Dilemma: If I provide a L2-Extended DCI solution within a traditional network architecture, how can I limit latency and ensure users at remote sites access compute resources at the correct data center?
For instance, if I have a single subnet stretched between DCs, the prefix would be advertised out both. If I have a server at the 2nd data center but the traffic comes ingress through the 1st, it would then have to route via the DCI and that produces a lot of latency, especially if the data centers are geographically disparate.
LISP doesn't appeal to me. What are some other solutions?
