[vdr] Re: performance during cutting

[Joachim Thees <thees@informatik.uni-kl.de>]

On 27 Mar, Arno Esser wrote:
> I got error messages on syslog saying something about disabling dma mode
> caused by problems with the driver. I did not further investigate how to solve
> this e.g. no kernel driver. Since I had a spare slot on the standard ide
> controller (UDMA33!) I connected my harddisk to that and the problems were gone.


Hello Arno, 

there are no such reports in my logfiles. But tonight I did some tests
with vdr and hdparm and I found out that without the "-K 1 -k 1"
parameters a possible soft-reset of the ide-interface will restore
default values (so this would be PIO-3 (i.e. no dma) on my system). So
maybe the system does a ide-soft-reset after some time (would this be
logged?) and so my interface falls back to non-dma mode.

After some tests I found the following settings to work perfectly for my
system:

	hdparm -a8 -A1 -m8 -c3 -d1 -X66 -u1 -W1 -k1 -K1 /dev/hda

	(note1: be careful with -u1) 
	(note2: use -k1 -K1 only if the other settings work stable)
	(note3: -a (sector  count  for  filesystem  read-ahead) has
	        no big impact on vdr-performance (I tested 1 .. 255))

Currently I get 20MB/s and cutting works with about 60% CPU load. But in
the last time it took some time before the problems appeared, so I don´t
know if it really solves my problem ... :-}

Perhaps someone who can reproduce temporary loss of disk performance
(especially during cutting) after runnig the vdr-PC for some time could
try a 

	hdparm -v -I -tT /dev/hda

before and after rebooting to see if settings were modified while the
system was running?


Greetings,

Joachim


-- 
 --------------------------------------------------
 Joachim Thees <thees@informatik.uni-kl.de>
 Univ. of Kaiserslautern, Computer Networks Group
 --------------------------------------------------