Sql-server – Backing up the database without compromising performance

I don't think this is a very good idea because replication is not just as bullet proof as a sound backup strategy. A blocking situation could stop replication jobs and you could easily be out of sync for some time. Or maybe a network card will fail and replication will be dead for a period. This means you won't be up to date for some time, and also, you will be without a good backup of the main, updated site.

Similar technologies, that will deliver updates to a copy of the main db, are mirroring and log shipping, but they both rely on full recovery model and log backups.

I'd suggest that you'd better study compression for your backups. Starting from SQL 2008 (Enterprise) and SQL 2008 R2 (from Standard) you have the option of compressing the backups using native SQL compression. In addition you should investigate data compression (in place compression for tables and indexes data). These will give you some help by making the backups smaller.

There are also 3rd party backup tools that will be able to help with compression if you need more complex solutions. See this site for question about how to make SQL Server backups as small as possible.

Also, you might think about moving old data from big tables to a history database, and play with connection strings in your app.

You could also try to make the backups locally, in a folder on the main server, and then just copy it to the tape (maybe backup to disk + direct copy will take less than backup to tape).

Sorry for the emphasis: YOU JUST WON'T BE SAFE WITHOUT TESTED PRODUCTION BACKUPS!

As long as the indexes that need to be rebuilt get rebuilt it should be fine to stagger them.

This will of course vary based on your usage - if you have a ton of updates/inserts/deletes run on Fridays then there may be a good reason that you currently reindex on Saturdays.

Something else you can do, instead of blindly rebuilding every index every week, is using some queries to check fragmentation levels and only reindexing when a specific index is over a certain threshold of fragmentation. The usual threshold (which was chosen somewhat randomly by Paul Randal many years ago) is 30%. There are dozens if not hundreds of sample scripts available for this.