On Tue, 13 Sep 2005, Bill Kendall wrote:
> It might be helpful to determine if xfsdump is slow because of the way
> it's interacting with the tape drive, or if it's just slow in general
> on your filesystem. Try using "-f /dev/null" and see how it performs.
> You don't need to specify -d/-S or -b in this case.
I guess it's not related to the interaction between xfsdump and the tape
drive: in my initial tests, I used dd to write on the tape.
Here is the result you asked for:
# xfsdump -J -p 300 -f /dev/null /raid
It gives me:
xfsdump: dump size (non-dir files) : 29369583312 bytes
xfsdump: dump complete: 4557 seconds elapsed
So an average of 64444938 bytes/sec. Not much better than when writing
to tape...
PS: as an additional test, on the same array/server, I created an ext3
filesystem, pushed the same files than for the XFS/xfsdump tests,
and ran a dump session:
# dump -0 -a -f /dev/nst0 /raid
I finally got this:
DUMP: 30325260 blocks (29614.51MB) on 1 volume(s)
DUMP: finished in 1792 seconds, throughput 16922 kBytes/sec
DUMP: Date of this level 0 dump: Tue Sep 13 14:34:57 2005
DUMP: Date this dump completed: Tue Sep 13 15:23:06 2005
DUMP: Average transfer rate: 16903 kB/s
--
Nicolas
|