Re: 1394 Performance

On 3 Feb, James Finnall wrote:
> I built a custom 2.4.24 kernel with two
> patches. I have noticed that my performance of the firewire was
> not what I think it should be using the included 1394 drivers.

Yes, ieee1394/sbp2 is not a big performer.

> So I tried using the drivers I had from my 2.4.22 kernel build (about
> 30Aug2003) from before the upgrade and I have checked out the
> latest revision for 2.4 (1119) as well and tested it. No
> differences in performance noted.

I believe the performance should practically the same as in the
earlier days of the 2.4 kernel series.

> I have also tried loading the
> modules with various parameters and can only degrade the
> performance but not increase it.

I can confirm this:
http://marc.theaimsgroup.com/?l=linux1394-user&m=107325612422364&w=2

> I have a 4 bay enclosure with two Oxford 911 bridges. The device I
> am using for the performance testing is a hard disk. Western
> Digital 7200 RPM, ATA100 40 GByte hard disk. The best perfomance
> I have been able to get using the "hdparm -t" command is about 15
> MB/s.

Mark already commented that you should use -Tt, not -t. This is
mentioned in hdparm's manual page. Anyway, 15MB/s is low for a
7200rpm disk even under Linux. I get more than 20 with a 5400rpm
disk (though it is a 120GB disk, therefore with higher data
density than your 40GB disk).

> I know I have seen this substantially higher before, about 3
> times better.

3 times? I very much doubt it. This would be dramaticly. As
mentioned above, I even doubt that 2.4 kernel revisions are
substantially different from each other.

The performance under kernel 2.6 is also basically the same, with
minor exceptions that are rooted outside of the 1394 driver suite
as far as I understand:
http://marc.theaimsgroup.com/?l=linux1394-user&m=107374425232357&w=2
http://marc.theaimsgroup.com/?l=linux1394-user&m=107377017922113&w=2

> I tested write performance by simply timing a file
> copy. A 3 GByte file took 5 minutes. That is only 10.5 MB/s
> writing speed or 84 mbps of the 400 mbps bandwidth.

This is consistent with your hdparm result.

> This problem I think is producing problems trying to burn DVD's
> above 2x speed. If I try to read a DVD using sdd it reads at 4.5
> MB/s. The drive is rated at 8x or 11 MB/s read speed. It burns at
> 2.4x or 3.5 MB/s. So the reading speed here is just barely above
> the writing speed of the drive. I have a 4x DVD burner also
> connected but it will not write above 2x while connected on the
> firewire bus. It only reads at 5.5 MB/s and it is rated at 16 MB/s
> read. It is suppose to write at 5.5 MB/s.

Forget about any ratings of CD-ROM-like devices. These are
meaningless peak numbers. A more sound comparison would be with
the same drive attached to an internal IDE channel.

-- 
Stefan Richter
-=====-=-=-- --=- --=--
http://arcgraph.de/sr/
-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
mailing list Linux1394-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux1394-user