Readdir ls -f graphs
From Linux NFS
(Difference between revisions)
(→NFS v4) |
(→NFS v3) |
||
| Line 24: | Line 24: | ||
* [[File:v3_ls_f_noplus_sys.jpg]] | * [[File:v3_ls_f_noplus_sys.jpg]] | ||
* [[File:v3_ls_f_plus_sys.jpg]] | * [[File:v3_ls_f_plus_sys.jpg]] | ||
| + | ** Using readdir plus for this command takes longer than using a simple readdir. | ||
| + | |||
=== NFS v4 === | === NFS v4 === | ||
* [[File:v4_ls_f_noplus_sys.jpg]] | * [[File:v4_ls_f_noplus_sys.jpg]] | ||
Latest revision as of 19:14, 7 April 2011
36: Linux 2.6.36
36NORC: Linux 2.6.36 (without readdir cap)
38: Linux 2.6.38
38RC: Linux 2.6.38 (with readdir cap added back)
38NB: Linux 2.6.38 (with Neil Brown's patch and loop detection)
Contents |
Real
NFS v3
-
- 2.6.36 based kernels seem to be slightly slower than 2.6.38 based kernels.
-
- 2.6.38 performs the same as 2.6.36 with the readdir plus cap removed. Using readdir rather than readdir plus is faster in this case.
NFS v4
-
- 2.6.36 based kernels take more time to finish the command than 2.6.38 based kernels.
-
- 2.6.38 based kernels can complete this command much faster than 2.6.36 based kernels.
Sys
NFS v3
-
-
- Using readdir plus for this command takes longer than using a simple readdir.
NFS v4
-
-
- Unmodified 2.6.38 gives the worst performance here. All other kernels seem to give roughly the same results.
RPC ops
NFS v3
-
- 2.6.36 kernels issue fewer RPC operations than 2.6.38 kernels
-
- Kernels with readdir plus limiting seem to have fewer RPC operations than kernels without the limit.
NFS v4
-
- 2.6.36 based kernels issue fewer RPC calls than 2.6.38 kernels.
-
- 2.6.36 based kernels give the fewest number of RPC calls. 2.6.38 with readdir plus limiting issues a few hundred more RPC operations, but not nearly as much as an unmodified 2.6.38 kernel.
