[Beowulf] Beowulf building book (fwd)

[Robert G. Brown]

And another.  Recycling old hardware into OTHER beowulfs is another
often held dream, with the usual problem of getting functioning hardware
from point A (where it is) to point B (where it is needed/wanted).
There are also energy efficiency/Moore's Law considerations that have
been discussed on list before.

For this sort of thing to work, there needs to be some sort of "cluster
exchange" site where people can post a list of nodes they'll give away
and other people can offer to pay the shipping... maybe CWM can set such
a thing up, or a list volunteer with a Green philosophy can contribute
and periodically advertise...

    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 at phy.duke.edu



---------- Forwarded message ----------
Date: Sat, 22 Jan 2005 02:01:19 -0500 (EST)
From: EAGLESON NATHAN A <nae119 at psu.edu>
To: rgb at phy.duke.edu
Subject: Beowulf building book

Dear Dr. Brown,
   My name's Nate Eagleson, and as you might gather from my email address, I'm
an undergrad student at Penn State. As you might guess from the subject line of
my email, I'm a computer science major.

   In particular, I'm a computer science major who has found himself with a
lightweight semester at the end of his senior year, and decided to hunt up a
good geek project to cap off my "career" as an undergraduate student. After
reflecting for some time on what to do, the answer sprang to me in a blinding
flash of the obvious:

   Build a Beowulf cluster.

   So, I took inventory of my resources, which happened to involve just about no
money.

   It was at that point that I had an interesting thought, which was more or
less as follows: "There's a fair amount of old hardware in the world that
people just throw out. This is bad for the environment, and it's also a waste
of resources that might have some possible use."

   This has provided me with something of a focus for the project, which is to
see whether a smallish set of machines that are decrepit by modern standards
can be used in parallel and/or pseudo-parallel fashion to make a useful working
environment of some sort, rather than simply allowing old hardware to be thrown
out and cause that little bit more of pollution.

   My success in collecting hardware at no cost has been surprising so far;
simply by poking my head into one recycling center twice in the past two weeks,
I've been able to obtain four machines in the 200 - 400 MHz range. Two of these
seem to be fully functional, while two others are not so, and will probably be
stripped for parts if this is needed. None of the machines have Ethernet cards,
but a little research has shown that this can probably be solved within even my
financial means.

   Well, you probably aren't terribly interested in hearing me rattle on about
hardware. In particular, I wanted to say two things:

1) I have found the book you've written on Beowulf cluster design to be
extremely helpful. Although parts of it were inapplicable to my case (I don't
have much choice in terms of hardware options; I'll take what I can get ;-),
overall, it gave me a very good introduction to the main ideas behind building
a Beowulf, and to some of the practical implementational details. Having read
through it once, I think I know more or less what to do, and once I've
accumulated all the important hardware, I'll be going through it again, as I
actually plan the cluster. I really appreciate the effort you've gone to in
helping people like me learn about cluster computing, and if ever I have a
chance to meet you, I'll most definitely remember the Beverage Modification to
the OPL.

2) At several points in your book, you asked for user input, if we had any to
give. At most points, being a complete novice, I of course have nothing useful
to say. However, on the subject of recycling computers, I do have one thought:

Give it away.

If your hardware is too old for your company to get good use out of it, or to
reasonably sell it, just offer it to those who are interested. To most people,
it will have no value, but to students like myself, or to people with children
who are bugging them for "my own computer", or to hardware hackers who could
use a spare case or three, a stack of old machines offered for free is a
wonderful thing, and every machine and/or part that someone takes home is one
fewer in the landfill, for as long as they use it. It would also be good PR for
a business or school, to be seen giving a block of old computers away to
members of the community who could use them.

I know of a large company in my area who throws out serviceable hardware, even
relatively modern hardware that's still useful to almost anyone (such as 21"
CRTs, when they bought a number of LCD screens). My friends who have
connections there try to salvage what they can of it, but a great deal winds up
being thrown out. It hurts to think that hardware is out there doing nothing
but busily polluting the groundwater table (as you pointed out in this
section), but it hurts almost as much knowing that people with an interest in
computing could be using this older hardware for something useful, such as my
current attempts to learn more about parallel processing.

Everything I've seen in the Beowulf design book indicates to me that you and
your colleagues do your absolute best to reuse old hardware and keep it working
usefully as long as you can, but perhaps people who may run across your work
will not do so. Reminding them that some people can make use of what they would
otherwise throw out might be a good idea.

Again, I enjoyed your book a great deal, and I think it has me pretty well
prepared to tackle this project. Thank you so much for taking the time to write
it.

Sincerely,
Nate Eagleson