Virtualization, NAS or SAN?

[ralfaro01zero]

Seems like nowadays there are more benefits from going NAS when virtualizing than going SAN since 10Gb Ethernet prices are similar to Fibre Channel and offers more benefits and simplicity (my humble opinion). I find myself involved more often that I would like to in discussions with people defending Fibre Channel over Ethernet storage, and now FCoE shows up to make things even more complex (and expensive); why not just make your life easier and go NFS? or iSCSI if you want SAN?

What is your opinion? Are you SAN or NAS?

Cheers.

Ricardo.

[scottgelb]

But nice that NetApp can do it all. I say coexist. San boot of servers is nice to have. NAS has many operational efficiencies. But some workloads work better block and some don't.

I have a NAS tendency whenever possible. Direct nfs with oracle and nfs with virtualization is great. Rumors are the next MS SQL can run over the new smb.

For some this is a Chevy vs Ford or Pepsi ve coke debate

[ralfaro01zero]

I heard about Hyper-V supporting CIFS as VM storage as well, seems like everybody is going that way.

You can PXE boot your ESX severs (and most Linux/Unix servers out there) from NFS too, i have done this several times with CentOS and it works great, I have yet to try this with VMWare, but it is supported AFIK.

There's a blog from a guy named Chris Wolf about PXE booting ESXi servers, it is old, but the procedure hasn't change much since it was written: http://www.chriswolf.com/?p=18

The procedure is similar on all Linux flavors.

[scottgelb]

I just heard about PXE being support with ESX 5 now… haven’t tried it yet but very cool.

[arunchak]

Few advantages of NFS:

• Provisioning is a breeze
• You get the advantage of VMDK thin Provisioning since it's the default setting over NFS
• You can expand/decrease the NFS volume on the fly and realize the effect of the operation on the ESX server with the click of the datastore "refresh" button.
• You don't have to deal with VMFS or RDMs so you have no dilemma here
• No single disk I/O queue, so your performance is strictly dependent upon the size of the pipe and the disk array.
• You don't have to deal with FC switches, zones, HBAs, and identical LUN IDs across ESX servers
• You can restore (at least with NetApp you can), multiple VMs, individual VMs, or files within VMs.
• You can instantaneously clone (NetApp Flexclone), a single VM, or multiple VMs
• You can also backup whole VMs, or files within VMs

<source> http://storagefoo.blogspot.in/2007/09/vmware-over-nfs.html (suggest you to go through this blog... whatever he talks makes real sense).

  So if NFS, NetApp does Lead

Still the choice of vmfs vs NFS is under your descretion.

-Arun

[ralfaro01zero]

I would like to add:

• Higher dedupe ratios
• You are not locking your data to a specific vendor
• Ethernet storage roadmap seems better than FC (100Gb Ethernet is not far away)
• Single, shared infrastructure (no more 2 teams doing basically the same thing, Networking guys and SAN guys connecting servers to each other)

[radek_kubka]

Are you talking about Autodeploy? It has few wrinkles in my opinion - it is 100% stateless, so e.g. you need to use Host Profiles to properly customize your ESX hosts once they boot, and forwarding logs is also beneficial (otherwise they are stored in memory & lost upon a reboot).

Nice write up could be found here:

http://www.gabesvirtualworld.com/my-first-auto-deploy-design-for-real-production-environment/

[radek_kubka]

I will repeat myself , as just responded in the same way in a different thread:

NFS is cool, but in most environments it is not enough - if SnapManager products are used, then disks with application data would normally be provisioned as Raw Device Mappings (over iSCSI or FC) - with one corner case of SnapManager for SQL, which can handle NFS VMDKs (well, and just Oracle sprang to mind, which can leverage NFS too).

Regards,
Radek