I like your ideas. I have been pondering something like this for some time (without the funding or time to seriously consider it however). High performance experts typically speak in terms of Giga- and Terra- Flops... which is a serious hurtle for this platform. With no floating point processor onboard, I'm not sure how bad the performance of this platform would be on the benchmarks in the field. Using a simple HZ*processors/sec does not accurately describe the performance of any computer, one without a floating point coprocessor in particular. To make this a more compelling argument (at least to the academic / high performance community you should consider metrics that take into account the poor performance of floating point emulation and slow commodity networks (10/100 networks being one of them).
Your ideas are good, and you certainly have the right idea about this, however.
Computing Systems Research Lab
Apologies for the crossposting, but I'm unsure how many of us have
signed up to the gumstix-wads list yet.
Anyway, it's a lazy Sunday afternoon and I'm bored so I've put together
a rough and poorly researched article on the advantages of gumstix
clustering over traditional methods. It's just a bit of fun and I'm not
really intending to take it much further at this point but I'd be
interested in any comments. It's on my webspace at:
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems? Stop! Download the new AJAX search engine that makes
searching your log files as easy as surfing the web. DOWNLOAD SPLUNK!
gumstix-users mailing list