i have to ask because i could not find a KB for this.
I have created an off-host proxy with storage snapshots (3PAR) and it backups well (most of the time).
But the main reason i configured this was i expected it would work like this:
But with Hyper-V the snapshot avhdx is still filled with data and not merged right after the snapshot (and the storage snapshot). So where is the big improvement for me then? I backup from the storage snapshot but i could also just take the "main" vhd. Simply makes no difference.
There is indeed a big difference between Hyper-V and VMware. The idea of an off-host backup proxy is to take data processing of the production Hyper-V host. In this mode, the off-host shifts the backup and replication load to a dedicated machine (the off-host proxy), Basically, it retrieves the VM data from the source, processes it and then transfers it to the destination.
This process is not the same as Backup from a Storage Snapshot integration that we have with VMware. The underlying backup architecture on Hyper-V (until 2012R2) is also completely different. Maybe in 2016, with the new backup architecture, things could change but that would require a few worlds come together
thanks for this quick answer.
Just having problems with the big VMs that generate a lots of data while backing up. (SQL Sharepoint (2TB) and Exchange Server (7TB)).
Is there any other way (except migrating to VMWare ) to speed things up?
What version of Hyper-V are you running. Might not be the easiest solution, but when you run Hyper-V 2016 and the VM's on ReFS you can take advantage of this. ReFS snapshot (sorry, checkpoint ) merge is significantly fast (because it is metadata operation, not I/O operation anymore) and causes less issues that way.
As I said, it is not the easiest solution, but it is the solution forward and makes a lot of difference for your virtualized environment
thank you. I will consider this for the SharePoint Machines this year. I think i will push Exchange to the cloud.
(Also took me a long time to say checkpoint instead of snapshot )
Just one other question about the storage snapshot. Why does it even use this technique with Hyper-V?
If the merge does not happen right away, you could just use the VHD file from the LUN?
I understand perfectly your confusion, but unfortunately it doesn't work like that, hence the reason why each storage vendor had to write a hardware VSS provider to make this even work through the hardware. As I said I can only hope we see some difference with 2016 but for now I don't see many hardware vendors doing something or even announcing something. One of the things that you need to keep in mind is that in non 2016 versions the VHD(x) is held by the VSS writer that won't release it until all the work is done. So even having the VHD on a lun, snapshot (hardware) that LUN would not work. Even worse, pre-2012 R2 things get even more complicated as every VHD(x) had to be mounted and rolled back to a certain point in time just to make the Guest and Host snapshot at the same time (which is basically unworkable with too many VM's on a host).
) to speed things up?