What are the downsides of PARTIAL backups?

Question

I have recently found online this amazing article about partial backups by Alexandr Omelchenko

-- this is how he would do a backup:
BACKUP DATABASE your_database TO DISK = 'full.bak'
BACKUP DATABASE your_database READ_WRITE_FILEGROUPS TO DISK = 'partial_backup_full.bak'
BACKUP DATABASE your_database READ_WRITE_FILEGROUPS TO DISK = 'partial_backup_diff.bak' WITH DIFFERENTIAL
-- this is how he would do a restore:
RESTORE DATABASE your_database FROM DISK = 'full.bak' WITH NORECOVERY 
GO 
RESTORE DATABASE your_dataabse FROM DISK = 'partial_backup_full.bak' WITH NORECOVERY

GO 
RESTORE DATABASE your_database FROM DISK = 'partial_backup_diff.bak' WITH RECOVERY

GO

It happens that I have a few databases that have 4 filegroups. One of these file groups is loaded with documents that are saved inside a table in that filegroup.

I think I could benefit from backing up the filegroups separately.

What would be the possible drawbacks in doing that? Other than the overhead of altering the restore verifyonly, including dbaTools, to accommodate this change and making sure taking backups with checksums?

Something I would need to check is compression. Another thing is checksum option on for the backups - as suggested by the man himself - Ola - here and Paul Randal here.

We definitely need to check if our backups are reliable.

Answer 1

What are the downsides of PARTIAL BACKUPS?

Generically:

• More complicated backup procedures that others may not understand
• More complicated restore procedures as now you have fulls, FGs, FG Differentials, Log (and potentially regular diffs, etc.) that others may not understand (does anyone use Hekaton?)
• Potentially longer restores due to backup procedure choices
• May require retention changes (for backup files) or have retention based on multiple factors
• More complicated automation routines for restoring
• Requires Enterprise Edition to really be useful (online restores)

It's all a toss between flexibility, size, and administration. If a database is a great candidate for this, the items need to be weighed to see if it truly makes sense.