Re: [Nolug] Beowulf Cluster and Thin Clients

When implementing TC, you typically get a beefy server for N number of
users, and a bunch of (usually low-power) diskless desktop workstations
w/ monitors. More users? More servers. But a specific user's "work"
is always done on the server that he (transparently) logs into. His
work (running GIMP, for example) is never distributed across multiple
servers. This is *not* a cluster.

A computational cluster (of which Beowulf is the most popular form)
typically has a "controller" (with lots of disk space) and the compute
nodes which are typically blades or 1-2U rack mounted machines w/ fast
CPUs and lots of RAM but not much disk space. There's no reason,
though, why an organization's desktop computers couldn't be
dual-purposed as compute nodes.

Note the difference between TC and Beowulf: the low power TC desktops
aren't beefy enough to effectively "compute".

On 04/22/2011 07:03 AM, Clint Billedeaux wrote:
> Well,
>
> I know absolutely NOTHING about both, but I've never considered ignorance a
> bar to learning from experimentation.
>
> So:
>
> Planned experiment one: Build a Beowulf Cluster.
> Goal: Learn about Beowulf Clusters
> Terms for success: develop an understanding of what EXACTLY a Cluster does.
>
> Planned experiment two: Build a Thin Client/Server Network.
> Goal: Run multiple clients from a single Linux Box. see www.ltsp.org
> Terms for success: develop the skills and confidence to handle Thin Clients
> and Servers.
>
>
> Premise of PE3: "One of the main differences between Beowulf and a Cluster
> of Workstations<http://en.wikipedia.org/wiki/Cluster_of_Workstations> (COW)
> is that Beowulf behaves more like a single machine rather than many
> workstations." (from the original how-to which was published by Jacek
> Radajewski and Douglas Eadline under the Linux Documentation
> Project<http://en.wikipedia.org/wiki/Linux_Documentation_Project> in
> 1998.)
>
> I understand that behaving LIKE a single machine and BEING a single machine
> are two different animals, but I'm willing to experiment.
>
>
> Planned experiment three: Build a Beowulf Cluster/ThinClient Server
>
> Goal: To successfully utilize the processing power of multiple computers to
> support the functions of the Server Node on a Beowulf Cluster working also
> as the Server Node on a Thin Client Network.
> Terms for success: Learn EITHER way whether or not such a merger of concepts
> is worth the effort.
>
>
> Summary: Instead of having multiple machines crunching away on individual
> graphics applications, it would be interesting if merging 18 desktops into a
> processing engine to support the computing needs of our business while
> lowering the cost of expanding the number of terminals in the business as we
> add employees.
>
>
>
> On Fri, Apr 22, 2011 at 12:22 AM, Ron Johnson<ron.l.johnson@cox.net> wrote:
>
>> On 04/22/2011 12:06 AM, Clint Billedeaux wrote:
>>
>>> Does anyone on this list know enough about Clusters and Thin Clients to
>>> really discuss it? Or am I just going to have to go experiment with it
>>> blindly? That's a trick question. I'm going to experiment blindly
>>> regardless of what I hear. It would just be nice if I could get some
>>> encouragement from someone who has some experience with either or both.
>>>
>>>
>> Since clusters and TCs are solutions to sufficiently different problems
>> that they don't overlap very much, please tell us your goals.
>>

-- 
"Neither the wisest constitution nor the wisest laws will secure
the liberty and happiness of a people whose manners are universally
corrupt."
Samuel Adams, essay in The Public Advertiser, 1749
___________________
Nolug mailing list
nolug@nolug.org