On Tue, 1 Jul 2014 10:21:40 -0400 Trond Myklebust <trond.myklebust@xxxxxxxxxxxxxxx> wrote: > On Tue, Jul 1, 2014 at 10:10 AM, Jeff Layton <jlayton@xxxxxxxxxxxxxxx> wrote: > > > > Quite right. I changed the program to be run as root and had each child > > process do an setfsuid/setfsgid to a different UID/GID combo: > > > > [jlayton@tlielax lockperf]$ time sudo ./opentest -n 128 -l 100 /mnt/rawhide/opentest > > > > real 0m3.448s > > user 0m0.078s > > sys 0m0.377s > > [jlayton@tlielax lockperf]$ time sudo ./opentest -n 128 -l 100 /mnt/rawhide/opentest > > > > real 0m3.344s > > user 0m0.053s > > sys 0m0.374s > > [jlayton@tlielax lockperf]$ time sudo ./opentest -n 128 -l 100 /mnt/rawhide/opentest > > > > real 0m3.550s > > user 0m0.049s > > sys 0m0.394s > > > > > > ...so the speedup seems to be quite dramatic, actually -- 3x faster or > > so with the patched kernel. > > > > The underlying filesystem is ext4 here, and the config is a rawhide > > debug kernel config. For my next trick, I'll build some non-debug > > kernels and replicate the test with them. Stay tuned... > > Just for grins, we should probably also try with the 3 client side > OPEN parallelisation patches in my 'devel' branch. I'd expect them to > give a slight performance improvement with your original test (i.e. > the one that doesn't play the UID/GID games) > Ok, I haven't tested that yet, but I did put together a f20 machine on bare metal to act as a server. Here are some results from a test that forks 128 child processes and has each one create and close 1024 files. Client is 3.14.8-200.fc20.x86_64 and using NFSv4.1: With a stock Fedora kernel on the server running 3.14.8-200.fc20.x86_64: $ time sudo ./opentest -n 128 -l 1024 /mnt/palma real 59m5.022s user 0m1.401s sys 0m17.561s ...with the patched kernel 3.16.0-rc2.jlayton.1+ (basically Bruce's for-3.17 branch with the rest of my patches piled on): $ time sudo ./opentest -n 128 -l 1024 /mnt/palma real 4m19.060s user 0m1.259s sys 0m13.059s ...so around a factor of 13x speedup on parallel creates. Again, this is not a terribly scientific benchmark, but it does suggest that this patchset helps immensely. -- Jeff Layton <jlayton@xxxxxxxxxxxxxxx> -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
