Skip to contentSkip to conversation

IDE Performance lack !

DISCUSSION Guido Stepken 7 participants 11 messages May 26, 2001 – May 28, 2001

lore
Share & tools
lore
#0 lore raw

Hi !

RedHat 7.1 - IDE IBM 41.1 GIG
Update to 2.4.5 -> noticed, that hdparm -t /dev/hda went down from 10 MByte/sec to 1.9 MByte/sec
Any special Options, beside ide-scsi driver activated ..

Anybody noticed the same problem ? Any clues ?

tnx in advance ...

#1 lore raw
Quoted 8 lines

RedHat 7.1 - IDE IBM 41.1 GIG
Update to 2.4.5 -> noticed, that hdparm -t /dev/hda went down from 10 MByte/sec to 1.9 MByte/sec
Any special Options, beside ide-scsi driver activated ..

Anybody noticed the same problem ? Any clues ?

yes , i am also not happy with IDE performance of Linux . That why i dont use Hard disk in my Target machines ;)

When ever i copy big data (around 400 to 700 MB ) from one partion to another my machine do not response at all (i can not work on another shell) during data transfer.

Thank you ,

Best Regards,

Jaswinder.
--
These are my opinions not 3Di.

#2 lore raw
In reply to Guido Stepken

* Guido Stepken schrieb am 26.05.01 um 22:19 Uhr:

Quoted 10 lines

Hi !

RedHat 7.1 - IDE IBM 41.1 GIG
Update to 2.4.5 -> noticed, that hdparm -t /dev/hda went down from 10 MByte/sec to 1.9 MByte/sec
Any special Options, beside ide-scsi driver activated ..

Anybody noticed the same problem ? Any clues ?

tnx in advance ...

Hi Guido,

This is a 20.1 GB Seagate Barracuda (with hdparm -d1)

homer:~ # hdparm -t /dev/hda

/dev/hda:
Timing buffered disk reads: 64 MB in 2.74 seconds = 23.36 MB/sec
homer:~ #

Did you turn on dma-mode?

-Marc

Signature · 4 lines
+-O . . . o . . . O . . . o . . . O . . .  ___  . . . O . . . o .-+
| Ein neuer Service von Links2Linux.de:   /  o\   RPMs for SuSE   |
| --> PackMan! <-- naeheres unter        |   __|   and  others    |
| http://packman.links2linux.de/ . . . O  \__\  . . . O . . . O . |
#3 lore raw
In reply to Guido Stepken

On Sat, May 26, 2001 at 05:16:42PM -0700, Jaswinder Singh wrote:

When ever i copy big data (around 400 to 700 MB ) from one partion to another my machine do not response at all (i can not work on another shell) during data transfer.

Both paritions on the same spindle? That's going to suck no matter what you do... but it shouldn't be as bad as you describe. Have you tried 2.2.x ?

--cw

#4 lore raw
In reply to Guido Stepken
Quoted 9 lines · Jaswinder Singh
Quoted 8 lines

RedHat 7.1 - IDE IBM 41.1 GIG
Update to 2.4.5 -> noticed, that hdparm -t /dev/hda went down from 10 MByte/sec to 1.9 MByte/sec
Any special Options, beside ide-scsi driver activated ..

Anybody noticed the same problem ? Any clues ?

Quoted 6 lines

yes , i am also not happy with IDE performance of Linux . That why i dont use Hard disk in my Target machines ;)

When ever i copy big data (around 400 to 700 MB ) from one partion to another my machine do not response at all (i can not work on another shell) during data transfer.

The answer for both of you is:

hdparm -d1 /dev/hd{whatever}

Without DMA enabled, performance is going to suck. 1.9 MB/sec is actually pretty
good without DMA ;)

Signature · 6 lines
................................................................
:   [email protected]   : And I see the elder races,         :
:.........................: putrid forms of man                :
:   Jakob Østergaard      : See him rise and claim the earth,  :
:        OZ9ABN           : his downfall is at hand.           :
:.........................:............{Konkhra}...............:
#5 lore raw
In reply to Guido Stepken
Quoted 7 lines

The answer for both of you is:

hdparm -d1 /dev/hd{whatever}

Without DMA enabled, performance is going to suck. 1.9 MB/sec is actually pretty
good without DMA ;)

I agree.

I get around 4MB/s without and 35.75MB/s with... hdparm -c3 -d1 -m16 /dev/hda
on a KT133 (A7V) and 45GB IBM DTLA hard drive.

no data corruption, even with a sblive, an scsi adapter, 2 hard drives...

François Cami

Quoted 8 lines

--
.................................................................
: [email protected] : And I see the elder races, :
:.........................: putrid forms of man :
: Jakob Østergaard : See him rise and claim the earth, :
: OZ9ABN : his downfall is at hand. :
:.........................:............{Konkhra}...............:
-

Signature · 4 lines
All that is gold does not glitter,
	Not all those who wander are lost,
The old that is strong does not wither,
	Deep roots are not touched by the frost.
Quoted 1 line

From the ashes a fire shall be woken,

A light from the shadows shall spring,
Renewed shall the blade that was broken,
The crownless again shall be king.

The Riddle of Strider JRR Tolkien

#6 lore raw
In reply to Guido Stepken
Quoted 8 lines · Chris Wedgwood

On Sat, May 26, 2001 at 05:16:42PM -0700, Jaswinder Singh wrote:

When ever i copy big data (around 400 to 700 MB ) from one partion to another my machine do not response at all (i can not work on another shell) during data transfer.

Both paritions on the same spindle? That's going to suck no matter what you do...

I am having only one harddisk of 20 Gb , so i made four partitions on it.

Quoted 1 line

but it shouldn't be as bad as you describe.

But it is more bad then i described , especially when i do telnet/ssh from my home .

I am not able to understand why Linux read and/or write harddisk after some time (after few hours ) , harddisk read/write leds keep on glowing for few minutes , even though nobody working on it and machine is in idle state.

Quoted 2 lines

Have you tried 2.2.x ?

yes i am taking about 2.2.12 .

And in my 2.4.2 :-

Harddisk read-write problem is also there , during harddisk read-write activity i took two hdparams and i got :-
i . 963.76 kB/sec
ii. 1.05 MB/sec

and when my machine becomes normal i got :-

9.07 MB/sec

and after hdparam -d1 i got :-

9.55 MB/sec

But my problem is why linux boxes do not response for few seconds (sometimes) and especially during telnet/ssh it looks more worst and looks similar to Microsoft Windows :(

there is problem in scheduling or what ?

Thank you,

Best Regards,

Jaswinder.
--
These are my opinions not 3Di.

#7 lore raw
In reply to Guido Stepken
Quoted 6 lines · From: "Jakob Østergaard" <[email protected]> wrote :

The answer for both of you is:

hdparm -d1 /dev/hd{whatever}

Without DMA enabled, performance is going to suck. 1.9 MB/sec is actually

pretty

Quoted 2 lines

good without DMA ;)

i think DMA or PCI is not my solution , atleast :)

Thank you,

Best Regards,

Jaswinder.
--
These are my opinions not 3Di.

#8 lore raw
In reply to Guido Stepken

In article <00c701c0e6d8$2b28ea40$4aa6b3d0@Toshiba>,

Quoted 3 lines · Jaswinder Singh

I am not able to understand why Linux read and/or write harddisk after some time (after few hours ) , harddisk read/write leds keep on glowing for few minutes , even though nobody working on it and machine is in idle state.

Are you sure it is idle. It might be running something from cron- say 'updatedb' or similar. That will cause a lot of disk i/o, and _ofcourse_ performance will be bad then - the machine is doing a lot of other things.

It might also be that you don't have enough memory and the machine is swapping itself to death. Running netscape or mozilla perhaps? These are known to blow themselves up to 50-100 MB (!). That will cause the exact symptoms you're seeing.

Quoted 4 lines
Quoted 2 lines

Have you tried 2.2.x ?

yes i am taking about 2.2.12 .

2.2.12 is old. In my experience, 2.2.19 is the first really good 2.2 kernel. Especially in low-memory situations like you might be experiencing..

Quoted 1 line

And in my 2.4.2 :-

2.4.2 isn't all that good either.. 2.4.x doesn't have VM sorted out just yet

Quoted 2 lines

and when my machine becomes normal i got :-
9.07 MB/sec

Sounds about right for a non-UDMA disk.

Quoted 4 lines

But my problem is why linux boxes do not response for few seconds (sometimes) and especially during telnet/ssh it looks more worst and looks similar to Microsoft Windows :(
there is problem in scheduling or what ?

Try 2.2.19

Mike.

#9 lore raw
In reply to Guido Stepken

On Sun, May 27, 2001 at 11:09:25AM -0700, Jaswinder Singh wrote:

But my problem is why linux boxes do not response for few seconds (sometimes) and especially during telnet/ssh it looks more worst and looks similar to Microsoft Windows :(

there is problem in scheduling or what ?

Sounds like IO starvation, if your reading from one partition to the other, it's going to suck for IO --- the head will be going nuts. Also, 2.2.x has a single IO queue for writes (doesn't it?), so that won't help.

What if you dd a file from one partition to another, is the problem lesser with larger buffers, say 64M or so?

What about dd from /dev/zero to a partition, so you get heavy writes, does this also suck?

--cw

#10 lore raw
In reply to Guido Stepken
Quoted 6 lines · Miquel van Smoorenburg

Are you sure it is idle. It might be running something from cron- say 'updatedb' or similar. That will cause a lot of disk i/o, and _ofcourse_ performance will be bad then - the machine is doing a lot of other things.

I am the only user who use my linux boxes , i never use cron , and i have no relation with database at all.

Quoted 5 lines

It might also be that you don't have enough memory and the machine is swapping itself to death. Running netscape or mozilla perhaps? These are known to blow themselves up to 50-100 MB (!). That will cause the exact symptoms you're seeing.

I never go to X-Windows , i do all my work from console.

cat /proc/meminfo
total: used: free: shared: buffers: cached:

Mem:  114475008 111816704  2658304  5234688 68067328 24711168
Swap: 583954432  2654208 581300224
MemTotal:    111792 kB
MemFree:       2596 kB
MemShared:     5112 kB
Buffers:      66472 kB
Cached:       24132 kB
SwapTotal:   570268 kB
SwapFree:    567676 kB
Quoted 4 lines

2.4.2 isn't all that good either.. 2.4.x doesn't have VM sorted out just yet

may be this problem of VM as suggested by Mark Hahn.

Thank you,

Jaswinder.
--
These are my opinions not 3Di.