| View previous topic :: View next topic |
| Author |
Message |
Tzuriel
Apprentice
Joined: 01 Jun 2004
Posts: 260
|
|
| Back to top |
|
 |
wmgoree
Apprentice
Joined: 08 Aug 2003
Posts: 246
Location: Alexandria, VA
|
 Posted: Thu Jun 02, 2005 3:10 am Post subject: Eh? |
 |
|
He's complaining about 1378 and 34? That sounds fine to me for a PowerEdge 1750.
I mean, I've got as much affection for my PowerEdges as the next guy, but the fact is these are "value" workhorses more than high-performance systems. 1378 / 34 is still twice as fast as any ATA drive you'll find. You could probably bring it up *some* by tweaking hdparm (incidentally, it's "hdparm" not "hdparam") or by getting new firmware for your SCSI controller, but at some point you have to ask if it's worth it. Unless you're doing hardcore journal-on-a-different-block-device kind of fine tuning you probably wouldn't notice much improvement.
_________________
vi? *snicker* it doesn't even include a mail reader... |
|
| Back to top |
|
|
Tzuriel
Apprentice
Joined: 01 Jun 2004
Posts: 260
|
| Posted: Thu Jun 02, 2005 3:38 pm Post subject: |
|
|
| Quote: | The first test line seems fine ... the second line is a bit low... did you try running this test a couple of times?
EDIT: Actually, your test are quite low...especially the second one. I am using ATA133 and I get about 2x the speed that you get for the second
line. Your first line out-does my hdd by about 200 MB/sec.
Considering that you have SCSI, you have very low speeds...I have never worked with SCSI, so I am not sure about what steps you can take to incresae thier speed, but if SCSI has DMA, then you should think about enabling it (if it is not already). |
Yeah, I see your point if it only brings it up minimally. I just thought there might be a huge performance gain to work for since it seems normal drives are doing better than this. |
|
| Back to top |
|
|
overkll
Veteran
Joined: 21 Sep 2004
Posts: 1249
Location: Austin, Texas
|
| Posted: Mon Jun 06, 2005 7:52 pm Post subject: Re: Eh? |
|
|
| wmgoree wrote: | | He's complaining about 1378 and 34? That sounds fine to me for a PowerEdge 1750. |
wmgoree,
Complaining, no. Puzzled, yes! tzuriel's drive is a U320, 10K rpm SCSI drive. Maybe this system was priced very good, but usually SCSI drives are more expensive because they are deemed to be faster (seek times and thoughput), more reliable and have better waranties than PATA and SATA.
| Quote: | | ...1378 / 34 is still twice as fast as any ATA drive you'll find. You could probably bring it up *some* by tweaking hdparm... |
Oh contraire! I have ATA/UDMA100 drives that post faster than that!
The first hdparm test is "Timing cache reads". From what I can tell, this is more a function of available memory bandwith/frontside bus. The second test is "Timing buffer disk reads" and is a good indicator of your drive's actual performace. I know that hdparm isn't perfect, but the second test gives you a decent baseline to judge the drive's speed. There are more accurate benchmarking tools. IMHO, Hdparm is more of a quick and dirty check.
I've got 4 AMD based Gentoo systems with diffenent hardware and kernels. I ran "hdparm" on all of them. The results make me wonder why an expensive 10K U320 SCSI drive isn't faster. For the record, I am NOT trying to prove that AMD is better than Intel. I only have one Intel box, and it's running Windows ME. BTW, I don't use any hdparm "tweaks" on these machines aside from making sure they are running in DMA mode. This is by no means a scientific test.
AthlonXP 2800 (2Ghz), ATA/UDMA133, 512mb ram
OS = Gentoo, 2.4 series kernel
| Quote: | /dev/hde:
Timing cached reads: 1968 MB in 2.00 seconds = 984.00 MB/sec
Timing buffered disk reads: 122 MB in 3.04 seconds = 40.13 MB/sec |
Dual AthlonMP 1800 (2x 1.4Ghz), ATA/UDMA100 768mb ram
OS = Gentoo, 2.4 series kernel
| Quote: | /dev/hda1:
Timing cached reads: 1068 MB in 2.00 seconds = 534.00 MB/sec
Timing buffered disk reads: 138 MB in 3.03 seconds = 45.54 MB/sec |
Athlon64 3000+ (1.8Ghz, 939 pin), SATA150, 512mb ram
OS = Gentoo, x86 32bit mode, 2.6 series kernel
| Quote: | /dev/sda3:
Timing cached reads: 3228 MB in 2.00 seconds = 1613.44 MB/sec
Timing buffered disk reads: 184 MB in 3.00 seconds = 61.32 MB/sec |
Athlon64 3500+ (2.2Ghz, 939 pin), !Gb ram
OS = Gentoo, x86_64 64bit mode, 2.6 series kernel
| Quote: | ATA/UDMA133
/dev/hde3:
Timing cached reads: 4560 MB in 2.00 seconds = 2278.07 MB/sec
Timing buffered disk reads: 166 MB in 3.03 seconds = 54.76 MB/sec
SATA150
/dev/sda3:
Timing cached reads: 4588 MB in 2.00 seconds = 2294.35 MB/sec
Timing buffered disk reads: 170 MB in 3.01 seconds = 56.41 MB/sec
And just for fun!
RAID5, 2x SATA150 + 1 ATA/UDMA133 ( only 2 onboard SATA ports )
/dev/md2:
Timing cached reads: 4584 MB in 2.00 seconds = 2290.06 MB/sec
Timing buffered disk reads: 320 MB in 3.01 seconds = 106.22 MB/sec |
So, IMHO either SCSI doesn't perform as advertised (not likely) or something is preventing that U320 10K SCSI drive from better performance. Maybe hdparm isn't suited to accurately gauge the performance of SCSI drives?
I may be wrong, and if I am, I welcome anyone to straighten me out in this issue. 34mb/sec for a 10K rpm U320 SCSI drive just seems too low to me.
Last edited by overkll on Mon Jun 06, 2005 11:49 pm; edited 2 times in total |
|
| Back to top |
|
|
plut0
Apprentice
Joined: 21 Dec 2004
Posts: 272
|
| Posted: Mon Jun 06, 2005 8:22 pm Post subject: |
|
|
I like to use dd for read/write, test for 1gb file.
time dd if=/dev/zero of=test bs=1024 count=1048576
| Code: | 1048576+0 records in
1048576+0 records out
real 0m29.917s
user 0m0.469s
sys 0m13.255s |
1000MB / 29.917s = 33.425 MB/s write
time dd if=test of=/dev/null bs=1024
| Code: | 1048576+0 records in
1048576+0 records out
real 0m23.620s
user 0m0.299s
sys 0m4.160s |
1000MB / 23.620s = 42.337 MB/s read
That is pretty bad performance you're getting. The one I just tested is a generic 5 year old IDE drive, most likely 7200rpm.
Are you using RAID by chance? |
|
| Back to top |
|
|
overkll
Veteran
Joined: 21 Sep 2004
Posts: 1249
Location: Austin, Texas
|
| Posted: Tue Jun 07, 2005 4:07 am Post subject: |
|
|
Pluto,
Ahhh! Good idea! I'm going to have to try the dd test. Thanks. |
|
| Back to top |
|
|
kash04
n00b
Joined: 08 Nov 2005
Posts: 5
Location: Houston
|
| Posted: Fri Feb 24, 2006 4:44 pm Post subject: |
|
|
did anyone find out anymore info on this? i'm getting the same results as him on my centos box.. maybe it is the poweredge :-\
an ide drive
aries akash # hdparm -tT /dev/hda3
/dev/hda3:
Timing cached reads: 2792 MB in 2.00 seconds = 1396.21 MB/sec
Timing buffered disk reads: 160 MB in 3.02 seconds = 53.01 MB/sec
and my scsi disk
[root@asterisk1 ~]# hdparm -tT /dev/sda3
/dev/sda3:
Timing cached reads: 2564 MB in 2.00 seconds = 1280.27 MB/sec
Timing buffered disk reads: 142 MB in 3.00 seconds = 47.29 MB/sec |
|
| Back to top |
|
|
tomatopi
Tux's lil' helper
Joined: 18 Sep 2005
Posts: 130
Location: Ottawa
|
| Posted: Sat Feb 25, 2006 11:41 am Post subject: |
|
|
| I have lots of servers with SCSI drives and hdparam always seems to show really low numbers when testing compared to ATA/SATA drives. It often shows me lower numbers on SCSI drives than old ATA laptop drives I test. I think SCSI excels more at low-overhead and high-transaction load than raw read-write speeds which is what hdparm seems to test. Overall, all the servers run smooth as silk even with supposidly low I/O rates. |
|
| Back to top |
|
|
plonka2000
Apprentice
Joined: 02 Dec 2004
Posts: 160
Location: UK, Surrey
|
| Posted: Sat Feb 25, 2006 8:45 pm Post subject: |
|
|
Hi all,
Interesting thread regarding SCSI and SATA/PATA performance.
I have a bit of a cunnundrum of my own though, as I wanted to tweak my hdparm settings for my SATA drives but am unable to it seems.
Can anyone explain why I get errors with hdparm and my SATA drives?
I cant get hdparm settings from them. Is this normal?
Its running on an a8n-sli deluxe motherboard (AMD64) on a Silicon Image 3114 controller.
All drives are Maxtor Diamondmax 9 SATA 160GB.
Here are my outputs:
| Code: | hdparm -tT /dev/sda /dev/sdb /dev/sdc /dev/sdd
/dev/sda:
Timing cached reads: 3620 MB in 2.00 seconds = 1810.28 MB/sec
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Inappropriate ioctl for device
Timing buffered disk reads: 170 MB in 3.01 seconds = 56.47 MB/sec
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Inappropriate ioctl for device
/dev/sdb:
Timing cached reads: 3640 MB in 2.00 seconds = 1819.37 MB/sec
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Inappropriate ioctl for device
Timing buffered disk reads: 170 MB in 3.01 seconds = 56.56 MB/sec
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Inappropriate ioctl for device
/dev/sdc:
Timing cached reads: 3652 MB in 2.00 seconds = 1825.37 MB/sec
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Inappropriate ioctl for device
Timing buffered disk reads: 170 MB in 3.01 seconds = 56.51 MB/sec
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Inappropriate ioctl for device
/dev/sdd:
Timing cached reads: 3628 MB in 2.00 seconds = 1811.56 MB/sec
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Inappropriate ioctl for device
Timing buffered disk reads: 172 MB in 3.03 seconds = 56.72 MB/sec
HDIO_DRIVE_CMD(null) (wait for flush complete) failed: Inappropriate ioctl for device |
| Code: | hdparm /dev/sda
/dev/sda:
IO_support = 0 (default 16-bit)
readonly = 0 (off)
readahead = 256 (on)
geometry = 19929/255/63, sectors = 163928604672, start = 0 |
| Code: | hdparm -i /dev/sda
/dev/sda:
HDIO_GET_IDENTITY failed: Inappropriate ioctl for device |
_________________
-Do not be afraid of what is different.
-Do not be afraid of being different.
-After all, ignorance is not an excuse.
Using Gentoo and Windows XPee.
Check my site here |
|
| Back to top |
|
|
augury
l33t
Joined: 22 May 2004
Posts: 722
Location: philadelphia
|
| Posted: Sun Feb 26, 2006 4:12 am Post subject: |
|
|
| A scsi w/ performance comparable to sata or even fast ata drives tends to run expensive (DoD expensive). I get about 35 MB/sec on fibre channel 10k cheetahs although this seems to be limited by either the pci bus or fibre interface (raid0 sees no improvement). In comparison I've gotten 2 sata's in raid0 to do just under 120 MB/sec. I got my fibre channel disk for free but it would have cost me 10k used and its close to being retired for being 400lbs. overbuilt. For the same sata space I would pay about $300-400. |
|
| Back to top |
|
|
linuxtuxhellsinki
l33t
Joined: 15 Nov 2004
Posts: 700
Location: Hellsinki
|
| Posted: Sun Feb 26, 2006 6:58 pm Post subject: |
|
|
plonka2000,
Hdparm is not working correctly with sata-drives or at least all the options are not supported.
There's also sdparm for scsi-drives, but the options are different 
_________________
1st use 'Search' & lastly add [Solved] to
the subject of your first post in the thread. |
|
| Back to top |
|
|
plonka2000
Apprentice
Joined: 02 Dec 2004
Posts: 160
Location: UK, Surrey
|
| Posted: Sun Feb 26, 2006 7:04 pm Post subject: |
|
|
| linuxtuxhellsinki wrote: | plonka2000,
Hdparm is not working correctly with sata-drives or at least all the options are not supported.
There's also sdparm for scsi-drives, but the options are different |
Yeah I've been looking into that, and you're correct.
However, I've been wresting with my wireless card for the last few days... *sigh*
...so my hdparm/sdparm issues will be on hold. 
Thanks for the info tho. 
_________________
-Do not be afraid of what is different.
-Do not be afraid of being different.
-After all, ignorance is not an excuse.
Using Gentoo and Windows XPee.
Check my site here |
|
| Back to top |
|
|
|
Kernel & Hardware
