Subject | Re: Multi-processor support with Linux? |
---|---|
Author | David Montgomery |
Post date | 2003-03-12T23:22:34Z |
> > Superserver is definitely my preference.In a nutshell, the performance we've seen with superserver (in
>
> Why's that?
>
production the past 18 months) has outstripped classic (the prior 4
years) by a significant margin in a Windows 2000 deployment. The
comparison has unfortunately been highly informal, and possibly
completely invalid since our last version change in production was
from IB 4.2 to FB 1.0, but running on the same hardware.
As far as typical server load is concerned, we usually have <= 25
concurrent database connections (via a web interface), but usually one
or two of those connections will be doing very large selects or inserts.
Our data is segmented into thousands of separate .gdb files (one .gdb
file for each of our clients), which are being connected/used by our
web server's applications per user page requests. Since every web
page request results in a connect to the client's .gdb, do stuff,
disconnect cycle, the primary performance boost we would receive would
be from the fastest possible connect/disconnect performance.
This said, we are in the process of developing an entirely new
application platform for our products, where we can look at more
advanced ways of managing database requests (like connection pooling)
which at this point I know absolutely ZERO about. Know any good
references I should buy? ;)
Best Regards,
David Montgomery