Difference between revisions of "MD BMG/roadmap"

From OpenSFS
Jump to: navigation, search
 
(3 intermediate revisions by the same user not shown)
Line 10: Line 10:
 
* First FS on 60 disk drives configured in 6 OST's with RAID-6 (8+2) and the MDT on a RAID-5 (4+1) flash devices
 
* First FS on 60 disk drives configured in 6 OST's with RAID-6 (8+2) and the MDT on a RAID-5 (4+1) flash devices
 
* Second on 20 SSD drives configured in 2 OST's with RAID-6 (8+2) and the MDT on a RAID-5 (4+1) flash devices
 
* Second on 20 SSD drives configured in 2 OST's with RAID-6 (8+2) and the MDT on a RAID-5 (4+1) flash devices
 
 
The results and some comments are in the pdf file format as follows:
 
The results and some comments are in the pdf file format as follows:
 
 
# [http://wiki.opensfs.org/images/3/38/MdtestSummary.pdf mdtest summary]
 
# [http://wiki.opensfs.org/images/3/38/MdtestSummary.pdf mdtest summary]
 
# [http://wiki.opensfs.org/images/d/df/Postmark_Results.pdf Postmark summary]
 
# [http://wiki.opensfs.org/images/d/df/Postmark_Results.pdf Postmark summary]
 
# [http://wiki.opensfs.org/images/d/d4/Netmist_Summary.pdf Netmist MPI summary]
 
# [http://wiki.opensfs.org/images/d/d4/Netmist_Summary.pdf Netmist MPI summary]
 +
# [http://wiki.opensfs.org/images/4/46/LUG14_MPEE_BWG_OpenSFS.pdf Slides for LUG14]
 
<br />
 
<br />
 
Please note that I used iostat for monitoring the utilization of the devices and that the metadata set was 3x larger than the cache size of the MDS.  
 
Please note that I used iostat for monitoring the utilization of the devices and that the metadata set was 3x larger than the cache size of the MDS.  
 
<br />
 
<br />
 
Return to [http://wiki.opensfs.org/MD_BMG Metadata Performance Evaluation] page.
 
Return to [http://wiki.opensfs.org/MD_BMG Metadata Performance Evaluation] page.

Latest revision as of 08:00, 28 February 2014

  • Collect benchmark tools candidates from OpenSFS
  • Evaluate all the tools and the workloads that can benchmarked
  • Recommend a small set of MD benchmark tools to cover the majority of MD workloads
  • Collect stats from users of MD benchmarks
  • Build scripts to allow ease of use of the recommended tools
  • Write documentation for troubleshooting MD performance problems using the toolset
  • Create a special website for MD tools


We ran 3 of the benchmarks proposed with similar configurations so we can compare the benchmarks. We ran the tests on 2 Lustre FS :

  • First FS on 60 disk drives configured in 6 OST's with RAID-6 (8+2) and the MDT on a RAID-5 (4+1) flash devices
  • Second on 20 SSD drives configured in 2 OST's with RAID-6 (8+2) and the MDT on a RAID-5 (4+1) flash devices

The results and some comments are in the pdf file format as follows:

  1. mdtest summary
  2. Postmark summary
  3. Netmist MPI summary
  4. Slides for LUG14


Please note that I used iostat for monitoring the utilization of the devices and that the metadata set was 3x larger than the cache size of the MDS.
Return to Metadata Performance Evaluation page.