• State Verified Answer
  • Locked Locked
  • Replies 3 replies
  • Subscribers 17 subscribers
  • Views 5098 views
  • Users 0 members are here
Options
Related
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

File-Level Restore of over 60k files in 1 folder - slow performance

leclaipa
over 6 years ago

Hey all.

I have an instance where I need to recovery about 120,000 1k-5k files between 2 folders (doing 1 folder at a time), but the restore process is taking an extraordinary amount of time.   I'm seeing a 1k file process in about 5-10 seconds with 2 of them running at the same time.

 

The CPU and memory load on the server is extremely low.  The back-end storage is on an ExaGrid storage device with SMB shares enabled.

   The Vizioncore.vRanger.Client.Shell.exe is hovering between 1% cpu to 3%.  The vznwinio.exe's (2 of them) go from 0-2%.

This is on 7.6.4 of vRanger.   Has anyone had a similar experience with a great # of files like this?

 

Thanks,

P

 

  • +1 over 6 years ago
    Observed this few times with FLR against web servers. This is usually the case - thousands of relatively small files.
    As I was told once, for Ranger it is easier to restore one 1MB file than 1000 1KB files. You remember everything is done on the fly. While it is transparent to the user, a lot of work is done in background: decompression, search for necessary blocks, reconstruction and so on.
    Another factor is if you are doing FLR against Diff or Incr savepoints: Ranger should merge data from the selected and parent SP(s). This is an extra overhead especially for Incr SP when data from multiple SPs in the chain all the way to the parent Full should be merged.
    Sometimes, if Hard Disk is of reasonable size, it will be faster to restore VM (probably with necessary disk), mount disk to available Windows VM and get necessary files.
  • 0 over 6 years ago
    This would be on a full backup and not an incr or diff. Yeah, I was afraid of that. It's the same type of thing I see with our enterprise backup solution that dumps to tape for our physical servers. Lots of 1k files take lots of time to recover.

    I guess I'll just have to wait it out.
  • 0 over 6 years ago
    You could also restore the entire VMDK file, mount it and copy the files over manually. a vmdk will restore more quickly than thousands of small files.