Every table or index is created in the form of OS files therefore max open files need to be set appropriately in order to achieve the larger table count. There would be no limitation for creation of tables in PostgreSQL and performance would be the major criteria as catalogs get overburdened.
You need to ensure OS parameters such as the max open files (fs.file-max if we're talking Linux) is set appropriately. Baring in mind each user will have an open file on each underlying datafile for the databases they're connected to.
THE INFORMATION CONTAINED IN THIS MESSAGE IS CONFIDENTIAL and is intended for the addressee only. If you are not the addressee, any disclosure, reproduction, copying, distribution, or other dissemination or use of this communication is strictly prohibited. Any views or opinions presented are those of the sender and do not necessarily represent Healthcare Software Systems.
HSS Ltd is a company registered in England and Wales with company number 2112370
On 10/04/2013 08:24, Vasilis Ventirozos wrote:
On Wed, Apr 10, 2013 at 10:10 AM, Vedran Krivokuca <vkrivokuca@gmail.com> wrote:
On Wed, Apr 10, 2013 at 9:07 AM, Vedran Krivokuca <vkrivokuca@gmail.com> wrote: > 1) we can go with different instances of PostgreSQL service, let's say > (for pure theory) 10 of them on the same HA cluster setup. Every > instance would hold let's say 1/10th of that big recordset, and around > 3.000 database tables (this apparently shouldn't be of any problem to > PostgreSQL, see below).
I am talking shit here, obviously, excuse me. We would go with, for example, 10 or 100 databases in the same PostgreSQL instance. Buzzwording got better of me. :) The rest of my initial e-mail still applies, I will cherish any input of experience on this subject.
V.
-- Pozdrav/Greetings, Vedran Krivokuća Disclaimer: This message may contain information.
I think that a limitation would be on the OS (filesystem and kernel) because each table
is a file (or more) under a directory (per database). I haven't got experience on how
modern Linux system cope with that but i think it would be something easy to test.
Database wise you shouldn't have a problem.