On Mon, 11 Aug 2003, Dennis Wicks wrote:

> Do you have some real numbers to back up that claim?
>
> Many people make the mistake of comparing the one-time-cost
> of a programmer changing a program to the recurring cost of
> hardware upgrades. There may be installation charges and there
> will most certainly be an increase in MMC for the foreseeable
> future.
>
> Also, once you have your programmers trained, (they still do
> that don't they?) they will not write any "bad" programs from
> then on out, but if you just throw hardware at the problem it
> is still a problem and will likely increase in severity as the
> "bad" programming practices proliferate throughout the system.
>

More generally, poor performance can arise because of poor code, but
there are other possibilities:

Hardware too modest for the job chosen.

Insufficient testing. Who here hasn't been caught out developing on the
Big Thumper, the deploying a working application to a less well-endowed
machine to find further find tuning is needed?

Unnecessary bells & whistles. This Pentium II is well able to serve my
office: I know because it's more powerful than the machine that does the
job. What kills it is the Red Hat's idea I have to have KDE or Gnome on
it.


Whether throwing more hardware at the problem is a question whose answer
is "It depends."

In my case, the performance problem can be fixed with (very few) tens of
Currency Units. However, if the heavyweight eye-candy were replaced
with something more modest, the cost savings would be tens of millions
of CUs over a million of these servers.

However, on an xBox where RAM is measured in Gigabytes, probably nobody
would notice the overheads.


--


Cheers
John.

Join the "Linux Support by Small Businesses" list at
http://mail.computerdatasafe.com.au/mailman/listinfo/lssb
Copyright John Summerfield. Reproduction prohibited.

Reply via email to