If we make a "mistake" and buy the wrong OS for a particular faculty
member, with the dells we just wipe the hard drive and start over with
the other OS. If you purchase non-PC's, this isn't an option.

We have a number of "clone" (high quality clones, in my opinion, but we
do have a number of "ICK" level boxes) systems. They're ok for
computational nodes where the loss of a single node is not a disaster.
However, we've come to rely on the 3yr part/labor warranty of dell so
much we're now requiring that level of warranty on ALL systems, and this
markedly reduces the price differences from the clones to the dells.

Why? We have 4 technical IT people here and hundreds of faculty,
students and staff to support. Our time is worth something. I HATE
having to figure out some other companies support policy or worm my way
into their RMA setup. Dell is far FAR from perfect -ask any dell support
manager and tell him/her Hunter sent you :). But they are fairly
consistent.

The workstations are nice, and are more powerful. But day in and day out
we can get most of what we need from the optiplex line. 

My one personal exception is laptops. The dell laptops just suck when
you sit them down and compare them to an apple or IBM unit.

That probably answered no ones question and will just get the
workstation/apple/stinkpad people mad at me. Oh well.


-- 
Hunter Matthews                          Unix / Network Administrator
Office: BioScience 145/244               Duke Univ. Biology Department
Key: F0F88438 / FFB5 34C0 B350 99A4 BB02  9779 A5DB 8B09 F0F8 8438
Never take candy from strangers. Especially on the internet.