Grid computing needs and education

[darose-prQxUZoa2zOsTnJN9+BGXg at public.gmane.org (David Rosenstrauch)]

Scott R. Ehrlich wrote:
> I know grid computing takes advantage of unused CPU cycles in other 
> people's systems.


Not meaning to be pedantic, but there's actually various types of grid 
computing - compute grids, storage grids, caching cluster, hosting 
cluster, etc.

I guess I would loosely define grid computing in general as pooling the 
resources of multiple computers in order to accomplish a large task that 
would be costly/lengthy/impossible on a single computer.  However, what 
the resource (or resources) are that are being "grid-ified" can vary 
depending on the type of grid.

The type of grid you're describing is generally termed a "compute grid" 
- i.e., a grid whose purpose is to garner a large amount of processing 
("compute") cycles.


> I have a network of heterogenious systems (Macs, Windows, Linux - Debian, 
> CentOS, Fedora) I'd like to utilize grid software on.  What is the best 
> way to approach this?  What grid software do people recommend?  How do I 
> configure the subnetting (the systems traverse at least three subnets), 
> and some are on a dhcp network with no corresponding hostnames - IPs 
> only).
> 
> Thanks.
> 
> Scott


Maybe try looking into Globus, and also Beowulf 
(http://en.wikipedia.org/wiki/Beowulf_(computing)).  You could also take 
a look at the Berkeley/BOINC stuff too - perhaps it's possible to create 
your own private cluster to run tasks on.

HTH,

DR