hi dario, to clear up some possible misunderstandings and naming issues: we have two different beasts here. one is the number of threads that zope uses while it runs. this can be tweaked by invoking the start script with an argument like "t <number_of_threads>" or by editing z2.py and changing the default number of 4 threads. again, these are application threads. the database connections that were talked about are a different issue. these refer to the number of connections zope creates to its underlying ZODB object store. you can see them if you go to <myzope>/Control_Panel/manage_debug in Zope < 2.2.0 or by using the link in the control panel in Zope > 2.2.0. the number of these connections is set in <zope>/lib/python/ZODB/DB.py and the default is 7. in general, when a thread attempts to do anything with zope it needs to get hold of a ZODB connection so it can read or otherwise manipulate objects in the ZODB. this is why the number of application threads should be smaller than the number of ZODB connections. tweaking these numbers is possible, but mileage may vary depending on your specific site and usage. jens ---- Jens Vagelpohl jens@digicool.com Software Engineer www.digicool.com Digital Creations (888) 344-4332 Got Zope? ----

-----Original Message----- From: zope-admin@zope.org [mailto:zope-admin@zope.org]On Behalf Of Dario Lopez-Kästen Sent: Friday, August 11, 2000 07:10 To: zope@zope.org Subject: [Zope] Defining High Load in Heavily DB Driven site (was: [Zope] zope2.2.0 and what is high load)

Hello!

I need to find more info on how Zope handles threading with relation to database connections.

I am currently considering and evaluating Zope as one of the options we have to build a really large, completely databasedriven "enterprise scale" web-platform. I am a bit worried about this "maximum of 7 threads per db connection" limitation mentioned previously, and I need to know a) what does this mean in terms of accesibility, b) how does it affect performace, and c) how does Zope work (in detail) with external database connections.

We are going to be using Oracle as our DB backend, all our served data will be database *only* (some of it will be fairly large), and efficient database connections is *crucial* to what we intend to do.

We are going to have between 10-20k users at most and I expect the maximun of simultaneous connections to be beween 400 - 900 under heavy load; "normal load" will probably oscillate around 25-200 simultaneous connections (these numbers are crude guesses, but take into account increased usage over time as more services are provided in the future).

Bandwith problems are not an issue, nor is computing power.

I would appreciate any pointers to documentation or people to ask.

Sincerely,

/dario