Fwd: Re: Room for Collections

From: Absurdly Obtuse <vance_at_ikickass.org>
Date: Wed Aug 22 09:47:32 2001

On Wed, 22 Aug 2001, Kent Borg wrote:

> On Tue, Aug 21, 2001 at 06:10:37PM -0400, Master of all that Sucks wrote:
> > I think the distinction between supercomputers and mainframes is even
> > deeper than that. Supercomputers are designed to do things very very
> > fast. Mainframes are designed to do many many things at once. The two
> > goals frequently aren't convergent.
>
> Very much so. Maybe a supercomputer is to a dragster as a mainframe
> is to an 18-wheeler. Both might have roughly the same magnatude
> engine with the dragster doing somewhat more power (I am making this
> up--don't fault me on the specifics), but the intended uses are very
> different.

Your analogy is quite good. If we continue this analogy, we can say that
your proverbial dragster has a lot more horsepower than the
tractor-trailer, and the semi has a lot more torque than the dragster.
You could force a supercomputer to act like a mainframe, or vice versa,
but it wouldn't work very well.

> > Mainframes are very good at serving thousands, if not hundreds of
> > thousands of *simultaneous* transactions.
>
> Yes, though earlier mainframes were doing batch jobs with cards and
> later mag tape. But always mainframes are big iron for business and
> that makes them very different from the shortlived big thoroughbred
> iron of supercomputers.

They were still sitting there chugging away doing many simultaneous (not
very quickly) with data obtained from mag tape or punched card. There
were, of course exceptions to this paradigm, but I still think it's a good
way of looking at it.

> -kb, the micro-oriented Kent who used to be puzzled over what made
> mainframes different, and though he still doesn't know that much about
> mainframe architecture, he at least understands their function is
> quite different.
>

Peace... Sridhar
Received on Wed Aug 22 2001 - 09:47:32 BST