Thanks Dmitry,

You are right. The performance on virtual machine is slower than on real
hardware, but in my case it was the opposite. FB on
VirtualBox(WindowsXP) was working faster than running on the host.

The task in beginning was simple. My friend told me, that his
application is working slow, working on real hardware. On this
hardware(as host) i've installed virtualbox and windows xp as guest and
when i tried, the application was working faster on virtual pc than on
the host. The VirtualBox HDD file was on the same array, as database on
real hardware. I was surprised and because of that, I began these tests,
but now with the IB test DB on virtual machines.

Then I've changed FB to classic, i've tried with compiled FB from me.
Then I start tests with virtual machines, 32bit, 64bit OSes, different
filesystems(jfs,xfs), even i've tried with ramdisk. I've tried with
different page sizes, dbcache pages, forcewrite - on/off. Every time the
result was that FB on Windows is better. May be, that is the true.

Thanks,
Ivaylo M. Ivanov

On 07.4.2012 ?. 00:29, Dmitry Kuzmenko wrote:
>
> Hello, Ivaylo!
>
> Friday, April 6, 2012, 5:48:51 PM, you wrote:
>
> II> I have a strange performance problem with fb 2.5.X. My friend told me
> II> that a program that he uses is working faster when database is on
> II> Windows XP compared to database on Linux. I didn't believe and did
> some
> II> tests. I can't explain the results to myself. For the tests I've used
>
> mostly performance depends on disk and processor. disk drivers
> on Windows can be better or worse on Linux, resulting better or worse
> performance. No surprise.
>
> II> wait to finish. The tests were made on different virtual machine on
> II> VirtualBox Host. The host is Intel i3 2100T with 4GB DDR3 RAM and
> raid 5
> II> array(linux software raid). Every virtual machine i've created was
> with
> II> similar parameters. 1GB RAM, 8GB HDD The results are:
>
> I'm very suspicious about test on virtual machines, especially
> VirtualBox. Again, performance depends on implementation of layer
> between virtual machine, used operating system, and hardware.
>
> By my tests performance on VMWare Desktop and Virtual PC is 1.5-2
> times worse than on pure hardware, mostly on disk performance.
>
> So, I consider your tests unclear.
>
> --
> Dmitry Kuzmenko, www.ib-aid.com
>
>



[Non-text portions of this message have been removed]