From Linux NFS

Revision as of 04:01, 27 August 2007 (view source) Chucklever (Talk | contribs) m (Reverted edits by IebKds (Talk); changed back to last version by Chucklever) ← Older edit	Latest revision as of 12:14, 16 July 2009 (view source) Trondmy (Talk | contribs) m (Reverted edits by AcelrOorli (Talk) to last revision by Chucklever)
(One intermediate revision not shown)

---

Latest revision as of 12:14, 16 July 2009

Contents 1 NFSv4 Robustness testing 1.1 testing Core file operations 1.2 Testing mount / export 1.3 Links 2 Main results

NFSv4 Robustness testing

testing Core file operations

There are 3 tools used to test NFSv4 robustness. This tools are used to tests core filesystems functions : read/write/create ... in various configurations.

• fsstress : random sequence of FS operations performed by concurent processes.
• fsx
• ffsb

Each time a NFSv4 is able to complete one test sequence, this test should be conserved for non regression testing.

Testing mount / export

Here is a small program to test that.

This test show nfsv4 can reach 2500 exported/mounted directories on a ia32 machine. When 2500 directories are exported/mounted we can see:

• rpc.mountd uses 40% CPU
• exportfs uses 14% CPU

Links

List of tests tools that should be usefull to test NFSv4.

Main results

Last kernel tested is Linux 2.6.12-rc3 more patches CITI_NFS4_ALL-4.

NFS v4 is able to complete this testes:

• fsx
• fsstress running 5 days in this configuration
  • 10 threads
  • 100 operation list lenght
• FFSB running 5hours with this profile:
  • num_files=1200000
  • num_dirs=100
  • max_filesize=4096
  • min_filesize=4096
  • num_threads=10
  • read_weight=3
  • write_weight=1
  • append_weight=1
  • write_random=1
  • write_size=4096
  • write_blocksize=4096
  • read_size=4096
  • read_blocksize=4096