Last changes:
Do Feb  5 09:40:02 CET 2004

• Choose a set of file systems to test (eg. reiser, reiser -o notail and ext2)
• Choose the directory structure for agesystem (eg. 1x1) and choose the file size distribution for agesystem (eg. the standard for a 1 GB partition) or measure the distribution of files per directory and file sizes according to an interesting directory (eg. /usr/src/linux or your home directory).
• Choose the minimum and maximum percentage between the file system should be aged (eg. between 60-80 %)
• Choose the number of aging runs (eg. 10)
• Choose the number of files to create through each aging run (eg. 1000)
• Loop over the file systems to test
  • Format partition one
  • Gauge it up to the chosen minimum percentage
  • Loop over the aging run number
    • Age partition one using the agesystem3 tool with the set of chosen options
    • Format partition two and copy partition one onto it (cp -R . /partition2)
    • Run fibmap to get the written block layout on both partitions
    • Run read to measure the read performance on both partitions

You may download the necessary files and try it yourself. After downloading the package create a directory where to untargz it. You probably need to adjust some paths, then type ''make agesystem; make fibmap; make read'' and look for the script ''agetest.sh'' that shows how to setup the options described above. If you like to get some graphs out of the resulting output, send the output files to me (targzed, please!).

The hardware and settings are the same as in the gauge section.

• AMD Duron 650 Mhz, 128 MB, 40 GB EIDE-harddisk for the system with the tests taken on
  • Adaptec AHA-2940U2/W host adapter
  • 9 GB IBM DNES-309170W SCSI-harddisk

• Software: SuSE 7.2 (gcc 2.95-3) kernel 2.4.8-ac1 restricted to 32 MB memory without swap.

• The write performance is measured in KB per second. It is the number of bytes created in total divided through the total runtime of the agesystem3 tool including the time for the deletes (but not counting the deleted number of bytes, maybe I should think of including it).
• The read performance is measured in KB per second measured with the timer starting just before the read system call and stopping right after it while recursively walking through all files in all directories.
• The random read performance is measured in KB per seconds. The files to be sequentially read are randomly chosen and the timer is started and stopped before and right after the read system call.
• The fibmap read performance is measured in KB per seconds. It is the read performance when the files are read in the order which is given by their logical block numbers sorted in an ascendant order. Files with logical first block number of zero (eg. tails) are omitted.
• The meaning of the fragmentation values is stated in the fibmap section.

• Partition Size: 1 GB
• File systems: Reiser, Reiser with -o notail and Ext2
• File size distribution: generic (-l 6:a:b:c:d:e:f option) with

  min_file_size: 2048
  mean_file_size: 65536
  min_mfile_size: 65536
  mean_mfile_size: 524288
  min_lfile_size: 524288
  mean_lfile_size: 4194304
  

• Directory structure: 1x1 (generic -c 1:1)
• Maximum number of runs: 25 (including the gauge run)
• Number of creations per run: 10000
• Minimum usage: 65%
• Maximum usage: 85%

• Ext2 (Ext2 copied)
• Reiser (Reiser copied)
• Reiser -o notail (Reiser -o notail copied)

The hardware and settings are the same as in the gauge section.

• AMD Duron 650 Mhz, 128 MB, 40 GB EIDE-harddisk for the system with the tests taken on
  • Adaptec AHA-2940U2/W host adapter
  • 9 GB IBM DNES-309170W SCSI-harddisk

• ReiserFS with and without tails with and without the bitmap hinting patch
• XFS (linux-2.4.10-xfs-2001-10-03.patch)
• JFS (jfs-2.4.7-1.0.7-patch
• Ext2 on a
  2.4.10 kernel (patched with the mentioned patches).

• Software: SuSE 7.2 (gcc 2.95-3) kernel 2.4.10 restricted to 32 MB memory without swap.

• The write performance is measured in KB per second. It is the number of bytes created in total divided through the total runtime of the agesystem3 tool including the time for the deletes (but not counting the deleted number of bytes).
• The file history is the fraction of the number of files created at run zero divided by the number of files which are still left at run x.
• The read performance is measured in KB per second measured with the timer starting just before the read system call and stopping right after it while recursively walking through all files in all directories.
• The random read performance is measured in KB per seconds. The files to be sequentially read are randomly chosen and the timer is started and stopped before and right after the read system call.
• The fibmap read performance is measured in KB per seconds. It is the read performance when the files are read in the order which is given by their logical block numbers sorted in an ascendant order. Files with logical first block number of zero (eg. tails) are omitted.
• The meaning of the fragmentation values is stated in the fibmap section.

• Partition Size: 1 GB
• File systems: Reiser and Reiser with Bitmap Hinting Patch
• File size distribution: generic (-l 6:a:b:c:d:e:f option) with

  min_file_size: 2048
  mean_file_size: 65536
  min_mfile_size: 65536
  mean_mfile_size: 524288
  min_lfile_size: 524288
  mean_lfile_size: 4194304
  

• Directory structure: 1x1 (generic -c 1:1)
• Maximum number of runs: 80 (including the gauge run)
• Number of creations per run: 500
• Minimum usage: 71%
• Maximum usage: 79%

• absolute compare RFS and BMP
• compare RFS and BMP normed
• BMP details
• RFS details