DFZ, pTLAs and the 6bone Michel Py Iljitsch van Beijnum ipv6mh mailing list

Goals of this presentation: Semantics clarifications. Provoke thoughts about the future of the IPv6 Internet backbone. 2

Semantics Problem at hand: define “IPv6 DFZ”. What does it mean? Default-Free Zone But what is it exactly? 3

Semantics IPv4 DFZ: The subset of routers that do not have a default route and do not receive a full routing table from a single peer. In the current IPv4 tiered system, the DFZ is the inter-connected full mesh of tier-1 providers. Tier-1 ISPs do not pay anyone for transit, resulting in them being fully meshed. 4

Semantics There is a difference between the IPv4 DFZ and the IPv4 DFZ’s routing table. The IPv4 DFZ’s routing table extends beyond the DFZ’s boundaries (for example, to multihomed customers). One of today’s IPv4 issues is the size of the DFZ’s routing table. 5

Semantics Several things have contributed to the large size of the routing table. Due to the shortage of v4 addresses, allocation policies tend to allocate barely what is needed at a given time, resulting in relatively poor aggregation. Multihoming has little influence on the size of the v4 DFZ’s routing table, since only ~12k ASes are currently in use. 6

Semantics Then, what is the IPv6 DFZ? If the v4 definition is to be used for v6, there is no such thing as an IPv6 DFZ today. The definition of the IPv6 DFZ is based on our guess of the evolution of the IPv6 Internet. 7

Semantics We see three possible scenarios: 1. Centralized backbone, Arpanet / NSFnet style. Unlikely. 2. Competing but interconnected backbones. This is the current tier-1 system. Likely. 3. No major backbone(s) but large scale direct interconnection between smaller networks. Is this a realistic possibility? 8

Semantics Consequences on semantics: _IF_ scenario 2 is what happens: The term “TLA” and the derived “8k DFZ” are used as described in RFC2373. The authors understand that draft-ietf-ipngwg-addr-arch-v3- 07.txt obsoletes this model. As soon as RFC2373 is becomes obsolete or historic, we will need a definition for “what used to be called a TLA”. 9

Semantics Consequences on semantics: _IF_ scenario 2 is what happens: Most TLAs (there could be ~8,189 of them with 2001, 2002 and 3FFE taken, the so-called “8K DFZ” ) will be tier-2 ISPs. Note that here, “8K DFZ” does not make any sense because only tier-1 are part of the DFZ. It should be “8K DFZ’s routing table”. There is no direct relation between being a Top- Level Aggregator and being a tier-1 ISP. 10

Semantics _IF_ scenario 3 is what happens: Then what is the definition of “IPv6 DFZ” ? (because the v4 definition does not match anything). A possible definition is that the DFZ will be a backup only localized at one or more exchanges. 11

Relation to the 6bone The question is not how to transform the 6bone into the IPv6 backbone. The question is not whether or not tunnelling v6 in v4 is bad. We know it is. But guess what? The flexibility and price are right. The question is what 6bone participants think about the future IPv6 Internet. 12

Food for thought The 6bone and the production v6 Internet are connected. How many 6bone pTLAs are we planning on and what is the impact of the 6bone on the DFZ’s routing table? IPv4 style multihoming depends on not aggregating, so doing this in IPv6 will either lead to an exploding routing table or to multihoming being restricted to a happy few. 13

Food for thought Is scenario 2 the only one? If yes, does the 6bone / IETF have a role in recommendations, or do we just accept the fact the current v4 tier-1 will become v6 tier-1 as well? If not, please speak up about what you think the peering structure should be. 14

Thank you for your time and attention