On 05 Jul 2004, at 11:07, Ugo Cei wrote:
IndeedIl giorno 05/lug/04, alle 00:39, Yves Vindevogel ha scritto:
There's a big postgresql database behind it, but the database will only be queried from time to time.
Most of the work is done by generating XML out-of that database and storing it into Cocoon's folders.
Do you mean that generation of the XML files will be done offline, in batch and Cocoon's job will only be to take XML files from the filesystem, possibly transform them and serialize the results?
Reason is simple: most reports are timebased. So, when you demand the report for eg. january, there will be no difference once we are february. So there's no need to query the database, there won't be any changes.
If so, performance and scalability will depend on the kind of transformations you are going to do, but given that source files are regenerated only from time to time, you could get a high cache hit ratio: most of the time, the output document will be retrieved from the cache in a matter of microseconds.Simple xml to html (generating tables mainly) and xml to xls (for download) and xml to pdf (using fo, also for download).
I may even consider to physically create the file instead of "on demand" in case of xsl and pdf if that takes the server too long.
You could even use Cocoon to pregenerate the entire site using the CLI and serve a static snapshot from Apache or some other web server.I know :-))
But not with Apache. There are logins involved and some dynamic queries. When I ask mypage.html, that page will be different according to your login.
App is completely running on Slackware, with "personal" configuration, that is, no gui, no overhead, just the basics.
Server will be a Compaq DL 380, 1 mb (or 2 when needed) ram, full scsi3 in raid 5, 3x 320u 72.8 HD.
I suppose you mean 1GB RAM.
Yes
Ugo
-- Ugo Cei - http://beblogging.com/
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
