FW: Total Cost of Ownership and porting (was FW: Can we have a mo ment of silence (or several million dollars) . . . please?)
Robert G. Brown
rgb at phy.duke.edu
Tue Jun 26 12:31:31 PDT 2001
On Tue, 26 Jun 2001, Schilling, Richard wrote:
> > -----Original Message-----
> > From: Greg Lindahl [mailto:lindahl at conservativecomputer.com]
> > Sent: Monday, June 25, 2001 7:57 PM
> > Even if you stay on a x86 32-bit platform, you have the "portability"
> > problem of > 4gig files, which sometimes makes you have to modify your
> > code. Even if all the system routines and the Fortran runtime use LFS
> > calls, you'll still have some places to hack in your C code.
> > -- g
> Exactly. There was recently a discussion on this list about keeping a
> registry of Beowulf clusters, and I would include porting costs in that
I think that I'll open up our site to a BRIEF round of RFC access, since
it is almost finished to the alpha stage and I'd rather incorporate
suggestions now before I quit working on it for a while.
Really interested beowulfers (only, please!) may therefore visit:
and look it over. Feel free to register a cluster for fun, remembering
that I will DELETE all registered clusters before putting the site up
for Duke. I haven't quite gotten the user survey password feature in
place but you can also register yourself as a "beowulf user" (as opposed
to and in addtion to being an "owner").
Note that my inclination thus far has been to keep the survey/registry
form >>simple<< in at least the first incarnation, and add detail and
additional reports as they really seem warranted. This is partly sheer
laziness on my part -- I don't want to mess with making the tables fully
relational and haven't even THOUGHT about building quantitative reports
yet (e.g. bar histograms, averages, trends).
What I would like to know is:
a) What am I missing that you REALLY would like to know about somebody
else's cluster(s)? Note that thus far I've eschewed e.g. enumerated
lists of CPU types and speeds and itemized counts and details of network
architecture because I think people would get bored trying to work their
way to the bottom of an eight page form. I do have a freeform
"description" field where folks can say as much or as little as they
wish in whatever level of detail they choose. There is also a field for
a cluster URI if you've got one so that you can refer people to the
cluster's own website for all the detail in the world, and a field for
your email address so they can contact you directly with questions.
b) What are the primary reports you'd like to see? So far I have only
a table view of the registered clusters. I assume that things like
"total number of e.g. XXX-based clusters", "total number of XXX-CPU
nodes", "total number of XXX-network sites" (where I'm referring to the
PRIMARY values of XXX in heterogeneous environments) would be of
interest. How about specific tools?
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 at phy.duke.edu
More information about the Beowulf