ZFS-Tests

From GridPP Wiki
Jump to: navigation, search

This page is to track different configurations that were tested under different setups, raid systems, and file systems. If one needs to add comments to the own test setup, please use the running number and add a comment to it under the table.


for tool testscript.sh:

  • averaged sequential/parallel read: sum of all bytes over all files divided by total time taken
  • averaged sequential/parallel write: bytes per file divided by (total time taken divided by total number of files written)
  • parallel read/write default: 10 parallel reads and writes for large files, 20 parallel writes for small files
  • other default: 5.1 TB used of mixed files to read, 100x54GB and 200,000x131kB files written


Running number 1 2 3 4
Filesystem ZFS XFS Ext4 ZFS
Special filesystem config xattr=sa, relatime=on xattr=sa,relatime=on
Kind of raid 3xraidz2 raid60 raid60 1xraidz3
Number of Disks 33, 11 per raidz2 33, 11 per raid6 33, 11 per raid6 35 in single ZFS vdev
Kind of Disks 2TB SAS, 7200k, 6Gbps 2TB SAS, 7200k, 6Gbps 2TB SAS, 7200k, 6Gbps 2TB SAS,7200k, 6Gbps
Controller PERC H800 (33xraid0) PERC H800 PERC H800 PERC H800 (33xraid0)
Controller Cache used yes (512 MB) yes (512 MB) yes (512MB) yes (512MB)
RAM 12GB 12GB 12GB 12GB
CPU 2xE5620, 2.4GHz, HT on 2xE5620, 2.4GHz, HT on 2xE6520, 2.4GHz, HT on 2xE6520, 2.4GHz, HT on
OS SL6 SL6 SL6 SL6
normal averaged sequential read 536MB/s 856MB/s 607MB/s 422MB/s
normal averaged parallel read 746MB/s 538MB/s 533MB/s 543MB/s
normal averaged sequential large writes 709MB/s 667MB/s 348MB/s 677MB/s
normal averaged parallel large writes 680MB/s 314MB/s 140MB/s 660MB/s
normal averaged sequential small writes 10MB/s 2MB/s 2MB/s 10MB/s
normal averaged parallel small writes 74MB/s 15MB/s 20MB/s 69MB/s
degrated averaged sequential read 532MB/s 274MB/s 266MB/s 423MB/s
degrated averaged parallel read 707MB/s 280MB/s 298MB/s 529MB/s
degrated averaged sequential large writes 706MB/s 657MB/s 298MB/s 644MB/s
degrated averaged parallel large writes 695MB/s 319MB/s 141MB/s 653MB/s
degrated averaged sequential small writes 10MB/s 2MB/s 2MB/s 10MB/s
degrated averaged parallel small writes 74MB/s 17MB/s 15MB/s 68MB/s
rebuild, averaged sequential read 401MB/s 247MB/s 246MB/s 316MB/s
rebuild, averaged parallel read 533MB/s 267MB/s 274MB/s 417MB/s
rebuild, averaged sequential large writes - 367MB/s 239MB/s 215MB/s
rebuild, averaged parallel large writes - 279MB/s 129MB/s 216MB/s
rebuild, averaged sequential small writes 6MB/s 3MB/s 2MB/s 7MB/s
rebuild, averaged parallel small writes 31MB/s 17MB/s 19MB/s 41MB/s


Comments:

  • for 1) Files for read test where on compressed area of ZFS (root files, tgz files, LSST data files)
  • for 1) ZFS write test during rebuild not comparable since the rebuild was always faster finished then the test (in 2h compared to 21h for hardware raid)
  • for 1-4) all read tests based on the same mix of files
  • for 1-4) done with testscript.sh as explained above the table