Restoring a full VM methods

[theviking28]

I was wondering which method is optimal for fastest time to restore an entire vm.

1) Perform a full vm restore directly to production storage

2) Perform an "instant recovery" and then have either veeam or Vsphere migrate the vm to production storage.

I have read from a few places that using instant recovery and then leaving the vm powered off and migrating it to production is actually much better option than performing a normal full vm restore.

The reason I am asking this is that I have a file server that is 2tb in size. I am performing a full fm restore on it today just as a test. Veeam says it's going to take 3 hours to perform, and it's running at 150MB/sec which seems decent.

In a crisis, 3 hours without the file server isn't horrible I guess, but not great.

Backup repository= HP x1600 NAS with 12 2tb 7200rpm sata disks. connected via 1gb to the network
Production Storage= HP P2000 SAN direct attached SAS to esxi hosts.

Veeam proxy is a vm on one of the hosts, and it's on the same host as the vm I'm restoring so it IS using hotadd mode to do the restore and not network mode, which is nice.

[Vitaliy S.]

Hi Derek,

I would say that using Instant VM Recovery is a preferred option (in most situations), because of the following reasons:

1. Very low RTO. You get your production server back online very quickly. Yes, it will be running off the temporary datastore which might not have decent performance, but your users will be able to operate without any interruption.
2. If you have a vSphere Edition which allows live migration, then the disaster recovery process will be almost unnoticed. Using migration tools guarantees minimal downtime during switch from vPower NFS datastore to production storage.
3. Using full VM restore via hotadd proxy, I agree, is a very nice option. This allows to access datastore directly via ESXi I/O stack, however your services will be offline during the entire VM restore process.

Full VM restore can be used if you need to restore a small VM and downtime of this VM is not very critical to your production.

Hope this helps!

[theviking28]

Hi Vitaliy,

Thank you for the information. That certainly helps, but I do have a concern with Instant Recovery. I don't believe running the VM out of it is a realistic option for me as when I tried an IR with my file server, (2.0tb) it was so horribly slow that I don't believe that would be viable. Even a pretty small (50gb) vm runs really sluggish. Now, I do understand that it depends on my backup repository. (and mine is 12 2tb 7200rpm sata disks) in raid5... I wouldn't think it would be horrible compared to people who are using a small nas or something.

Anyway, my overall question is still this. If I did an instant recovery but did NOT power up the vm, and then I migrated that instant recovered vm to production storage and THEN turned it on... Would that be faster than a full restore of the same vm?

In other words, is hotadd speed of 150MB/sec any slower than what a veeam or VMware storage migration would run at?

[Vitaliy S.]

Yes, you're right that Instant VM Recovery heavily depends on the repository, raid 10 would give you a better performance.

As regards your question, then I believe hotadd restore should be quicker, as migration will most likely go via network stack.

[theviking28]

Thank you Vitaliy!

Very helpful!