But why would you do such a thing?? You'd have to be really frickin' stupid to burn dollar bills to light a fire!!
Yeah, I knew you'd agree with that. Fortunately, NetBackup customers are smarter than the average bear (or, let's say, the average backup administrator). NetBackup customers know that backing up ESX servers twice (once for the VM image, once for all the individual files) is like burning dollar bills, and additionally, backing up without using incrementals is like throwing a few twenties on top.
Let's explore the incrementals angle a bit. NetBackup 6.5.2 is the only backup product available that supports incremental backups with VCB. How does this work? Using VCB at the VMDK level, a full image is backed up initially, and then catalogued on the proxy server. Subsequently, incremental backups are performed by comparing the contents of a VM against the stored catalog of the initial full image. If changes are detected, then VCB is used to back up only the changed files, at the file level (not the block level).
Technically, how does this work? With NBU 6.5.2 we combine two types of VCB backup. The first method involves copying over the complete vmdk files to the Backup Proxy. The second method simply creates and mounts a "shortcut" of the virtual machine (no data is copied to the Backup Proxy during this mount operation). This shortcut represents the individual files that exist inside that virtual machine. Single file backups are possible using this second method but vmdk file backups are not possible.
The FULL backup uses the first, vmdk level backup. The INCREMENTAL backup uses the second method. We can do this because we have already mapped and know exactly which individual files (e.g. Word docs) exist inside the vmdk files which we designate as the FULL backup. For the INCREMENTAL backup, we mount the shortcut of the VM, search for files that have changed since the FULL (vmdk) backup, and then backup *only* the files that have changed. The advantage with this technology is that for incremental backups, we copy to the Backup Proxy only the files that have changed since the vmdk FULL. This significantly reduces backup time, I/O load on the VMware Datastore and amount of data written to tape or disk, while retaining the ability to restore individual files regardless of which backup method was used. This technology does not rely on any deduplication technology but can be used with any deduplication product.
So, why is this a good thing? Well, running incremental backups instead of fulls saves a lot of money in storage space. It also dramatically shrinks the backup window and the performance impact of the backup itself.
Why would you light your fire with dollar bills? Give them to your friendly Symantec sales rep or (cough) Product Manager instead :)
