Skip to Content
avatar image
Former Member

The big difference between backup time on Windows and Linux SAP systems

Oct 09, 2017 at 07:29 AM | 83 Views

Hello to all! Please help me with my question, I have two systems:

ERP (Windows 2k8)

EWM (RHEL 6)

And backup server on Debian 6 with SAMBA, so I make a backups on network share.
I can't understand why ERP backup produced 8 hours, and EWM - 0.5 hour. Backup server is the same, datastore is the same, this virtual machines placed on the same host and it has the same performance (cpu and ram). Database size = ERP 1TB/ EWM 0.5TB, exec_parallel =16. Only difference is OS, could it be an answer? Thanks!

screen-shot-2017-10-09-at-101602.png (21.8 kB)
screen-shot-2017-10-09-at-101945.png (21.7 kB)
Oracle Database
Add comment
10|10000 characters needed characters exceeded

  • Get RSS Feed

4 Answers

  • Best Answer
    avatar image
    Former Member
    Oct 18, 2017 at 01:22 PM

    Thanks James! With your help I realise that SAP kernel on Windows uses mkszip and on Linux uses gzip.

    Compression command looks like :

    Windows: compress_cmd = "mkszip -c $ > $"

    Linux: compress_cmd = "gzip -c $ > $"

    Gzip default compression level is 6,

    -# --fast --best<br>Regulate the speed of compression using the specified digit #, where -1 or --fast indi‐ cates the fastest compression method (less compression) and -9 or --best indicates the slowest compression method (best compression). The default compression level is -6 (that is, biased towards high compression at expense of speed).

    and mkszip default compression is 9

    <em><strong>-l</strong> comp_level<i>sets the level of compression to comp_level. Higher values lead to greater compression but result in more time taken during the compression process. The value comp_level may be an integer from 0 to 9. <strong>The default is 9</strong>.</i></em><i></i>

    So the problem was in default level of compression which SAP do not manage. Now I am trying to find the optimal compression level.

    Add comment
    10|10000 characters needed characters exceeded

  • avatar image
    James Zhang
    Oct 10, 2017 at 12:44 AM

    Hi Igor,

    The backup time is mainly depending on size, hardware/i/o, since you have different OS, so I assume this could be a proper different reason.
    You may check from OS or I/O point of view.

    Best regards,
    James

    Add comment
    10|10000 characters needed characters exceeded

    • Former Member
      Oct 10, 2017 at 06:08 AM

      Hello James!

      The problem that I use the same storage for both systems, so I think that isn't I/O question, and I can't believe that this OS can make so big difference. How can I check it ?

  • avatar image
    James Zhang
    Oct 10, 2017 at 08:34 AM

    Hi Igor,

    Then please perform testing:

    1. put a big file to oracle data file directory
    2. use cp or os command to manually move the testing file
    3. then please check whether the speed is the same from both the system

    By doing this, we can roughly knows the pure OS speed to back up.

    Best regards,
    James

    Add comment
    10|10000 characters needed characters exceeded

    • James Zhang
      Oct 13, 2017 at 12:20 AM

      "copy processes together", can you please let me know the details on this?

      Also please attach the faster and slower's brbackup trace, let us compare them again, whether it is certain data file backup was slow.

  • avatar image
    Former Member
    Oct 16, 2017 at 02:07 PM

    Hello James!

    "copy processes together":

    • I made two copies of a big datafile and send it to my servers (into oracle directory)
    • I started this copy processes simultaneously

    I attached *.and files from two servers.

    Add comment
    10|10000 characters needed characters exceeded