I am paranoid about backups, probably because of my 18 years' experience working with mainframes and minicomputers. Things go wrong. Systems fail. Machines stop. And then the call goes out, "Where's the backup tape?" All too often, the response is a blank look, followed swiftly by a realization that the tape device on the Windows NT server is for something more than copying files from laptops. Sadly, often the first time users consider backups is after their hard disk crashes.
I've heard of several recent catastrophic hardware failures involving Exchange servers. An editorial in PC Week ("24-by-7 minus 10 days of data = ouch!" by Brett Arquette, http://www.zdnet.com/pcweek/opinion/0504/04corner.html) describes one very public failure. The article tells how data loss occurred because the company couldn't restore the Information Store from backup tapes. Hardware failures are always a pain, but they become even more destructive if the backup tapes don't work and you can't bring databases back on line quickly after you replace the failed components. In the situation the article describes, the company didn't lose data because its IS department didn't perform backups. Instead, a disk drive had been having problems for 10 days (without error messages) before it eventually collapsed, and all the backups made during the 10-day period were invalid because the data was corrupt.
A set of databases forms the core of Exchange Server, and you must protect the databases at all costs. Good disk I/O subsystems equipped with features such as RAID arrays and a UPS provide a degree of resilience, but a good backup brings a feeling of security as no other measure can. Mark Ott's article "Backing Up an Exchange Server" describes some of the essential information you need to know about Exchange backup processes. But what about the hardware?
The company in the PC Week article seemed to do everything right; staff faithfully made backups every day. However, the article didn't mention the need to verify the backup tapes, a practice that could have caught the problem before it went so far. Verifying means that you take the time to test that the backup tapes can restore the Exchange Information Store and the Directory on to another server. More important, you need to test that Exchange can start up and use the restored databases. You can't test restores unless you have a server available for this purpose, and this requirement probably deters systems administrators from carrying out this procedure.
The question everyone must answer is, can you cope with a situation like the company in the PC Week article faced? Will your operational procedures pass the test in case of a hardware failure? No software can guard against hardware failure, and Exchange is no different in this respect. But with good planning, a disciplined approach to backups, and the good sense not to depend on everything going right all the time, you'll have a better chance of avoiding problems when you have to restore your databases.