Messages in this thread | | | Date | Tue, 7 Jan 2014 00:10:42 +1100 | From | Dave Chinner <> | Subject | Re: [numa shrinker] 9b17c62382: -36.6% regression on sparse file copy |
| |
On Mon, Jan 06, 2014 at 04:20:48PM +0800, fengguang.wu@intel.com wrote: > Hi Dave, > > We noticed throughput drop in test case > > vm-scalability/300s-lru-file-readtwice (*) > > between v3.11 and v3.12, and it's still low as of v3.13-rc6: > > v3.11 v3.12 v3.13-rc6 > --------------- ------------------------- ------------------------- > 14934707 ~ 0% -48.8% 7647311 ~ 0% -47.6% 7829487 ~ 0% vm-scalability.throughput > ^^ ^^^^^^ > stddev% change%
What does this vm-scalability.throughput number mean?
> (*) The test case basically does > > truncate -s 135080058880 /tmp/vm-scalability.img > mkfs.xfs -q /tmp/vm-scalability.img > mount -o loop /tmp/vm-scalability.img /tmp/vm-scalability > > nr_cpu=120 > for i in $(seq 1 $nr_cpu) > do > sparse_file=/tmp/vm-scalability/sparse-lru-file-readtwice-$i > truncate $sparse_file -s 36650387592 > dd if=$sparse_file of=/dev/null & > dd if=$sparse_file of=/dev/null & > done
So a page cache load of reading 120x36GB files twice concurrently? There's no increase in system time, so it can't be that the shrinkers are running wild.
FWIW, I'm at LCA right now, so it's going to be a week before I can look at this, so if you can find any behavioural difference in the shrinkers (e.g. from perf profiles, on different filesystems, etc) I'd appreciate it...
Cheers,
Dave. -- Dave Chinner david@fromorbit.com
| |