yarym said:
So that means if server crashes, i would not be able to restore the exchange back to its original settings? What should I start backing up?
You need to consider a) how to restore the environment, and b) how to restore the latest email data.
Lets take them in reverse order. To be able to restore your latest email data, you need a recent copy of the databases. An online backup of these is by far the best, it checks the database for corruption every time it runs, and purges away the transaction logs that have built up since the last full online backup (you have turned off circular logging, right?). We do one every weekday night (our servers don't get much use over a weekend). With a nightly backup, you're safe against a corrupt database file - if a backup fails one night because the priv.edb file is corrupt, it will also leave all the transaction logs on disk. You simply restore your last good online backup, and the remaining logs will replay and bring your restored db back up to date. This is why it's a real good idea to check your backups worked every single day, and also a good idea to keep your transaction logs on a totally physically different drive(s) from your databases (use a mirrored pair of disks for your transaction log volume for extra security).
Now, back to part a), restoring the environment. To do this, you could use an NT/file-level backup of every file on all disks (inc your registry). Because you're taking seperate backups of the Exchange databases, you can exclude the *.edb and transaction logs from this backup if you wish. It works slightly better if you stop your Exchange services before you do this kind of offline backup - so out of hours, unless you want lots of unhappy users. You can either do these backups on an ad hoc basis (whenever you make changes to a server, like patches or SPs etc), or do it semi-regularly - we do one at 4am local time every Saturday morning on each server, but this is probably overkill, once monthly would probably be better.
Or, and here you might want to experiment in a test lab, you can avoid having to restore a backed-up environment. You'll need a good documented copy of your server setup for this to work. When disaster strikes, and your server dies, simply turn it off, and find another. Build a replacement server from scratch, with the same name, same settings, same organization name, same site name, same SP level, everything the same as the sick server. Make it a new server in a new site, but you must have the directory name for the Exchange organization and site names EXACTLY the same as before (even the case is important), then when the server is up and running and you've tweaked Exchange to exactly the same as your documented settings (circular logging, etc etc), do a full online restore of the last know good databases - including the directory. This will bring the entire identity and mail environment of the sick server back to the replacement server at the point of the backup. So you'll have some lost mail (depending on when your sick server went down), but I imagine most of your users will be releaved that their mailboxes have come back almost totally intact.
This approach is more difficult if this server also hosts your IMS, but this is all the more reason to break this out to a seperate IMS (non-mailbox) server.