CGI question

[Pierre Sahores]

Richard MacLemale a écrit :
> 
> I've been using metacard to do cool cgi tricks on my network and web server
> for 2 years now and love, love love it.  However, I'm about to set up this
> new big project dealing with student grades, and am wondering what the
> "best" solution would be.
> 
> This is an OS X Mac with 512 MB of RAM, running apache for web services and
> using metatalk cgi scripts.
> 
> When a student hits a web page, I need to be able to search for and pull
> info out of a 700K text file, to get their class schedule.  The problem is
> that this cgi script may be called simultaneously from 40 different
> computers, at times.  My concern is that I'll end up with 40 copies of the
> script running at the same time, each loading 700K into a variable, which
> will eat RAM.  I've got quite a bit of RAM, but still...
> 
> So what is the best option of the following:
> 
> 1.  Bust the 700K file down into different files, maybe A-L and M-Z, and
> then only hitting the file I need?  (I could make one file per alphabet
> character, but that's not going to be fun to maintain...)
> 
> 2.  Search the file and pull out info without loading it into a variable,
> which would use far less RAM but would take longer?
> 
> 3.  Load the whole file into a variable and not worry about it?
> 
> 4.  Use some type of index file which tells me specifically the location of
> the user in the big 700K file?
> 
> 5.  Some other solution I haven't thought of?
> 
> The thing that is confusing is the speed issue.  The more I load into a
> variable, the faster the script can run, the more memory it uses... But if
> it runs fast enough it will be done quicker and get out of the way of the
> next request.
> 
> Anyone have any thoughts?  As usual, thanks in advance...
> 
> --
> :)
> Richard MacLemale
> Network Administrator
> J. W. Mitchell High School
> 
> _______________________________________________
> metacard mailing list
> metacard at lists.runrev.com
> http://lists.runrev.com/mailman/listinfo/metacard

Hi Richard,

You could use there tree alternative ways to optimise your production
process :

1.- Set up an mc-based long running application server witch will load
once the 700 k flat-file and maintain and serve it from ram (best in
terms of speed and number of concurents connexions). Search the
lists-archives about the way it can be done under unixes. First builded
for linux, my examples are, with some configuration adaptations, usables
under Jaguar, and even Win32, too.

2.- Let your CGI speaks to an SQL databases server to avoid to have 700
k to load but only the needed database records each time the CGI starts.

3.- Combine the two previous ways to get the more secure and fast
solution to use MC datas over the web.
--
Cordialement, Pierre Sahores

Inspection académique de Seine-Saint-Denis.
Applications et bases de données WEB et VPN
Qualifier et produire l'avantage compétitif