On Fri, Apr 22, 2011 at 3:22 PM, Brad Bendily <bendily@gmail.com> wrote:
> My understanding of a Beowulf cluster is that you have to have
> software written specifically for a "cluster".
> So, i don't think you can just run any software program on a Beowulf
> cluster and it will go out and use all the nodes.
>
This is one of my suspicions. I didn't list what I would define as failure,
but if I had to pick anything, it would be that special programming would be
necessary making even the experiment too time consuming.
> From the wiki article:
> "There is no particular piece of software that defines a cluster as a
> Beowulf. Commonly used parallel processing libraries include Message
> Passing Interface (MPI) and Parallel Virtual Machine (PVM). Both of
> these permit the programmer to divide a task among a group of
> networked computers, and collect the results of processing. Examples
> of MPI software include OpenMPI (OpenMPI) or MPICH (MPICH). There are
> additional MPI implementations available."
>
I keep having to read that part 3 times to get what they're saying, but
yeah...there seem to be libraries for doing things, but there's not a real
demand for what I'm looking at doing. No biggie. I'm feeling better about
not being in rush to do this. I still want to, just for the experience of
doing it.
>
> You can potentially make those $1000 desktops serve up 2-4 Terminals
> (on thin clients) but that may be more
> trouble than it's worth. Part of the problem with the whole "Thin
> client" is that they're not that cheap.
> You're talking $200-$400 a piece. When you can buy a usable dell for
> $200-$400 new or used.
>
>
Yeah, I've found some clients for as low as $50. So prices on thin clients
is still attractive. If nothing else, I could reduce the useless overhead
in our customer service department. And depending on growth in Production,
thin clients may still have a place.
So I still want to set up a small thin client network at my house to build
my understanding of the limitations involved.
> I'm not sure how you would spread ltsp over a cluster.
> I have setup ltsp, we used one server to host a web browser only
> desktop for 20-40 clients.
> As I recall it worked pretty well.
>
Glad to hear there was success. I need that pep talk because as you say,
how to do it over a cluster is going to be the question to answer.
>
> I would think that getting a beowulf cluster of LTSP nodes would be a
> real pita.
> however, after googling "beowulf ltsp" people have asked the question
> before.
> Here is the middle of a thread on another lug list:
> http://www.dcglug.org.uk/archive/2008/06/msg00116.html
>
>
Hot dog. Appreciate the leg up on the research. Even if there's no
finished success there, it will be interesting to see what others have
accomplished on this question. Thanks.
> good luck!
> let us know how it goes!
> bb
>
>
Oh if it goes well, I'll be sure to boast with specifications. Never know
who might actually find it useful.
If I fail, I will probably continue asking questions to pick the group
brain.
___________________
Nolug mailing list
nolug@nolug.org
Received on 04/22/11
This archive was generated by hypermail 2.2.0 : 04/22/11 EDT