All of lore.kernel.org
 help / color / mirror / Atom feed
* IDE performance problems on 2.6.0-pre1
@ 2003-07-15 21:49 derek
  2003-07-15 22:03 ` Felipe Alfaro Solana
  0 siblings, 1 reply; 4+ messages in thread
From: derek @ 2003-07-15 21:49 UTC (permalink / raw)
  To: linux-kernel

Hello,

My ide performance seems to have dropped noticably from 2.4.x to 
2.6.0-pre1...

the controller is an hpt-366, there's only one drive connected to it.  
It's on an 80 pin cable, and the cable is plugged in the right way.

hdparm reports that the drive is using udma4 (it's udma5 capable, the 
controller is not)

from 2.4.x, hdparm -t
 Timing buffered disk reads:  140 MB in  3.02 seconds =  46.36 MB/sec

and vmstat 1 while it's running shows
 0  0      0 349064   1964   9784    0    0     0     0  101     4  0  0 100  0
 1  0      0 335360   1964  23480    0    0 11648     0  282   356  0  9 90  0
 1  0      0 286976   1964  71864    0    0 48384     0  863  1469  1 38 61  0
 1  0      0 239360   1964 119480    0    0 47616     0  844  1451  1 33 65  0
 0  0      0 349108   1964   9784    0    0 35840     0  664  1095  0 31 69  0

also,

time dd if=/dev/hde1 of=/dev/null bs=4096
768088+0 records in
768088+0 records out

real    1m6.079s
user    0m0.850s
sys     0m16.720s


under 2.6.0-pre1, I get

 Timing buffered disk reads:   88 MB in  3.02 seconds =  29.14 MB/sec

vmstat says
 0  0      0 334748   2172  10816    0    0     0     0 1002     4  0  0 100  0
 1  1      0 329180   5636  12860    0    0  3464     0 1031    64  0  2 93  4
 1  0      0 298140  35692  12860    0    0 30080     0 1244   483  1 20 50 30
 1  0      0 266972  65924  12888    0    0 30208     0 1239   480  0 19 50 30
 1  0      0 334108   2172  10884    0    0 26496     0 1253   426  0 26 48 26
 0  0      0 334180   2172  10884    0    0     0     0 1010    18  0  0 100  0

(this is a dual processor machine, so this looks to me like I'm cpu bound 
where I wasn't before?)

time dd if=/dev/hde1 of=/dev/null bs=4096
dd: reading `/dev/hde1': Input/output error
768088+0 records in
768088+0 records out

real    1m33.977s
user    0m0.967s
sys     0m26.695s

the input/output error is new too...

I also tried with everything in the Kernel Hacking sub-menu disabled, and 
got results from hdparm as high as 31.53MB/sec

Any idea what's wrong?

^ permalink raw reply	[flat|nested] 4+ messages in thread
* Re: IDE performance problems on 2.6.0-pre1
  2003-07-15 21:49 IDE performance problems on 2.6.0-pre1 derek
@ 2003-07-15 22:03 ` Felipe Alfaro Solana
  2003-07-15 22:27   ` derek
  0 siblings, 1 reply; 4+ messages in thread
From: Felipe Alfaro Solana @ 2003-07-15 22:03 UTC (permalink / raw)
  To: derek; +Cc: LKML

On Tue, 2003-07-15 at 23:49, derek@signalmarketing.com wrote:
> Hello,
> 
> My ide performance seems to have dropped noticably from 2.4.x to 
> 2.6.0-pre1...

What does "hdparm /dev/hde" tell us?


^ permalink raw reply	[flat|nested] 4+ messages in thread
* Re: IDE performance problems on 2.6.0-pre1
  2003-07-15 22:03 ` Felipe Alfaro Solana
@ 2003-07-15 22:27   ` derek
  0 siblings, 0 replies; 4+ messages in thread
From: derek @ 2003-07-15 22:27 UTC (permalink / raw)
  To: Felipe Alfaro Solana; +Cc: LKML



On Tue, 16 Jul 2003, Felipe Alfaro Solana wrote:

> On Tue, 2003-07-15 at 23:49, derek@signalmarketing.com wrote:
> > Hello,
> > 
> > My ide performance seems to have dropped noticably from 2.4.x to 
> > 2.6.0-pre1...
> 
> What does "hdparm /dev/hde" tell us?

hdparm /dev/hde  from 2.6.0-pre1

/dev/hde:
 multcount    = 16 (on)
 IO_support   =  1 (32-bit)
 unmaskirq    =  1 (on)
 using_dma    =  1 (on)
 keepsettings =  0 (off)
 readonly     =  0 (off)
 readahead    = 256 (on)
 geometry     = 50765/16/63, sectors = 117231408, start = 0

---

hdparm /dev/hde from 2.4

/dev/hde:
 multcount    = 16 (on)
 IO_support   =  1 (32-bit)
 unmaskirq    =  1 (on)
 using_dma    =  1 (on)
 keepsettings =  0 (off)
 readonly     =  0 (off)
 readahead    =  8 (on)
 geometry     = 116301/16/63, sectors = 117231408, start = 0


I hadn't noticed the change in readahead previously, so I tried setting it 
to 8 in 2.6.0-pre1 with hdparm -a 8 /dev/hde

hdparm -t /dev/hde

/dev/hde:
 Timing buffered disk reads:   32 MB in  3.01 seconds =  10.64 MB/sec

hdparm -a 512 on the other hand...

 Timing buffered disk reads:  140 MB in  3.03 seconds =  46.18 MB/sec

and I get my previous numbers back.

I guess the meaning of the parameter has changed dramatically?  (what was 
once sectors is now bytes?)

^ permalink raw reply	[flat|nested] 4+ messages in thread
* Re: IDE performance problems on 2.6.0-pre1
@ 2003-07-15 23:16 Voluspa
  0 siblings, 0 replies; 4+ messages in thread
From: Voluspa @ 2003-07-15 23:16 UTC (permalink / raw)
  To: linux-kernel


2003-07-15 22:27:19 derek wrote:

> hdparm -a 512 on the other hand...
> Timing buffered disk reads:  140 MB in  3.03 seconds =  46.18 MB/sec
> and I get my previous numbers back.

Ahhh thank you derek! I followed your example and got _my_ old numbers
back, from 22 to 28 MB/sec. Surely this is a bUg. Been there for a long
time.

Mvh
Mats Johannesson

^ permalink raw reply	[flat|nested] 4+ messages in thread
end of thread, other threads:[~2003-07-15 23:00 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-07-15 21:49 IDE performance problems on 2.6.0-pre1 derek
2003-07-15 22:03 ` Felipe Alfaro Solana
2003-07-15 22:27   ` derek
2003-07-15 23:16 Voluspa

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.