If your servers have local disk, it is possible you have mirrored system disks, in which case, the system can be recovered in minutes in the event of a disk failure. But for the restore of a single file due to deletion or corruption, the mirrored disk will be of no benefit.
To determine if snapshotting and replicating data can provide the same level of data protection as backup applications, we have to understand what backup applications offer. Backup applications can be said to offer a form of scheduling tasks which, at defined times, will run a set of rules. This will culminate in the transfer of data from a host to labelled media, while recording the transferred files in some sort of database which can be queried for restoration purposes over days, weeks, months or years.
Arrays basically offer one of two methods of protection spanning hours or days: snapshot or full copy. With a snapshot, a quick photograph is taken of the data. With a full copy, a complete copy of the data is taken).
Advantages of snapshots
Snapshots have two advantages: the speed at which they're created (allowing applications to be quiesced during the snapshot at minimal disruption to the application – quicker than a backup application) and the minimal disk space required per snapshot (at a small percentage of the capacity of the source data), allowing many snapshots to be taken. The time in which a system can be recovered (RTO) in the event of a failure is reduced, along with the point (RPO), depending on how often snapshots are taken (dependent on costly array capacity!).
The disadvantages? The reliance on the source data which snapshots point to, and capacity for the number of required snapshots. Essentially, this will suffice for hours (or, with luck, a couple of days), but it's not a long-term solution.
Advantages of full copies
A full copy requires an equal capacity of the source data, but with the advantage of containing a full copy of the source data, with no reliance on the source. The disadvantage? The space required per full copy and the time taken to perform the copy -- which may not allow for applications to be quiesced. Again, the physical size of each full copy will depend on costly array capacity, so it will not provide recovery points spanning more than days.
On the other hand, NAS devices perform snapshots and replication based on files changed, allowing for the individual recovery of a file -- an obvious advantage over an array.
NAS can fulfil single file recovery. However you have to ask yourself if NAS is good enough for your tier 1 storage, which for most of you will be based on modular or enterprise arrays.
If you were designing the mother of all backup applications, you would utilise array-based copies for short-term speedy RTO and shortened RPO timelines, with a backup application to fulfill restoration requests spanning weeks, months and years.
About the author: Hywel Matthews is a senior consultant at Glasshouse Technologies (UK), a global provider of IT infrastructure services, with more than 12 years experience in the IT industry and nine years experience in backup, recovery, disaster recovery, systems and storage.