Before I get into a lot of edits and thoughts -- In section 2: == In the above scenario, the term "route aggregation" refers to the case where a node situated at the WAN edge of the data center network behaves as a default gateway for all the destinations that are outside the data center. The absence of route aggregation refers to the scenario where NVEs within a data center maintain individual (host) routes that are outside of the data center. ==
Reading through the rest of the draft, there seems to be a lot of interchange between "WAN gateway," "route aggregation," and "default gateway." Let me try describing two switching paths, and then try to sort if they actually cover all the different scenarios given in the draft. Assume you have: TS1--NVE1--R1--R2--NVE2--TS2 TS1 wants to send a packet to TS2; however, TS2 is on a different subnet. TS1 examines its local tables, and determines that TS2 is on a different subnet, so it transmits the packet to the default gateway (it actually doesn't matter if TS1 itself sends this packet, or if TS1 is a switch, and some device behind TS1 sends this packet -- the result is the same). In this case, let's assume NVE1 is the default gateway, so it receives and must process the packet. There are two options at this point. NVE2 could have a routing entry directly to TS2 which includes layer 2 information through an EVPN advertisement in BGP (or other means outside the scope of the document). In this case, NVE2 uses the layer 2 information contained in the BGP advertisement to build a MAC header rewrite string, stuffs the packet into the correct outer header (MPLS or VXLAN or whatever else), and ships it towards NVE2. NVE2 pulls the tunnel headers and fowards based on the layer 2 information. NVE2 could have a routing entry directly to TS2 which does not include layer 2 information. In this case, NVE2 simply forwards the packet based on any local layer 3 information it might have. Let's say this layer 3 information is either a default or more specific route advertised by R2. In either case, NVE1 wraps the packet in the correct tunnel information to reach R2 and ships it. R2 removes the tunnel information and forwards it to NVE2 based on local forwarding information (which could be a layer 2 advertisement or a layer 3 advertisement -- from NVE1's perspective this isn't important). These two seem to cover all five of the possibilities covered in section 2. Where TS2 is located -- within or outside the local data center -- doesn't really make any difference (unless you assume there will never be a default route advertised between IP subnets within a data center, or there will never be IP only connectivity between two different subnets within the same data center, both of which seem like a bad assumptions to me). Is there a specific reason why the location of the "wan default gateway" makes any real difference in the different scenarios? Or why the default route is a special case (other than requiring the device advertising the default route to be able to interconnect an EVI and an IP-VPN)? Russ _______________________________________________ BESS mailing list [email protected] https://www.ietf.org/mailman/listinfo/bess
