Interesting article on traffic management between multiple data centers in case of failures