There's nothing like failover similar to what you have in the PIX, but the 30x0 has two features that work similarly.
Load Balancing:
Set up a group of 2 or more concentrators in a load-balancing format. This group shares a logical IP address that the user actually connects to, abd between them the concentrators will actually connect the user to the least-loaded concentrator. In the event of a failure of one concentrator, all those users connected to just that one are disconnected, but they can reconnect straight away without making any client changes and they'll get re-connected to one of the other concentrators.
Redundancy.
Similar to load balancing where the group of concentrators shares an IP address that the user connects to, but in this case the concentrators decide on a primary and a backup. all users connect to the primary, if that fails they'll get disconnected but again, they can re-connect back in without making any client changes, and they'll get connected to the backup concentrator.
Load balancing is better than redundancy (in my opinion) cause if you have a failure, at least you only lose some of your users, not all of them. With L2L tunnels the switchover in both scenarios is seamless and requires no user-interaction.
Check the Config Guide for details (http://www.cisco.com/univercd/cc/td/doc/product/vpn/vpn3000/3_6/config/index.htm).
