SQL Server Incremental Backups – Handling Race Conditions

editted based on feedback in comments

1. If you could stay readonly than a good alternative to replication could be log shipping. The vendor would create transaction log backups. And send those backup to you. You then apply those backups.You can't restore differential only, you would still need to reply a full first. Another benefit of (delayed) logshipping is that, since the source is at the vendor (where you have no control over what they do with it, you might want to build in some buffer (in time) that when they make a huge mistake, you simply pause the logshipping redo, so you still have a good copy)
2. For restoring filegroups/files you need transaction log restores as well. It can work but it highly depends on the growth rate vs the transaction log rate if I would prefer this above option 1, simply log shipping. Furhtermore, partitioning your database and table just for the sole purpose of solving your replica problem wouldn't be the route I would pick first but I have some questions for you later.
3. Replication sounds like a very good option if you think the restoring of transaction log backups becomes to much work.
4. What prohibits sql 2000 from doing Bulk Inserts? However, if you think logshipping is tedious, keeping track of updating every single object sounds like a potentially more troublesome strategy.
5. I have no experience with cdc. somebody else has to help out on that.

Couple of questions for you:

1. How big is the initial database size?
2. What will be the daily, monthly growth rate?
3. What will be the daily growth rate of the transaction log?
4. is it possible to have both locations in the same domain?
5. Is it possible to build a secure connection between the two locations?
6. What is the bandwidth between the to locations? what is the latency?

If we have some numbers, it's easier to provide you with a better advise.

First of all grab Ola Hallengren's scripts to do the backups (http://ola.hallengren.com) and drop the maintainance plans. Ola's scripts have much better retention settings and will allow you to set a mark in hours how long to store the backups and know how to take the backups of the secondaries.

And for your questions:

1. Set up a small script in SQL Server agent or Windows task scheduler that copies a backup file to a secondary folder where you keep it longer than the others or do a mirrored backup to a secondary folder every six months. A small powershell script or xcopy command that runs every six months and copies the last backup file will do just fine
2. Backup log DatabaseName to disk ='NUL' will make a log backup to the windows equivalent of /dev/null, no disk writes or space wasted. Nogood to restore from though.