what is the difference between snapshot and backup?

[gkman]

Hello,

this subject has probably been disccused quite extensively, but I am still not able to understand it.

I am familiar with the concept of snapshots from netapp storage system, but I cannot figure out what they mean in netbackup. 

I am new to netbackup, and looking through the activity monitor I see many jobs type of snapshot and right after a backup job. both belong to the same policy.

what is the difference between snapshot and backup? are they different types of backups or are they part of the same process? 

please explain in short preferably a dummy friendly explanation (please don't refer me to another 1000 page pdf, i'v read about three of those these past weeks) 

all help is deeply appreciated.

[sdo]

Maybe this may help:

http://www.computerweekly.com/feature/Backup-vs-replication-snapshots-CDP-in-data-protection-strategy

...and this old post, but the VSS principles are still relevant (ignore about VSP, which is no longer relevant):

https://www.veritas.com/community/forums/snapshot-explanation

[sdo]

Notice how the VMware snapshots are actually "write pending logs", whereas VSS (and thus Hyper-V) snapshots are actually "delta change logs".  Ultimately they achieve the same thing, but they work very differently:

https://www.veritas.com/community/forums/vmware-backup-failure-error-code-156

[Marianne]

Have a look at this current 'featured' Connect article: 

Copy Data Management is not a Replacement for Backup 

[sdo]

This looks useful, re different snapshot technologies:

http://searchdatabackup.techtarget.com/tip/Using-different-types-of-storage-snapshot-technologies-for-data-protection

[RiaanBadenhorst]

The short answer is, to get a consistent copy of the data (Vmware, Hyper-V, Exchange) the application needs to be stopped / paused momentarily so that we can get a "view" of what it look like at the specific point in time. Since backups are performed while the application, database, or virtual machine is online this snapshot process is critical to getting a consistent backup. Each of these applications have snapshots and all of them are different, and provided by the application vendor (for MS we use VSS, for VMware the VMware APIs/vCenter handle the snapshot).

So once you've got the snapshot it gets accessed and the data is backed up as it was at THAT point in time.

Ok so maybe the answer wasn't very short :p

HTH

[sdo]

Some say that a snapshot is not a backup.  And I have to agree.

Remember, as soon as a snapshot is instantiated, then any new writes (puts/updates/deletes) are effectively being squirelled away... and so, whilst the snapshot exists the 'pool/log/capture/delta' of write IO is growing all the time and will never cease until the snapshot is removed.  Plus, for simple snapshot technologies, the squirelled away list of writes is usually on the same storage... so, if this storage is lost, then you've lost your original point-in-time and all changes since the snapshot was instantiated.  Plus all snapshot technologies ultimately result in increased IO... think about it... all writes have to eventually be allowed/consolidated, but any overwrites have to first be captured.  This is why VM admins do not like long life snapshots.  This is why snapshots are not backups.

Whereas a backup is usually a copy to another storage medium or device, and can be transported or replicated off-site and does not cause all write IO to be accumulated on the original (or differencing store) storage.

There are some very advanced snapshot technologies these days, but they usually cost $$$$.

In summary then:

- a snapshot is usually an accumulating "difference" between a point in time, and now.

- a backup is usually a complete copy at a point in time.

[gkman]

so if I understand correctly, snapshots aren't acctually a netbackup process but rather a host (vmware, hyper-v) feature netbackup triggers as part of it's backup process?

does this happen only for virtual machines and not physical servers?

[RiaanBadenhorst]

Windows File Systems will create a snapshot but its not a separate job, you'll see it in the detailed status of the backup job. bpfis is the process that is responsible for all the snapshot "creation" (requests) in Netbackup.

Exchange will create a separate snapshot job like the virtuals do (its hanled by VSS too,unless you've enabled some 3rd party snapshot provider)

[sdo · Accepted solution]

Q1) so if I understand correctly, snapshots aren't acctually a netbackup process but rather a host (vmware, hyper-v) feature netbackup triggers as part of it's backup process?

A1) yes

Q2) does this happen only for virtual machines and not physical servers?

A2) I would expect it to happen for any "VM style" backup (in ESX (via VADP) and Hyper-V (via VSS)).  For physical, typically yes (via VSS) for "file system style" backups of Windows, typically no for "file system style" Unix/Linux, and typically yes for "database agent" style backups via internal database mechanisms within the database "application software" (i.e. MS SQL, Oracle, MS Exchange, etc...).