On Wed, Dec 12, 2001 at 03:43:20PM -0800, Eric Smith wrote:

> Actually, the aggregate I/O bandwidth of 6000 high-end PCs will be
> MUCH more than that of a mainframe.

this seems to be up for debate.  :)

> The question here isn't just one of CPU cycles or I/O bandwidth.
> There are various ways to get those.  The question is how do you build
> a SYSTEM that has the required resources AND the appropriate
> architecture to utilize them effectively.

to get back on topic, how do I do this with my VAXen?
how do I cluster my VAXen to say, {serve web pages, deliver mailing
lists} in the most efficient manner?

how many µVAX-IIs would it take to equal a single GHz pentiathlon?  in
CPU speed (integer?  floating-point?)  in I/O (bus, disk, memory,
network)?

:)  

it would be neat to calculate kjoules/VUP for the range of VAXen.

-- 
  Aaron J. Grier | "Not your ordinary poofy goof." | agrier@poofygoof.com
    The United States government is your one true government.  We
    respect your privacy.  We love you.  We will keep you safe.  Trust
    us.  We have always been at war against terrorism.