[ardour-users] disk throughput problems with beta

[Peter Lutek]

On Thu, 31 Mar 2005 16:17:20 +0200, Karl Hammar <karl at aspodata.se> wrote:

>
>> On Wed, 30 Mar 2005 22:45:49 -0500 (EST), Jesse Chappell  
>> <jesse at essej.net>
>> wrote:
>>
>> > On Wed, 30 Mar 2005, Peter Lutek wrote:
>> >
>> >> just looked at some recent session directories and, while a separate
>> >> peak
>> >> directory does exist, in fact it is empty, and the peak files are  
>> being
>> >> written in the same directory as the wave files.
>> >
>> > This is not good, and almost definitely a bug.  Withhold complete
>> > judgement until we fix this, because it definitely helped out
>> > with ext3.
>> >
>>
>> in my most recent tests, peak files ARE where they should be. don't know
>> what was going on before. i'll keep monitoring to make sure.
>>
>> sorry to have raised the alarm. however, the problem with recording  
>> while
>> recording waveform display is ON still remains. same results for me on
>> both ext3 and xfs.
>>
>> -p
>
> It is my understanding that a journalling filesystem is worse
> latencyvise compared to a traditional one like ext2.
>
> If you have a latency problem and suspect the filesystem;
> take a spare disk, make a few partitions with ext2, ext3, xfs, ...,
> stop all daemons (else you might be measuring latency problems in
> /var), unmount whatever you can.
>
> Then mount them one at a time, start recording, umoun, take next
> filesystem. Btw., there was a bug in ext3 last summer, which caused
> a custemer get 2s latency from time to time.

ok, i tested ext2. same problem.

-p