Randy, On Thu, 2011-06-30 at 08:45 +0900, Randy Bush wrote:
My understanding is that routing table growth is largely fueled by traffic engineering rather than multihoming.
got cites?
As you know, I'm not a routing guy, so I tried to be careful with my language here. I'm neither an operator nor a researcher, so if I sound ignorant it is because I am. Having said that... One simple way to look at it is to consider how much PI space there is, versus routing table size. Looking at the latest copy of the RIPE Database, we find just under 29k PI assignments. I am told that the RIPE PI policy is the most lenient in the world, but if we assume that similar policies are in place throughout the world that still leaves less than 100k routes for all PI holders combined, compared to the 350k routes total. This does not tell us anything about the *growth* exactly, but it serves as a worst-case analysis: in the current Internet, most routes come from non-PI allocations. Further, a quick google for "routing table growth deaggregation" turns up the following paper: http://people.ee.ethz.ch/~wmuehlba/jsac-10.pdf It is a couple years old, and I don't know if the authors are trustworthy, but it seems nicely done. ;) The authors were looking for trends in changes in traffic engineering practices, and summarize: To sum up, given the differences in the two plots of Figure 7, we conclude that traffic engineering is a significant driver for address space deaggregation. However, based on the evolution since 2001, we cannot identify any sudden shift towards a more aggressive and widespread use of prefix deaggregation in order to achieve traffic engineering. This suggests that it is not an increasing demand for traffic engineering that inflates routing tables: rather, inflation is a consequence of the topological growth of the Internet combined with the lack of support for traffic engineering in the current routing system. Nevertheless, if we look at Fig. 3., we see that the percentage of prefixes from deaggregated routes has gone from like 17% in 2001 to 31% in 2008 - almost double. So while the amount of routes due to traffic engineering is not due to a fundamental change in practice, it seems probable to me that it is nevertheless the source of a lot of routing table growth. Regarding the implications on policy... we seem to have a situation where the people for and against increased IPv6 PI seem to think that the other side must prove their position somehow. These ideas are all based on potential future impact, and historically predictions about the future have been very difficult. Since we are talking about theories about future impact, we're kind of like economists: the only real way to test a theory is to implement it and see what happens. I will note that *not* changing the policy is also implementing a policy, so no matter what we do, we're running an experiment based on half-baked theories of how the Internet works. More research will certainly inform the discussion and possibly lead to better policies, but policies will never be based on perfect understanding. -- Shane