Update on disk IO slowdown using udev

Update on disk IO slowdown using udev

[Ake]

Hi!
After some more experimenting i did the following,

mkdir /tmpfsdev
mount -t tmpfs tmpfs /tmpfsdev
mknod /tmpfsdev/sdb b 8 16 (This is my test disk)
dd if=/tmpfsdev/sdb of=/dev/null bsÅ92k count 
20+0 records in
20+0 records out
167772160 bytes transferred in 11.006714 seconds (15242711 bytes/sec)

mkdir /rawdev
mknod /rawdev/sdb b 8 16
dd if=/rawdev/sdb of=/dev/null bsÅ92k count 
20+0 records in
20+0 records out
167772160 bytes transferred in 1.668631 seconds (100544793 bytes/sec)


So tmpfs is the culprit.
So, yes using udev i DO get a slowdown due to its use of tmpfs.

-- 
Ake Sandgren, HPC2N, Umea University, S-90187 Umea, Sweden
Internet: ake@hpc2n.umu.se	Phone: +46 90 7866134 Fax: +46 90 7866126
Mobile: +46 70 7716134 WWW: http://www.hpc2n.umu.se


-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id149&alloc_idÅ66&op=click
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: Update on disk IO slowdown using udev

[Harald Hoyer]

[-- Attachment #1: Type: text/plain, Size: 754 bytes --]

Try it the other way round. I guess the data is cached :)
Use bonnie or hdparm -Tt to measure throughput!

Ake wrote:
> Hi!
> After some more experimenting i did the following,
> 
> mkdir /tmpfsdev
> mount -t tmpfs tmpfs /tmpfsdev
> mknod /tmpfsdev/sdb b 8 16 (This is my test disk)
> dd if=/tmpfsdev/sdb of=/dev/null bs=8192k count=20
> 20+0 records in
> 20+0 records out
> 167772160 bytes transferred in 11.006714 seconds (15242711 bytes/sec)
> 
> mkdir /rawdev
> mknod /rawdev/sdb b 8 16
> dd if=/rawdev/sdb of=/dev/null bs=8192k count=20
> 20+0 records in
> 20+0 records out
> 167772160 bytes transferred in 1.668631 seconds (100544793 bytes/sec)
> 
> 
> So tmpfs is the culprit.
> So, yes using udev i DO get a slowdown due to its use of tmpfs.
> 


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 252 bytes --]

Re: Update on disk IO slowdown using udev

[Ake]

On Tue, May 25, 2004 at 12:58:36PM +0200, Harald Hoyer wrote:
> Try it the other way round. I guess the data is cached :)
> Use bonnie or hdparm -Tt to measure throughput!
> 
> Ake wrote:
> >Hi!
> >After some more experimenting i did the following,
> >
> >mkdir /tmpfsdev
> >mount -t tmpfs tmpfs /tmpfsdev
> >mknod /tmpfsdev/sdb b 8 16 (This is my test disk)
> >dd if=/tmpfsdev/sdb of=/dev/null bsÅ92k count 
> >20+0 records in
> >20+0 records out
> >167772160 bytes transferred in 11.006714 seconds (15242711 bytes/sec)
> >
> >mkdir /rawdev
> >mknod /rawdev/sdb b 8 16
> >dd if=/rawdev/sdb of=/dev/null bsÅ92k count 
> >20+0 records in
> >20+0 records out
> >167772160 bytes transferred in 1.668631 seconds (100544793 bytes/sec)


To prove my point.

reboot

dd if=/rawdev/sdb of=/dev/null bsÅ92k count 
20+0 records in
20+0 records out
167772160 bytes transferred in 1.572609 seconds (106683964 bytes/sec)

mount -t tmpfs tmpfs /tmpfsdev
mknod /tmpfsdev/sdb b 8 16
dd if=/tmpfsdev/sdb of=/dev/null bsÅ92k count 
20+0 records in
20+0 records out
167772160 bytes transferred in 10.733471 seconds (15630746 bytes/sec)


There is no filesystem on this device so bonnie is no good.

ido-i [/]# hdparm -Tt /rawdev/sdb 

/rawdev/sdb:
 Timing buffer-cache reads:   2276 MB in  2.00 seconds = 1137.60 MB/sec
 Timing buffered disk reads:  292 MB in  3.01 seconds =  97.15 MB/sec

ido-i [/]# hdparm -Tt /tmpfsdev/sdb 

/tmpfsdev/sdb:
 Timing buffer-cache reads:   2276 MB in  2.00 seconds = 1137.04 MB/sec
 Timing buffered disk reads:   42 MB in  3.01 seconds =  13.94 MB/sec


-- 
Ake Sandgren, HPC2N, Umea University, S-90187 Umea, Sweden
Internet: ake@hpc2n.umu.se	Phone: +46 90 7866134 Fax: +46 90 7866126
Mobile: +46 70 7716134 WWW: http://www.hpc2n.umu.se


-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id149&alloc_idÅ66&op=click
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: Update on disk IO slowdown using udev

[Harald Hoyer]

[-- Attachment #1: Type: text/plain, Size: 956 bytes --]

Ake wrote:
> To prove my point.
> 
> reboot
> 
> dd if=/rawdev/sdb of=/dev/null bs=8192k count=20
> 20+0 records in
> 20+0 records out
> 167772160 bytes transferred in 1.572609 seconds (106683964 bytes/sec)
> 
> mount -t tmpfs tmpfs /tmpfsdev
> mknod /tmpfsdev/sdb b 8 16
> dd if=/tmpfsdev/sdb of=/dev/null bs=8192k count=20
> 20+0 records in
> 20+0 records out
> 167772160 bytes transferred in 10.733471 seconds (15630746 bytes/sec)
> 
> 
> There is no filesystem on this device so bonnie is no good.
> 
> ido-i [/]# hdparm -Tt /rawdev/sdb 
> 
> /rawdev/sdb:
>  Timing buffer-cache reads:   2276 MB in  2.00 seconds = 1137.60 MB/sec
>  Timing buffered disk reads:  292 MB in  3.01 seconds =  97.15 MB/sec
> 
> ido-i [/]# hdparm -Tt /tmpfsdev/sdb 
> 
> /tmpfsdev/sdb:
>  Timing buffer-cache reads:   2276 MB in  2.00 seconds = 1137.04 MB/sec
>  Timing buffered disk reads:   42 MB in  3.01 seconds =  13.94 MB/sec
> 
> 



Ok... you got a point! very odd!

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 252 bytes --]

Re: Update on disk IO slowdown using udev

[Harald Hoyer]

[-- Attachment #1: Type: text/plain, Size: 450 bytes --]

Ake wrote:
> 
> ido-i [/]# hdparm -Tt /rawdev/sdb 
> 
> /rawdev/sdb:
>  Timing buffer-cache reads:   2276 MB in  2.00 seconds = 1137.60 MB/sec
>  Timing buffered disk reads:  292 MB in  3.01 seconds =  97.15 MB/sec
> 
> ido-i [/]# hdparm -Tt /tmpfsdev/sdb 
> 
> /tmpfsdev/sdb:
>  Timing buffer-cache reads:   2276 MB in  2.00 seconds = 1137.04 MB/sec
>  Timing buffered disk reads:   42 MB in  3.01 seconds =  13.94 MB/sec
> 
> 


what about ramfs??

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 252 bytes --]

Re: Update on disk IO slowdown using udev

[Harald Hoyer]

[-- Attachment #1: Type: text/plain, Size: 47 bytes --]

ok, I can reproduce it. FC2 kernel 2.6.5-1.358

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 252 bytes --]

Re: Update on disk IO slowdown using udev

[Ake]

On Tue, May 25, 2004 at 02:05:44PM +0200, Harald Hoyer wrote:
> what about ramfs??

ramfs has exactly the same problem.

-- 
Ake Sandgren, HPC2N, Umea University, S-90187 Umea, Sweden
Internet: ake@hpc2n.umu.se	Phone: +46 90 7866134 Fax: +46 90 7866126
Mobile: +46 70 7716134 WWW: http://www.hpc2n.umu.se


-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id149&alloc_idÅ66&op=click
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: Update on disk IO slowdown using udev

[Oliver Neukum]

Am Dienstag, 25. Mai 2004 14:49 schrieb Harald Hoyer:
> ok, I can reproduce it. FC2 kernel 2.6.5-1.358

That is very very odd. Is there anything running in the background?
To rule out the obvious, dd opens the devices only once, doesn't it?

	Regards
		Oliver


-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id149&alloc_idÅ66&op=click
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: Update on disk IO slowdown using udev

[Olaf Hering]

 On Tue, May 25, Oliver Neukum wrote:

> Am Dienstag, 25. Mai 2004 14:49 schrieb Harald Hoyer:
> > ok, I can reproduce it. FC2 kernel 2.6.5-1.358
> 
> That is very very odd. Is there anything running in the background?
> To rule out the obvious, dd opens the devices only once, doesn't it?

just try it yourself. init=/bin/bash is enough. its scsi only.
you better bring that to lkml.


-- 
USB is for mice, FireWire is for men!

sUse lINUX ag, n√úRNBERG


-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id149&alloc_idÅ66&opÃk
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: Update on disk IO slowdown using udev

[Oliver Neukum]

Am Dienstag, 25. Mai 2004 20:42 schrieb Olaf Hering:
>  On Tue, May 25, Oliver Neukum wrote:
> 
> > Am Dienstag, 25. Mai 2004 14:49 schrieb Harald Hoyer:
> > > ok, I can reproduce it. FC2 kernel 2.6.5-1.358
> > 
> > That is very very odd. Is there anything running in the background?
> > To rule out the obvious, dd opens the devices only once, doesn't it?
> 
> just try it yourself. init=/bin/bash is enough. its scsi only.
> you better bring that to lkml.

Using an USB memory stick I see sys time fall by 50%.
Mightily odd.

	Regards
		Oliver



-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id149&alloc_idÅ66&op=click
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel