I've installed System Center Data Protection Manager 2012 SP1, created a Storage Pool assigning five hard disks.
What would happen if one of those drive suddenly crash and become totally useless?
Is there some kind of data redundancy that prevent the loss of my backup in such case?
In which extend it's working ? (i.e. how many dead drives it can recover)
The storage pool is a set of disks on which the Data Protection Manager (DPM) server stores replicas, shadow copies, and transfer logs. Before you can start protecting data, you must add at least one disk to the storage pool. Disks added to the storage pool should be empty. To prepare for data protection, DPM reformats the disks and any erases any data on them.
The DPM server must have at least two disks installed — one dedicated to the startup, system, and DPM installation files, and one dedicated to the storage pool. In the context of DPM, “disk” is defined as any disk device manifested as a disk in the Windows Disk Management tool.
(Meaning it can be a RAID array that isn't software-based RAID.)
DPM will not add any disk containing startup files, system files, or any component of the DPM installation to the storage pool.
The level of fault tolerance/redundancy isn't set by Data Protection Manager. The fault tolerance/redundancy is provided by the storage system that you've set up. This can be a RAID array, which is an array of mirrored, striped, or distributed parity array of disks. A RAID array can be provided by a dedicated hardware card, or it can be software driven. (But keep reading because DPM doesn't work with software RAID.) Either way you go, the idea is that you take multiple physical disks, and create an orchestrated array out of those disks such a single logical storage device is presented to the operating system.
So if you assign just a single physical disk as part of the storage pool, and you lose that disk... then there is no inherent redundancy in that. It was just one disk, and now it's gone. That means any backups that were stored on that disk are now gone.
You'll want to do some RAID config with your multiple disks for any serious level of redundancy.
Here is the canonical Serverfault question on RAID:
What are the different widely used RAID levels and when should I consider them?
OK, so Microsoft also says this about DPM:
Do not create a software-based RAID configuration on disks that you will add to the storage pool."
From http://technet.microsoft.com/en-us/library/ff399569.aspx
As well as:
DPM does not support software-based configurations of RAID.
From http://technet.microsoft.com/en-us/library/cc161569.aspx
Oh, and here's the one for DPM 2012 specifically, though it hasn't changed:
http://technet.microsoft.com/en-us/library/hh757992.aspx
So, yeah, if you want any amount of fault tolerance and/or redundancy in your backup data with DPM... you have to go with hardware-supported RAID.
I'm sorry... I think Microsoft jerked our chains a little bit with the contradictory documentation there.
All that being said, even if you cannot do hardware RAID for whatever reason and only have a JBOD (just a bunch of disks) configuration... well, it's pretty easy to replace a bad disk and rebuild backups from scratch. So there's that...
If you have a sufficiently complex backup system that it's not easy for you to just rebuild them all when you need to, then there's really no excuse for you not being able to build a hardware RAID either, regardless of platform or product.
