> I have a question about OSPF cost. According cisco, to calculate cost in > OSPF, we use the formula: 10^8/BW. What's OSPF cost in Gigabit/10Gigabit

AFAIK, cost for interfaces with bandwidth equal or larger than 10^8 bps is normalized to 1. It is not a good practice to left it that way because it will not "see" the difference between (high bandwidth) circuits. The main reason for all this is that people which created OSPF spec in the first place, didn't expect to have such a great bandwidth that is available today. Seems it was a common mistake in IT in the past (just remember DOS and 640K limit). To fix problems you could use manual 'ip ospf cost' interface command which will set appropriate cost to specified inteface. This might look as a solution to a problem, but really it is not because you need to lower bandwidth on Gig interfaces and you can't go into negative cost. Probably the best way would be to use 'auto-cost reference-bandwidth' command under your routing process command mode. It will put specified bandwidth as a reference. The important thing is that you have to put the same reference bandwidth to all OSPF routers in your OSPF domain to get desired results. You should be warned that while doing this, you might loose connectivity on your routers for a while because routers have to run SPF algoritm again. It could also create additional connectivity problems because of possible asynchronous routing while not all the 'reference-bandwidth' commands are in their place.