Forum Discussion

manatee's avatar
manatee
Level 6
8 years ago

restore speed question

NBU 7.7.3

two scenarios, when an RMAN restore was done directly from a disk partition, it took under 4 hours to restore a db of size around 1.6TB. unable to find out the transfer rate though.

when an RMAN restore was done from NBU, the transfer rate is around 4 to 6 KB/sec. so to restore a file size of about 1.6TB will take more than 4 hours. the backup image is still on disk.

both restore uses 3 streams to restore.

question is, why restoring directly from a disk partition is faster?

7.7
Solaris
  • manatee's avatar
    manatee
    8 years ago

    over the weekend, we found the source of the problem. it's the zfs storage. even file to file copy is slow.

  • Deb_Wilmot's avatar
    Deb_Wilmot
    Level 6
    8 years ago

    A few questions:

    Is the image on disk restore a local read/write?  

    Are you reading from the same disk partitions with the NBU restore and the direct read from disk?  Or is the netbackup on a different system?  

    Is it a single image that you are restoring from and is the image on disk the equivalent?

    Assuming the scenario where the data is stored on the same partition for both the basic read from disk and the restore through NBU - NBU is going to add overhead as the bptm process transfers the data to tar, and the processes also send meta data to the master server processes.

    The expectation is that NetBackup will add 10 to 15 % overhead to a backup or restore (possibly more or less depending on underlying hardware - this is just a ballpark figure).  This is due to the additional processing done which allows the applicaiton to track file metadata.

    If the time difference is within 15 percent, I would say that is within normal range.

    Hopefully this 'makes sense.'

     

     

     

    • manatee's avatar
      manatee
      Level 6
      8 years ago

      over the weekend, we found the source of the problem. it's the zfs storage. even file to file copy is slow.