On Dec 10, 2003, at 4:02 PM, R. Joseph Newton wrote: [..]
Most CPUs in use average about 99% idle time, at least on the computers [some running up to 20 open windows] on which I have checked these stats.
Not wishing to get us bogged down in a convention of the IEEE Transactions of Distributed Processing are you talking here in terms of the average 'desk top' unit that most persons have access to? Back End headless servers in dedicated database services mode? yada-yada-yada...
{ forgive me, I was traumatized by major data center issues
at a formative age, where optimizing CPU utilization on
crays was an Imperative...
But my therapist says I'm getting better...8-}To me, the more important issues in normal practice have to do with comprehensibility, and thus maintainability, than with technical performance comparisons.
Is that 'normal' as in 'bell curve shaped' or as in 'normalized data', as in DB speak? 8-)
I think a part of the underlying set of questions here are when is it time to move one's data sets out of the simpler perl tiehash with a generic DB_FILE and off into a 'real database' - such as Postgres, as Randall Schwartz recently noted some of the newer benchmark numbers. The simplest answer is
when the DB portion of the process is the log_jam
Which then takes us back into what I wish to underscore in R. Joseph's point,
IF one has written modular, and
maintainable, DB interface code
Then
one simply opts to change the guts
on the inside of a Module.If one has built it in such a way that one already is using a 'network query' interface, where one has already abstracted the database semantics, one can do this without changing anything on the 'client side' of that code.
ciao drieux
---
-- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] <http://learn.perl.org/> <http://learn.perl.org/first-response>
