[Bizgres-general] Sizing a server to host Bizgres...

Charly Clairmont charly.clairmont at altic.org
Tue May 2 23:45:18 UTC 2006


Hi,

First I must thanks all of you for your sugestions. I improve my disk
performance and the results are the following :

# hdparm -tT /dev/hda

/dev/hda:
 Timing cached reads:   2552 MB in  2.00 seconds = 1275.92 MB/sec
 Timing buffered disk reads:  102 MB in  3.03 seconds =  33.68 MB/sec

# time dd if=/dev/zero of=bigfile bs=8k count=524288
524288+0 enregistrements lus (read).
524288+0 enregistrements écrits (written).

real    2m11.164s
user    0m0.108s
sys     0m16.113s

# time dd if=bigfile of=/dev/null bs=8k
524288+0 enregistrements lus (read).
524288+0 enregistrements écrits (written).

real    2m44.466s
user    0m0.124s
sys     0m4.576s


So I just retry some queries on my Bizgres and see if it is faster... I
speak to you soon about it 

Thanks a lot...

-- 
Charly CLAIRMONT 
________________________________________________________________________
CTO
ALTIC, une ALTernative à  votre système d'Information et de
Communication
mail : charly.clairmont at altic.org
mob : +33(0)661736013
tel : +33(0)870286663
site : www.altic.org 

Le mardi 02 mai 2006 à 09:56 -0700, Luke Lonergan a écrit :
> Yes, and do you remember the days on Windows when enabling dma was an
> experiment to be tried with some fear of data loss, etc?  That’s why
> RedHat was conservative, but that was 8 years ago. 16 bit IDE?
> Pullease! Multi-sector PIO transfers, it’s been safe for 10 years.
> Interrupt handling was an afterthought on desktop drives about 10
> years ago, it’s been solid over the last 8.
> 
> BTW, none of these problems happen on SATA drives because they use the
> Linux SCSI layer.
> 
> - Luke
> 
> On 5/2/06 9:44 AM, "Steve Atkins" <steve at blighty.com> wrote:
> 
>         -d1 - use dma
>         -c1 - use 32 bit IDE support
>         -m16 - burst transfer multiple sectors
>         -u1 - service interrupts during disk interrupt processing
>         -t and -T - time cache and device reads.
> 
> _______________________________________________
> Bizgres-general mailing list
> Bizgres-general at pgfoundry.org
> http://pgfoundry.org/mailman/listinfo/bizgres-general



More information about the Bizgres-general mailing list