[arin-ppml] Does Moore's law help with routing table growth? Get thee to the RRG!
rw at firstpr.com.au
Tue Dec 22 19:39:19 EST 2009
Short version: The tension between DFZ scaling vs. end-user network
portability/multihoming/TE need not be a zero-sum
The IRTF Routing Research Group has multiple
proposals to enhance the Internet's architecture to
enable much wider use of portability, multihoming
etc. without seriously burdening the DFZ.
Discussions continue to mid-February and the
final recommendation to the IETF needs to be
finished by the end of February.
I hope anyone with an interest in this field will
There are various administrative and economic ways of limiting the
growth in the DFZ routing table, and so reducing DFZ router costs and
limiting the increase in DFZ convergence time. This would directly
protect two sets of interests:
1 - Costs of ISPs.
2 - Costs of end-user networks with ASNs which currently advertise
their space in the DFZ AND run their own DFZ routers.
It would also tend to improve everyone's Internet use if the DFZ
stability is improved, or at least not worsened as much as it would
be with many more DFZ routes. Also, reducing ISP costs is generally
a benefit to all Internet users, since we all depend on ISPs in one
way or another and ultimately pay all their costs. (Perhaps end-user
networks with their own DFZ routers are not so dependent on ISPs.)
Its not just a question of advertising a prefix - it is how often
that advertisement is changed. Rapid TE, such as to accommodate
traffic patterns which change on a 24 hour cycle, or faster, will
change the advertisements frequently and so burden the DFZ's BGP
control plane - leading to potentially greater instability.
We accept that ISPs need to advertise their prefixes in the DFZ and
we accept that this alone is scalable - but this depends on some
limit to the number of ISPs, which so far has not been a problem.
Any ISPs which chop and change their advertisements would arguably be
as much of a burden on the DFZ as a bunch of extra end-user network
Your solution - administrative and economic push-back - involves
restricting end-user networks from getting something they can get
now, so it is harder or impossible to get.
The RRG is trying to find a way to alter the architecture of the Net
so the end-user networks, potentially right down to SOHO (and mobile
devices, though mobility is outside the RRG's goals) can get what
they want, inexpensively, without causing a burden on the DFZ.
There are multiple proposals - 22 December is the final day for
accepting them and I guess they will soon be listed in the wiki.
I think you and any other people who have thought about this DFZ vs.
multihoming/portability/TE problem would be able to make a
contribution to the RRG's decisions on what to recommend to the IETF.
(I am just an RRG member.)
More information about the ARIN-PPML