On Sun, Mar 26, 2006 at 09:18:28AM -0500, Justin Piszcz wrote:
> Does anyone have any benchmarks for the XFS filesystem contrasted
> with the amount of fragmentation on the disk? I have spoken with
> someone who believes defragmenting filesystems with XFS is probably
> a waste of time unless there are benchmarks to back it up.
it depends, badly fragmented files do hurt performance if it's
pathalogically bad, i've worked with a couple of people who do real
time video to help reduce this and the differences can be enourmous
> Benchmarks that show that performance degrades significantly as
> fragmentation increases.
benchmarks are highly subjective to the usage pattern and the job at
hand, so you need to state what you are trying to measure here in
> Has anyone seen or performed any benchmarks? Is it worth defragging
> XFS filesystems?
again, it really depends, for some people it is a win, for many people
i doubt it's worth it
also, right now i'm a little concerned that there is some odd
interaction with one of my systems and xfs_fsr, if i copy several
hundred GBs of data to another volume and then run xfs_fsr afterwards
i will get the odd (md5) checksum mismatch --- i've not had a chance
to figure out why the odd file breaks but the bulk are ok though