Folks,

I am considering the feasability of building a distributed system out of
a network of low-cost single-board computers.

Physically constructing such a system is a straightforward enough
exercise, but finding and configuring a truly distributed OS (like
AMOEBA or some such) is a rather daunting prospect.

I suppose creating a Linux Beowulf cluster is one option, but I would
really like to hear some sound alternatives.

Thanks guys,
Gavin Brewer.

Gavin Brewer wrote:
Some background here:

http://en.wikipedia.org/wiki/Cluster_computing
http://en.wikipedia.org/wiki/Beowulf_%28computing%29

Paul

On Fri, 24 Nov 2006 15:34:23 +0000, Gavin Brewer <gb (AT) axicon (DOT) com>
wrote:

Have a look here: http://www.shadowflux.com/xbox.html - Xbox 1s are
real cheap at the moment - must have cost him a fortune at that time.

"Gavin Brewer" <gb (AT) axicon (DOT) com> wrote


Nope, not worth it...
if you use a bunch of *free* machines...

vs just one very high end such as a dual core...

the energy they will take will make the whole project un-feasible

On Mon, 27 Nov 2006 15:32:45 -0600, "philo"
<philo (AT) privacy (DOT) net> wrote:


Depends on how much processing power is needed, as well as
memory bandwidth, etc. I don't recall seeing any "low cost
single-board computers" though, they tend to cost quite a
bit for what they are though the lower volume sales might
easily account for that.

<snip>
Well when the X2's first came out...I waited a while and built a machine
as soon as the first price break came...
but now an X2-3800 is dirt cheap ...

Even though the electric bill to run a computer is not terribly high...
If one ran a dozen or so lower end machines...the cost of electricity would
make
the project a bit unattractive.

On Fri, 24 Nov 2006 15:34:23 +0000, Gavin Brewer wrote:

Have you seen:

http://www.mini-itx.com/projects/cluster/

Good luck,
Steve

On 2006-11-29, SteveSch <THISEMAIL (AT) IS_FAKE (DOT) COM> wrote:
12 nodes? That's a slightly generous definition of massively
parallel...

I have quite a bit of experience with the VIA EPIA boards, and
generally rate them, but I wouldn't consider them for performance.
Added to that, if you don't need everything (video, audio etc) on
board each system, a standard mid-range ATX mobo and chip will be
both cheaper and faster.

--
Andrew Smallshaw
andrews (AT) sdf (DOT) lonestar.org

"SteveSch" <THISEMAIL (AT) IS_FAKE (DOT) COM> wrote


That was a good article and I'm sure it must have been a lot of fun building
the machine.
It almost seemed like a worthwhile project until I read the "maintenance"
section and saw that it is *typical* to
have two nodes down (and I quote ):
"Plan on having about ten percent of the cluster failed or failing at any
given time. If you need a machine with 10 nodes operational, you had best
plan on having 12 nodes, and some spare parts. The larger the cluster is,
the more failed hardware you can expect. Really large clusters have hardware
failures on a more or less continuous basis. Alternatively, you can just
build a lot of extra nodes and take bad nodes offline as the cluster "burns
in" (this seems expensive and wasteful to me). Run the cluster on a good
UPS. It is not an option. You need clean power to get good hardware life,
and with this many computers the investment in a UPS will pay off in terms
of longer hardware life."



That's *totally* unacceptable. On my own machines I'd consider two failures
a year to be excessive!@!!

GB> Folks, I am considering the feasability of building a
GB> distributed system out of a network of low-cost single-board
GB> computers.

What is the intended use or purpose of this network?
- Best MFLOPS/$ ?
- Fault tolerant computing?
- Experimentation?


--
Politics doesn't make strange bedfellows, marriage does.
Groucho Marx