Email address obfuscation in effect -- please
click here to turn it off.
[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
On Tue, 4 Apr 2006, Mark Rages wrote:
On 4/4/06, Dave Lloyd <EMAIL:PROTECTED> wrote:
On Tue, 4 Apr 2006, Mark Rages wrote:
About filesystems, I have benchmarked them in the past, and my
conclusion was: When you're just writing a single file full of 10
gigs of zeroes, the filesystem doesn't matter. It's when you have a
lot of bitty files that the filesystem design choices start to become
apparent. For this application, I'm not running a mail server or
compiling the kernel, I'm storing huge video files.
If you're using the filesystem for storing, streaming and/or editing
video files, XFS will offer excellent performance. It was designed with
this in mind. You may want to look at the xfs man pages to see all the
bits you can twiddle when you make the filesystem.
Can you expand on what you mean by "excellent performance"?
Throughput appears to be limited by the disk hardware, not the
filesystem in use. So what other specs are critical to video use?
True, your hardware may be the limit here. Other systems I've seen have
streaming video served from big storage arrays going through multiple
fiber channel interfaces. At that point, the ability for the the kernel
and the filesystem code to cope with large amounts of I/O becomes the
bottleneck.
I guess a better filesystem could give less latency when you seek
in the large video file. Maybe it uses less CPU. I know deleting
a 10G file takes a good bit of time in ext3. What should I look
for when I bench test filesystems?
It's difficult to say what to benchmark as your workflow will vary. I'd
suggest looking at your typical workflow and find a few key areas that
may stress the filesystem. Produce reproducable testcases and run and
benchmark those on a variety of filesystems.
--dlloyd
_______________________________________________
members mailing list
EMAIL:PROTECTED
http://mlug.missouri.edu/mailman/listinfo/members