Embedded Perl VS usertag comparision results

[George Law <glaw@hammer.thebook.com>]

Hi Folks -

First off, thanks for the answers which steered me towards using a
UserTag instead of a GlobalSub!

I am working with a minivend database (no sql) catalog  of 1926 items.
This catalog is strictly search driven - the user enters their search
criteria - there is no way to browse the catalog.  

We started out with about 10k of embedded perl in the search results
page to parse out incorrect matches and display a list of confirmed matches.

We time some tests from the click of Submit, to when the results were
fully displayed.  One of our test searches took 26 seconds - another 
took 108 seconds.  During this time the CPU load climbs through the roof!!!

I have managed to get identical output replacing the embedded perl with a
usertag.  Searches can be done on artist, title, product id, category and
style - so far the usertab only contains the artist logic- so there will be
some additional code which may increase my preliminary results a few seconds,
but the results using the Usertag over the embedded Perl are staggering.

The embedded perl search that took 26 seconds took 3 with the usertab,
the 108 second search took 8 seconds.  during the 108 second search the
server load climbed to as high as 1.97 from .20 and was using 49% of the cpu - 
The usertab search used the same % of the CPU - but completed in 8 seconds 
and only caused the load to increase .04 - the server didn't even blink.


I had read the the Safe.pm checks really affected the performance of embedded
perl - but never imagined I would see such dramatic increases in the use
of Usertags.

Cordially,

                                          George
____________________________________________________________________________
Customer Service              		            Software Workshop Inc.
glaw@thebook.com 315.635-1968 (x213)             "software that fits!" (TM)
http://www.thebook.com