Subject Re: [ib-support] Multiple Databases?
Author Paul Schmidt

See comments below:

On 14 Mar 2001, at 20:01, s.beames@... wrote:

From: s.beames@...
Date sent: Wed, 14 Mar 2001 20:01:13 -0000
Send reply to:
Subject: [ib-support] Multiple Databases?

> I like some advice on my first c/s db project please.
> I have two separate groups of users of my client program, who will
> maintain separate sets of records in their own databases, but the
> database structure is identical for both groups. One group is a
> mechanical workshop, the other electronics. A few tables, eg client
> details, will be identical for both groups, so it seems sensible to
> share these, rather than have duplicates in two distinct db's, but
> when I looked at splitting these out to their own 'COMMON' db, it
> appears I'll have to lose the referential integrity constraints I had
> established with the other tables. So, question 1/ Can you have R.I.
> constraints between different databases?

You can't have R.I. between two databases, so you either need one
database for both, or two completely seperate databases, trust me you
don't want to go there.

> I suppose I could have one large db, with an extra column in some
> tables to identify each group's data, and rely on the logon to
> identify the group, and have an extra parameter in the SELECTs...... I
> was however considering having the two databases running on two
> separate (currently W95) PC's, to share the load and reduce the db
> sizes. Q2/ Which approach is best? Q3/ Does posting from yahoo-groups
> web page produce HTML posts?

Dump Win95, Win9x is a client O/S not a server O/S, your best bet is
Linux, it was originally designed as a server OS, has better
security, is more efficient, and can be scaled up to some pretty
serious hardware.

As for database size, there are two limitations, disk file size, and
row count, a database can consist of multiple .gdb files, which can
live on different partitions, and even different physical drives.
This can be much easier with a system like Linux that uses a
homogenous file system (all drives are part of the same file

As for row count, a dedicated Unix database server, with a decent
amount of RAM and fast drives (Ultra Wide SCSI II or III prefered),
can handle a pretty big database, even with a relatively anemic CPU.
The key is good indexing and SELECTs that don't return large numbers
of rows with big field counts.

> I'm using Firebird

That's a good combination you can drop Firebird onto a Linux
box, add it to your network, and point your queries to the server
rather then your workstation.

Paul Schmidt,
Tricat Technologies
Email: paul@...