Ask HN: Serious Disk Backup, How to?

What are the better options for doing disk backup of a large PC, say 3 TB of data, up to a small Web server farm, say, 20 TB of data? For sizes in the Exabyte range we will leave to a second Ask HN! :-)!!

Some of the issues: Removable media, backup software features, various relevant techniques (synchronization, incremental, full), restoring, on-line, relational database, bootable drive partitions, disk to disk (all internal, on same computer or a different one), disk to removable media, elapsed time, bookkeeping, automation, multi-volume files, multi-file volumes, RAID (and RAID rebuilding times), role of drives connected via UBS, off-site and/or remote backup, SQL Server log shipping (backup and recovery), and security (in a server farm).

I'm developing only for Windows and not at all for Unix (with the resulting pros and cons).

Currently for my development computer, I am well backed up, but my tools and techniques are simple and already are failing to 'scale' well and will not be useful as my needs grow.

For bootable Windows XP disk partitions: I use NTBACKUP keeping 'system state' and with 'volume shadow copy' (VSS) that permits backing up a bootable partition as it is running. Apparently the backup can be restored to the same drive letter on the same computer but is not really a 'disk image' and the new drive letter need not be the same size as the old one. Even as I grow, for bootable partitions, NTBACKUP (or the newer replacements from Microsoft) may be what I stay with.

Related will be a 'deployment' issue: Given 20 new servers and a Windows 'site license', how to 'deploy' Windows Server 2008 and SQL Server 2008 to all 20, quickly? That is, install on one server, take a backup, and the 'restore' it to the other 19? Microsoft has some tools for large scale deployment, but so far I'm not familiar with them.

E.g., I plan to have quickly over 10 TB of data in SQL Server but so far have no good experience doing backup and restore for SQL Server. Of course a standard feature of a RDBS is ability to back up while the database continues to do transactions. But in case of a disk failure, RAID rebuilding elapsed times and/or SQL Server restore elapsed times could mean my Web site was too slow or down for too long. So, I need some good approaches for the SQL Server part of the whole. Of course additional, special details are handling the SQL Server system databases and the log files.

Anyone who has been there, done that, and has the T-shirt knows much more than I do. Fire at will: Many others will also learn.