1

We all know that once a disk (or storage system for that matter) gets introduced into use, the performance degrades due to fragmentation of files. This seems to be why disk defragmentors are in fairly wide use on Windows boxes. And they do increase the performance substancially. As an aside, I haven't heard of many defragmentors in the Unix/Linux area.

Despite the claimed WAFL protections for the NetApp, file fragmentation still will occur, especially with all the sparsely crated VMs. My question is does anybody do any sort of defragmention of such a storage system? Do you notice any measurable degration/improvement of either not doing/doing anything to address this situation? Does anybody do anything about it? If so what?

Thanks

Improve this question
1
  • Huge topic - for example, are you using NFS or SAN. Have you ensured block alignment on your VMs? Are you using any cloning technologies? NetApps have historically had performance degredation - primarily when aggregates/volumes get full, but it's unlikely to be your biggest concern running a virtualised environment.
    – ChrisH
    Jul 25, 2012 at 17:10

3 Answers 3

Reset to default
3

Potentially unrelated, but if you are using NetApp filers as storage for VMs you need to be aware of the I/O alignment issues outlined in TR-3747. In a scenario with multiple misaligned VMs hosted on one filer, the potential impact on iops would probably be a greater concern to me than fragmentation.

I've personally spent a lot of time recently realigning practically every VM in our infrastructure because proper alignment of VM partitions was not accounted for in the initial build-out.

Improve this answer
1
  • You are correct that alignment issues are a BIG problem in any sort of virtual infrastructure, although it's not just limited to NetApp.
    – Jeremy
    Sep 23, 2010 at 23:33
2

Have a read of To Defrag or Not to Defrag and Should I defragment SAN-based volumes?.

Are you talking about NetApp CIFS or NFS volumes presented to a Windows host or NetApp FC/iSCSI LUNs?

Yes they are all built on top of WAFL but in the case of the LUNs you are being presented with a raw block device which is then formatted by the host. This means that the fragmentation of these file systems is dependant on the host to maintain. I believe WAFL will still do some of its block deframentation but not at the logical file system layer.

With CIFS and NFS volumes then the file systems they are on are NetApp's own and host file system defragmentation doesn't come into play.

Should you defrag your Windows VMs that live on a NetApp filer? Good question, I guess it depends on if you see a performance benefit from doing it regularly compared the amount of IO that it takes to defrag them.

Improve this answer
3
  • 1
    The use is of the NetApp that I have experience is purely NFS and CIFS volumes ONLY. BTW, they are presented to all sorts of hosts Windows, and several commercial UNIX and several varients of Linux.
    – mdpc
    Feb 25, 2010 at 5:40
  • Just to makes this clear: on a NetApp all LUNs are on WAFL data structures(their're essentially files) that adhere to the same WAFL allocation mechanisms as any other data that resides on WAFL.
    – pfo
    Dec 14, 2010 at 14:35
  • @pfo a WAFL based block device that is formatted with NTFS would still be subject to fragmentation at the NTFS layer, WAFL allocation notwithstanding.
    – Sim
    Dec 15, 2010 at 1:30
2

Just run/schedule reallocate on your LUNs and they should stay fine.

Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .