Hi Everyone
I transfered a perl script driven web site ( not mod-perl ) to a new server.
On the old server the performance where quite good and on the new one
they sucks big time.
Old Server is:
CPU: Intel(R) Pentium(R) 4 CPU 2.00GHz with 512KB cache
MEM: 1GB DDR ( i think 200mhz )
On Thu, Sep 29, 2005 at 12:22:31PM +0300, Michael Ben-Nes wrote:
Hi Everyone
I transfered a perl script driven web site ( not mod-perl ) to a new server.
On the old server the performance where quite good and on the new one
they sucks big time.
Old Server is:
CPU: Intel(R)
Ummm Sorry for pointing the obvious... but:
Old server has 10K U1600 SCSI RAID, while the new one has normal IDEs
with a software RAID 1. (RAID1 is known to reduce performance)
Second, did you check that the IDE drivers are running at UDMA133/32bit
mode? (hdparm /dev/hdX)
Gilboa
On Thu,
On Thursday, 29 בSeptember 2005 12:22, Michael Ben-Nes wrote:
Its not the Disks as i made few tests that where successful ( under load
) and the respond was very fast.
I beg to differ. The single biggest difference between the two
servers is:
DISKS: HW SCSI RAID ( mirror ) U160 10K RPM
Versus
Yes, both using using_dma and 32bit
I also mentioned that there is no CPU wait, so i assume the Disks do
their jobs quite well
Cheers
Gilboa Davara wrote:
Ummm Sorry for pointing the obvious... but:
Old server has 10K U1600 SCSI RAID, while the new one has normal IDEs
with a software
But the the CPU user was not on 100% but on 40% the wait as on 60%
I was also able to duplicate a file ( 700mb ) quite fast under the load.
this prooved me that the disks are not the issue.
Unless the software raid is very hungry when it comes to kernel
calculation ? though i would expect
Here is the output of dstat:
http://www.canaan.co.il/users/miki/dstat.html
By the way, I really like dstat :)
Michael Ben-Nes wrote:
Hi Everyone
I transfered a perl script driven web site ( not mod-perl ) to a new
server.
On the old server the performance where quite good and on the