On the subject of VMs and Firebird installations...

I have about 15 production servers running that I am responsible for.
About 85% of them are Linux (mainly CentOS & Fedora) and the rest are
Windows. All run on VMWare ESXi 3.5. All have their own local Firebird
Super Servers (and there's a dedicated database server in there as well
on VMWare).

Why did we move EVERYTHING to VMWare? Well the main reason is that in
the case of a disaster, I can simply restore a VM disk image on a
different hardware server, and it will just come up. I can't do that
easily with Windows if the hardware that I'm restoring to is radically
different to the hardware that died. The reconfiguration headaches, BSOD
issues, etc. just is a risk I can't afford to take in a stressed out
production world. Much is the same with Linux. Restoring to a new
hardware system with disks with different device references can be a
huge pain. We do use MondoArchive for backup which helps, but I can't
guarantee that the hardware for the destination server will match
exactly to the hardware from the source, so its going to be difficult no
matter what happens.

And then there is the issue of efficiency. If I can run 8 servers on
one 'maxd out' hardware box, its just better for me. Better on my
wallet, better on my power consumption, etc. My servers are colocated
so my data center gives me a limited ration of power for the cost of the
colocation, so power is an issue. By leveraging multiple servers per
hardware box, I have a workable and cost effective solution.

This is all being done with VMWare ESXi 3.5. We backup using the
ghettoVCB scripts (open source) to NFS servers.

Performance actually improved over dedicated hardware. I suspect this
has something to do with optimized drivers for the virtualization
environment, that just seemed to be more efficient. We store our
Firebird data on either local SATA disks or via iSCSI and although the
iSCSI target is connected on its own LAN and is a bit slower than local
disk, its still very efficient

Of course everyone's mileage may vary. The one thing I would use as a
decent test is to time your recovery from a dead hard disk or dead disk
array with or without VMWare. I don't use the commercial VMotion
solutions etc. because they are very expensive. But even with our
configuration, my restore time is under 1 hour. I've had to do this a
few times, and using the right tools for this makes it a breeze.

When I put all of this together, I blogged about the whole thing in
detail. If you are considering moving to VMWare environment, this might
help.

http://www.techsolusa.com/ts_vmware_experience.html

Hope this helps someone out. From our perspective, the ultimate server
platform on top of VMWare ESXi 3.5 is a CentOS 5.3 Linux distro, Apache
2, PHP5.2.9 and Firebird Super Server. This configuration has become
our 'standard' server footprint, and I can run about 8 of them per
hardware box without raising a sweat on it. CPU performance shows
consistent at 50% or under even with this number of VMs on a box. We
typically install 12-16 GB of RAM per server for this and run dual Intel
CPUs. My preferred hardware is SuperMicro systems, but we also have
Dell servers and they have been pretty reliable. We had a Dell server
have a power supply die on it once, and replacing that was a pain
(required Dell specific parts), but still we just moved the VMs from
that server to another, fired up the 2nd box and the VMs just 'came up'.
No issue, and no real downtime to speak of.

Good luck.
Myles

--
=======================
Myles Wakeham
Director of Engineering
Tech Solutions USA, Inc.
Scottsdale, Arizona USA
http://www.techsolusa.com
Phone +1-480-451-7440