20 Jan
2005
20 Jan
'05
3:14 p.m.
Florin Andrei <florin(a)andrei.myip.org> writes:
On Thu, 2005-01-06 at 10:40 -0500, Rick B wrote:
How is its low-latency realtime behavior?
In recent tests of his experimental isochronous scheduler, Con Kolivas
uncovered a periodic 6msec scheduling delay every 20 seconds while
running reiserfs with logging enabled...
http://ck.kolivas.org/patches/SCHED_ISO/iso2-benchmarks/jack_test3-2.6.11-r…
Without logging it ran much better...
http://ck.kolivas.org/patches/SCHED_ISO/iso2-benchmarks/jack_test3-iso2-40c…
I'd love to see some similar tests using XFS. We should probably
collect audio performance data on all the available filesystems.
--
joq
From what I've read XFS main strong point is
*reading* data from the
disk faster than other FSs', especially as it pertains to large video
files. But when it comes to writing and deleting, its performance is
worse than other FSes. I've personally tried it and it didn't seem any
faster than ReiserFS.
Actually, XFS is very fast at deleting very large files. I know, i do
DVD authoring all the time, Ext3 is simply ludicrous when deleting a lot
of very large project files, you have to wait and wait and wait... With
XFS, even a 50GB project gets deleted instantaneously.
Also, when writing real fast to really big files (think: video capture)
pretty much nothing beats XFS. 
