Readdir ls -lU graphs

From Linux NFS

Revision as of 19:19, 7 April 2011 (view source) Amschuma (Talk | contribs) (→NFS v3) ← Older edit		Revision as of 19:22, 7 April 2011 (view source) Amschuma (Talk | contribs) (→NFS v4) Newer edit →
Line 16:		Line 16:
	=== NFS v4 ===		=== NFS v4 ===
	* [[File:v4_ls_lU_noplus_real.jpg]]		* [[File:v4_ls_lU_noplus_real.jpg]]
		+	** 2.6.36 kernels seem to take more time than 2.6.38 kernels.
	* [[File:v4_ls_lU_plus_real.jpg]]		* [[File:v4_ls_lU_plus_real.jpg]]
		+	** Using readdir plus on large directories helps the command finish faster.
	== Sys ==		== Sys ==

---

Revision as of 19:22, 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 1 Real 1.1 NFS v3 1.2 NFS v4 2 Sys 2.1 NFS v3 2.2 NFS v4 3 RPC ops 3.1 NFS v3 3.2 NFS v4

Real

NFS v3

• 
  • 2.6.36 kernels take longer than 2.6.38 kernels.
• 
  • Falling back to readdir operations for large directories makes this command take longer. Removing the readdir plus cap from 2.6.36 improves performance, but doesn't match 2.6.38 kernels. Adding a readdir plus cap to 2.6.38 doubles the time needed for this command to complete.

NFS v4

• 
  • 2.6.36 kernels seem to take more time than 2.6.38 kernels.
• 
  • Using readdir plus on large directories helps the command finish faster.

Sys

NFS v3

• 
• 

NFS v4

• 
• 
  • Using readdir plus on 2.6.38 kernels decreases time spent in the kernel

RPC ops

NFS v3

• 
  • Kernels issue the same number of RPC operations in this case
• 
  • Falling back to readdir on large directories can lead to thousands of extra RPC operations

NFS v4

• 
• 
  • Using readdir plus results in fewer RPC operations