Backup Advice-Recommendation Needed

[rciscon]

Our company has decided to transition from Backup Exec to Veeam for backups of our VMware environment and I need some advice on how to implement our regular backups. Here's a quick precis of our environment:

3 - Clustered ESX Hosts running vSphere v4.0 Update 1
All VMs connected to a NetApp FAS3140 SAN via NFS
90% of Production VMs are running on Tier 1 (FC) disk. The remaining VMs and all backups will be stored on Tier 3 (SATA) Disk.

The environment is:
Veeam Backup & Replication v5.0.1.198 64 bit running on a VM with 4 virtual CPUs, 4GB of RAM with Windows 2008 R2 (64-bit).

My Goal:
Backup VMs running on the same ESX cluster as the Veeam B&R VM to Tier 3 disk on the NetApp SAN via NFS.

Here are the two scenarios for this that I can determine. Can you provide recommendations on which scenario will provide the best performance, would be safest, and follows "Best Practices":

Scenario 1 - VMFS disk containers directly connected to the Veeam VM:
In this scenario I carve out NFS volumes in VMware and directly connect them to the Veeam VM as additional hard disk drive volumes, i.e. Drive G:, Drive H:, Drive I:, etc. I can then create Veeam Backup jobs and point them at the Windows volumes mounted in the VM and perform the backups in this manner. Does Direct SAN Access "Processing Mode" work in this scenario?

Scenario 2: - Write directly to a NetApp NFS volume:
In this scenario, I create a large NFS volume on the NetApp and mount that NFS share to all three of the ESX Hosts in the cluster. In the Veeam Backup Job Wizard, I choose Direct SAN access as the Processing Mode, and select the source from the Virtual Center server. For the Backup Destination, I choose one of the three ESX Hosts and select a path like this: "/vmfs/volumes/veeam_nfs_target_volume".


For either scenario, my preferred situation would be to have a backup that I could restore as either a full VM or drill into and restore individual files. For this scenario, using "Reverse Incremental" seems to provide the best of both worlds. Are there any downsides to using Reverse Incremental, i.e. disk space usage, safety, etc?

Other questions:

Should I use Veeam deduplication, should I rely on NetApp's ASIS deduplication on the NFS volume itself, or do both?

If neither of these scenarios would be considered "Best Practices", can you recommend a scenario that is?

Thanks in advance,

Ray Ciscon

[Vitaliy S.]

Hello Raymond,

First of all, provided that you're using NFS datastore to host your production VMs, I would strongly recommend applying a VMware patch that improves snapshot removals behavior: vSphere 4.1 snapshot removal related patch (NFS)

Also you definitely want to use Virtual Appliance mode with NFS production storage, as based on the existing feedback you should get a pretty decent speed with a setup like yours.

As to SAN mode then it will be possible if you could configure direct SAN access using Microsoft iSCSI Software Initiator bundled with Windows. Not sure it is a way to go since your production is on NFS storage already.

Besides, I'm not sure it is worth choosing ESX host as destination target for your backups. What would you do if this host goes down? If I were you I would most likely mount this NFS LUN to a Linux box and then choose it as a destination storage.

In regard to what backup mode to choose...well basically reversed incremental backup mode is quicker to restore to the latest VM state, on the other hand forward incremental mode is at least 3 times faster than reversed incremental because of 3x less I/O on the target storage. Should more details be required, please have a look at our one-to-one feature comparison (see v5 Choosing Backup Mode.zip)

Finally I wouldn't recommend disabling Veeam deduplication mechanism (use both).

Hope this helps!