Subscribe

FlexClone vs A-SIS Deduplication for Filesystem-Based VM's

I have a series of ~10 virtual machines, all of which share a common history.  The machines are have Debian Linux filesystems, the base of which is about 30G with 1.8m individual files.

  1. I could rebuild a "master" machine, FlexClone it 10 times, then rebuild each VM's personality.  Each VM will have it's own volume with size limit. New data won't be scanned for duplicates.
  2. I could stuff all 10 existing machines into a single flexible volume (e.g. /vol/xen_vm/machine1 /vol/xen_vm/machine2 ), then use A-SIS deduplication to bring the size down.  I'll be stuck with a limit of 500G total on an FAS2020.  All volumes will share a single size limit.  But any new data added will be scanned for duplication.  My understanding is that A-SIS works only on a single volume.

Then

  1. I wish to SnapMirror backup the entire set.

I'm seeking some experience or thoughs on these options.  I have read http://blog.scottlowe.org/2008/08/05/flexclones-versus-deduplication-with-vmware-infrastructure/

The same question is at https://forums.netapp.com/message/18419#18419 (I'm a bit fuzzy on why there are two separate netapp supported forums)

Re: FlexClone vs A-SIS Deduplication for Filesystem-Based VM's

Various comments....

  • Don't forget single-file flexclone -- you can use that inside a volume and may be exactly what you want here.
  • 2020 Dedup Volume Size Limit -- on 7.2.x the 2020 is limited to 500 GB volume size for dedup. Under 7.3.x that limit is raised to 1 TB.
  • Dedup and file count -- given dedup works at a block layer and the virtual machines are encapsulated inside virtual disk files, dedup/ONTap won't care at all about the file count inside the virtual machine.
  • SnapMirror -- no issues there if doing volume SnapMirror as that works at the block level (so if data is already deduplicated it's not "inflated" for transfer).

I'd probably go with putting them all inside a single volume and potentially look at single-file FlexClone to make that easier.

Re: FlexClone vs A-SIS Deduplication for Filesystem-Based VM's

Thanks for the 7.3 note on the size.

But single file FlexClone won't work.  VMWare uses a large virtual file per VM.  With Xen I have thousands of files per VM.

We'll be going with A-SIS, and lamenting that since all the VM's will be part of the same volume, a "disk full" condition in any one of them will bring them all down.