strange performance degrade on 3Ware
- From: Yakov Hrebtov <hrebtov@xxxxxxxxx>
- Date: Tue, 31 Oct 2006 13:01:08 +0500
Hello!
I'm building server using 3Ware 9550SX-4LP SATA RAID controller.
OS is FC5.
I have noticed strange performance degrade, when writing to the partition of raid device compared to writing to device itself.
Let me explain.
There is only one partition created (/dev/sda1) on my raid device (/dev/sda).
I decided to measure write speed to this partition, using dd command.
dd if=/dev/zero of=/dev/sda1 bs=1M seek=100 count=4096
Writing 4G of zeros to /dev/sda1 takes about 64 seconds, so write speed is about 67Mb/sec.
Then I ran this test on raid device itself (/dev/sda):
dd if=/dev/zero of=/dev/sda bs=1M seek=100 count=4096
In this case, dd finishes in 34 seconds, write speed is about 124 M/sec!
(I repeated tests several times in both cases)
What causes so huge difference?
If I do similar write test on regular hard drive attached to this system (/dev/sdc), then I have not measured any performance difference in writing to /dev/sdc compared with writing to /dev/sdc1.
Kernel version is 2.6.17-1.2187_FC5
I'm ready to provide other info, that you need to help me to solve this.
Thank you!
--
fedora-list mailing list
fedora-list@xxxxxxxxxx
To unsubscribe: https://www.redhat.com/mailman/listinfo/fedora-list
- Prev by Date: Re: e2fsck: Bad magic number in super-block
- Next by Date: Re: FC6 fuse problem
- Previous by thread: Re: (fedora) Re: Cisco VPN Client on FC6 failed, missing config.h
- Next by thread: fc6, livna and 'spurious' totem-xine update
- Index(es):
Relevant Pages
- Re: Does FAL problem exist in WinCE 5.0?????
... I was perfoming partition driver test on NAND flash.The partition ... >>
Writing known data to partition 0x87db123a ... the compactor can't be started ...
>> sectors in a critical compaction stage. ... (microsoft.public.windowsce.platbuilder) - Re: Does FAL problem exist in WinCE 5.0?????
... Override common setting to enable automatic formatting of unformatted ... I
was perfoming partition driver test on NAND flash.The partition ... >>> Formatting
store 0x27c4435a ... >>> Writing known data to partition 0x87db123a ...
(microsoft.public.windowsce.platbuilder) - Does FAL problem exist in WinCE 5.0?????
... I was perfoming partition driver test on NAND flash.The partition driver ...
Requesting partition "TestPart1" of size 7895 sectors ... Writing known data
to partition 0x87db123a ... (microsoft.public.windowsce.platbuilder) - Re: Does FAL problem exist in WinCE 5.0?????
... I was perfoming partition driver test on NAND flash.The partition driver ...
> Requesting partition "TestPart1" of size 7895 sectors ... > Writing
known data to partition 0x87db123a ... (microsoft.public.windowsce.platbuilder) - Re: CHS - LBA problem
... > When i use parted to write partition tables on my disks ... writing a defined
partition table does not solve it. ... necessary when dealing with "large" drives
(back in the 20th century, ... Writing them is worse. ... (alt.os.linux.suse)