Hi all, With distributed mapping database, such as
CRIO, the mapping tables are smaller but paths are longer. With anycast
mechanism, such as default-mapper in eFIT-APT, the paths are shorter but the
mapping tables are larger. How about integrating both of them into one
solution? For example, set up multiple PoPs, within the transit network, which
advertise the same super-prefix, and ETRs register EID-RLOC mapping info to the
nearest PoP. Those PoPs advertising the same super-prefix should synchronize
the mapping database. In this way, the traffic with destination to the same
edge network can be shared among those PoPs and the average path is shorter
compared with CRIO. Best regards, Steven XU |