Subject Linux server optimization for FB 1.5 SS? Best practice guide?
Author Myles Wakeham
I have a new FB 1.5.6 Linux Super Server installation going online.
This will eventually be upgraded to 2.1 or 2.5 but we have legacy code
that keeps us in 1.5.6 for now.

I have installed a 'default' build of FB 1.5.6 on CentOS 64 bit server,
and its running as a virtual machine on OpenVZ/Proxmox. The
configuration of the server is like this:

Dell PowerEdge 2950, Dual Xeon 2.6Ghz CPUs (2 core per CPU I believe)
RAID 5, with 3 x 1TB drives, 1 x 1TB Parity, and 1 Hot spare
16GB of RAM total for server, with 6GB allocated to CentOS for FB Server

The database feeds a web application running on a separate box.

I have not applied any changes to the default firebird.conf setting.

The server, under light load (about 8 users), is reporting a higher than
normal CPU load, but almost no RAM load.

The most intensive queries involve a lot of row retrieval, sorting and
display, and I do not believe I have optimized this for sorting at all.
Nor have I optimized for CPU affinity (which I'm not even sure I can
do with SS 1.5.6 on Linux). I have also not optimized Cache size, and
the DB is reporting page size at 16384.

So can anyone give me a suggestion as to where to start to get more
performance out of this beast? I know that FB upgrade would help a lot,
but I can't address that in the next few months at least so I have to
work with what I have got right now. I'm seeing that simple queries
appear to be pretty fast, but the more complex ones that are already
highly optimized are loading the server.

Any thoughts?

Myles Wakeham
Director of Engineering
Tech Solutions USA LLC
Phone +1-480-451-7440