Yes, VRF-lite SHOULD be used in an Enterprise environment to isolate the different security classes of devices.
In the past you would isolate different groups of users using Layer1, i.e. separate hubs either totally isolated or connected together by a router with ACLs. Since the PCs were only connected at shared 10 Mbit and the routers were such low performance and worms weren't really prevalent, this was not a big security issue at the time.
Then we migrated to VLANs, which essentially allowed Layer2 isolation within the same switch to provide the same functionality of separating different classes of users and to break up broadcast domains. Unfortunately, everyone connected the VLANs together at Layer3 with a router (or SVI) which essentially connected everything together again! And almost no one gets the ACLs right (if at all) to isolate the VLANs from each other. In fact, in most cases every VLAN can automatically reach every other VLAN from a Layer3 or IP perspective. This is a huge security problem.
Enter VRF-lite, essentially created by Cisco as their tag switching migrated to standards based MPLS and had a need to isolate Layer3 security domains from each other within the same switch (or router). Think of VLANs for routing tables. VRF stands for 'Virtual Route Forwarding', which basically means separate routing tables. Since VRF-lite is a per-switch feature (running locally to the switch) you will need to use other technologies to connect multiple VRF-lite switches together and keep the traffic isolated, see below.
What makes this so secure is that there is no command within the switch to connect different VRFs together within the same switch. You would need to connect a cable between two ports on the same switch configured in different VRFs to be able to communicate between them (recent IOS 12.2SR allows tunnels with different source VRFs but that is a corner case). The reason for this is simple, remember the basis for VRF (and VRF-lite) is for a service provider to isolate multiple customers from each other within the same switch. Just like an ATM, Frame-Relay, SONET, or Optical switch, the command line makes it very difficult (or impossible) to accidentally connect 2 different customers together.
Think about that. Even if someone was able to get ssh enable access to your switch (you aren't running telnet anymore, right?!), they CAN'T connect 2 VRFs together with any command.
And, yes, this is highly recommended by Cisco Engineers and is actually deployed far more than you think. I have VRF-lite running on at least 10 client's networks and those are LARGE networks. VRF-lite was integrated into the environment purely to solve a Layer3 security class isolation issue. I have used Layer3 dot1q trunks on c6500 switches and tunnels to keep isolated connectivity between VRFs between switches.
In Cisco speak, VRF-lite falls under the topic of 'Path Isolation' which is combined with other features that isolate traffic within the same network such as dot1q trunking, tunneling, VPN, policy-routing, and MPLS. Do a search on Cisco's web site for 'path isolation' and you will find a bunch of info.
See the following URLs for a good start:
http://www.cisco.com/en/US/netsol/ns658/networking_solutions_design_guidances_list.html
http://www.cisco.com/en/US/netsol/ns658/netbr0900aecd804a17db.html
http://www.cisco.com/en/US/netsol/ns658/networking_solutions_white_paper0900aecd804a17c9.shtml
As always, rate all posts appropriately, particularly those that provide value and don't be shy about following up with additional questions or comments.
Good luck!
