[Bugs] [Bug 1467614] Gluster read/write performance improvements on NVMe backend

[bugzilla at redhat.com]

https://bugzilla.redhat.com/show_bug.cgi?id=1467614



--- Comment #57 from Mohit Agrawal <moagrawa at redhat.com> ---
Hi,

After use splice call(for github
https://github.com/gluster/glusterfs/issues/372) on server side(only for write)
I got some performance improvement in writing of, though it is initial level of
testing but I hope we can get some improvement in IOPS if we use splice in both
sides (client and server and for both fops read and write).



Without splice patch

for i in `(seq 1 5)`
do
rm -rf /mnt/rpmbuild/glusterfs-4.0dev1/*
echo 3 > /proc/sys/vm/drop_caches
time cp -rf rpmbuild_build/glusterfs-4.0dev1/* /mnt/rpmbuild/glusterfs-4.0dev1/
done

real    22m51.189s
user    0m0.086s
sys    0m1.511s

real    22m57.531s
user    0m0.090s
sys    0m2.025s

real    22m57.845s
user    0m0.115s
sys    0m1.834s

real    22m51.257s
user    0m0.113s
sys    0m1.966s

real    22m57.857s
user    0m0.113s
sys    0m1.966s

# After applying splice patch

for i in `(seq 1 5)`; do rm -rf /mnt/rpmbuild/glusterfs-4.0dev1/*; echo 3 >
/proc/sys/vm/drop_caches; time cp -rf rpmbuild_build/glusterfs-4.0dev1/*
/mnt/rpmbuild/glusterfs-4.0dev1/; done

real    17m51.073s
user    0m0.104s
sys    0m1.862s

real    17m50.057s
user    0m0.097s
sys    0m2.005s

real    17m50.022s
user    0m0.096s
sys    0m1.928s

real    17m49.073s
user    0m0.101s
sys    0m1.828s

real    17m47.594s
user    0m0.077s
sys    0m1.753s


Regards
Mohit Agrawal

-- 
You are receiving this mail because:
You are on the CC list for the bug.
Unsubscribe from this bug https://bugzilla.redhat.com/token.cgi?t=uvZI1NSCKR&a=cc_unsubscribe