Test conditions
This test aims to document the performances of NFSv4 running on different filesystems on the server. Iozone has been run four times. Each time, the directory exported by the server uses a different filesystem between:
kernel is 2.6.9.
|
write
|
rewrite
|
read
|
reread
|
randwrite
|
bkwdread
|
recrewrite
|
strideread
|
fwrite
|
frewrite
|
fread
|
freread
|
Comments
|
NFSv4
filesystem
|
|
|
 |
|
|
|
|
|
|
|
|
|
results
for rsize=32kb
|
local
filesystem
|
|
|
|
|
|
|
|
|
|
|
|
|
No
latence between 2 tests.
rsize=32kb
|
Warnings
- Tests are done with read
size of 32KB (default NFSv4). You
should not extrapolate these results
for other
read/write sizes. In
particular you should not use these results to optimize local
applications running on a local filesystem.
- Be aware to the different
scales used for NFSv4 and local
performances.
Conclusion
NFSv4 performances do not depend on the local file-system used.
Software
configuration
Local
benchmarks
- iozone is ran
over
each partition.
- iozone command
:
iozone
-ace -f /mnt/filesystem_XXX/fichier_test
NFSv4
benchmarks
-
Server:
- export is /export.
- A /mnt/filesystem_XXX
is
binded on the
/export directory.
- export options are :
/export
<world>(rw,wdelay,insecure,no_root_squash,no_subtree_check,fsid=0)
- Client:
- iozone command: iozone
-+q
1 -ace -U /mnt/benchs_nfs -f
/mnt/benchs_nfs/fichier_test
hardware
configuration (Client and server)
2 processors : Intel(R)
Xeon(TM) CPU 2.80GHz, cache 512 KB
Total memory: 2Gb
Ethernet: 1Gb/s link
Distribution: modified Fedora Core 2
