Hi, On Sonntag, 13. März 2011, Ben Hutchings wrote: > > how inefficient is it? how large? > The larger the file, the more benefit there is from mapping its blocks > with extents. The cost in terms of disk space is minimal. The cost in > terms of seek operations is probably more significant. So how much faster is xfs with files of 10-20gb? Something like 10% or 20% or rather something like 200%? > XFS should be fine, but doesn't flush data to disk as eagerly as ext3, > i.e. you need an explicit sync before you can be sure the data will > still be there after an unclean shutdown. Same is true for ext4 (at > least with default settings). The problem with unclean shutdowns is, that you hardly ever can do a sync in advance... And really, even if xfs is 300% faster, I prefer 10% more reliability. (Especially in debconf-video setups where we dont have a backup of our 2-5tb raid several times a day...) cheers, Holger
Attachment:
signature.asc
Description: This is a digitally signed message part.