Thank you Anderson,
It seems that is file system problem. I will check with XFS !

--- In firebird-support@yahoogroups.com, Anderson Farias <peixedragao@...> wrote:
>
> Hi,
>
> According to this benchmark
>
> http://firebirdsurgeon.blogspot.com/2009/11/firebird-25-benchmark-firebird-shows.html
>
>
> Firebird is a lot slower on Linux boxes.
>
>
> I, as much as you, was surprised by that. This benchmark has cleared for me why my application runs slower on a Linux machine even the hardware being beter than a Windows box.
>
>
> Regards,
> Anderson
>
>  
> --- Em seg, 28/2/11, Kok_BG <kok_bg@...> escreveu:
>
> De: Kok_BG <kok_bg@...>
> Assunto: [firebird-support] Linux vs Windows performance FB 2.1
> Para: firebird-support@yahoogroups.com
> Data: Segunda-feira, 28 de Fevereiro de 2011, 20:22
>
>
>
>
>
>
>
>  
>
>
>
>
>
>
>
>
>
> Hi Folks
>
> I made some tests on windows and linux these days and I was surprised that the linux build is sometime much slower than the windows.
>
> The setup:
>
> Dell Precision Workstation 650 2x Xeon with 533MHz FSB and 512KB L2 @ 2800MHz. Storage: Integrated Ultra 320 SCSI controller with one 36GB HDD @ 10000 rpm. 2 GB RAM.
>
> I used a backup of one database 1.6GB sized.
>
> Windows setup - default SS on windows 7 ultimate;
>
> Linux setup - CS on gentoo 2.6.36-gentoo-r5 #2 SMP
>
> On Linux, firebird was installed via gentoo portage system and was properly complied with the correct CFLAGS in the make.conf
>
>
>
> Have someone the same observation?
>
> I always thought that the Linux is the better solution for FB.
>
>
>
> Some of the measurement:
>
>
>
> 1. restore database - windows was ~10% faster;
>
> 2. Long stored procedure that reads one big table and pumps the data in another table - 30% faster
>
> 3. create shadow - 30% faster;
>
> 4. same check as (2) on the shadowed DB - 240% (!) faster
>
>
>
> It seems that the HDD write speed is the problem. All linux drivers seems correct installed.
>
> Can someone advise - is it normal, or there is a problem with something?
>
>
>
>
>
>
> [Non-text portions of this message have been removed]
>