Setting up a new backup scheme

The full is "leading". You're think of the week starting on a Monday. In the backup's "mind", the week starts on Friday or Saturday. Your weekly sets would be all the tapes from the full through the last incremental or differential before the next full. Don't think about "calendar weeks", think about "backup weeks".

It doesn't matter which day of the week you run the full on-- you'll still start a restore by bringing back the last full and then the last differential (or all incrementals) taken after that full.

Typically, Friday or Saturday is chosen to give the longest window of non-business hours time to take the full backup. If you have a better day of the week for the full, use it.

BTW: If you can get away with differential backups use them instead of incrementals. You have much less risk exposure.

Edit: An incremental backup contains only the changes since the last full OR incremental. A differential backup contains all the changes since the last full. Assuming you're piling a week's worth of incrementals or differentials on the same tape cartridge, they're functionally very close to equivalent.

If, however, you're using a different tape cartridge each day for the incremental, you have a dependence on each of those cartridges holding the incremental backups to work for a successful restore. With a differential backup scenario using a different cartridge each day, only the last full and differential cartridges are needed for a restore.

There are some good discussions on backup already on Server Fault. I'd have a look at them:

• my local backup request

• Is our planned backup strategy adequate for my new server insfrastructure?

• Recommended Backup Media for Circa 2009?

I notice that you're already ruling out tape. I'd do a cost-benefit analysis of the alternatives pitted against tape before dismissing tape out of hand. It sounds to me like you've either got a bum SCSI controller, cable, or bum tape drive and it's unfortunate that it has soured you on tape because tape can be very reliable and robust.

I did some calculations a few months ago (posted on Server Fault in one of the questions above, but not updated) and found that tape wasn't necessarily the most expensive route. Call me a traditionalist, but tape has worked very well for me and my Customers, and has proven to be reliable and suitable for disaster recovery. I tend to think that a lot of the "horror stories" associated with tape being unreliable often stem from backup strategies that don't involve testing of the backups after-the-fact. As is often said, backup isn't about backup-- it's about being able to restore.

Having a maintenance agreement for your hardware is, obviously, a component of any robust solution as well. You state that you need a backup you can restore to other hardware, but I'd caution you that your SBS server's "System State" backups aren't restorable by any Microsoft-supported means to alternative hardware. Hopefully your other server is a domain controller, since you'll be losing Active Directory (and, thus, making your Exchange data fairly difficult to get at) if you lose that SBS Server computer.

Having a maintenance agreement on the tape drive and / or server would get you out of the mess that you're in, in this case. It's particularly critical since you've currently got a backup, assuming you don't have another domain controller, for which you really need the identical server hardware available for restore.

You shouldn't center your whole backup regime around writing backups to an on-site storage device. Backup is off-site and offline. Anything less isn't backup.

I'd question if you can restore fast enough to get back up and running in a timely fashion if your main backup is an off-site device and accessible only through your Internet connection. If your renmote backup provider has a provision to ship you a physical storage device I could see that working. If you have to download your entire backup corpus over a consumer Internet connection, though, I'd think that you'd be talking about at least a couple of days to get everything back up and running. Ouch!