Hi Sam, At 16:28 28/07/00 +0200, Sam Critchley wrote:
These are interesting data - looking at the information on the web, it's remarkable also to see the change in the ratio of originating ASes to transit ASes in each of the regions. Note that over the past year, in the ARIN area the ratio of originating to transit has gone from about 3.25:1 (not an exact figure, perhaps someone could come up with the precise calculations) to around 9:1, whereas in the APNIC region this is approximately 4.75:1 -> 6:1, and in the RIPE region has moved from around 4.75:1 to 5:1.
Thanks for this - I hadn't even noticed these changes but then haven't studied the data in depth for a while now. I had noticed what appeared to be an acceleration in the number of ASes being distributed by the registries, and I think these numbers confirm it.
Whilst these are not precisely-calculated ratios at all, and there may be many reasons for this change, I would be interested in hearing what is behind this shift. Anecdotally, I think we are beginning to see a large increase in the number of enterprises (rather than ISPs) who are running BGP.
I think you are right. More and more non-SPs are multihoming now. It is certainly educational to look through the recent assignments made by ARIN (given that is the region which is "burning" ASes the fastest) - very few SPs in there (going by the names anyhow).
1. Increase in the number of prefixes in the global BGP-table - a move from linear increase to exponential? What knock-on effects would this have on the way routing functions? I'll be interested to see what happens when the global table reaches 100,000...
Should it be a problem? I'm told that several providers are already running well over 100k prefixes in their iBGP. The knock-on effects I'm expecting are more to do with stability of the routing table. So far, there is no evidence to suggest there is any instability, even with this growth.
2. Increase in the number of ASes in the Internet Routing Table. What are the statistics looking like here? Do we have an acceleration in growth here as well? If so, is there a risk of running out of ASNs in the BGP IPv4 routing-table?
It is accelerating, but note that 8000 ASes are present in the routing table out of approximately 16000 which have been assigned... I wonder where the others are, apart from in back pockets? ;-) Looking at the previous 12 months, we're seeing about 680 new ASes per quarter. It is accelerating slowly, and there are over 47000 left, so would last around 12 years at a sustained assignment rate of 1000 per quarter. Should we worry - well, probably not with the current usage pattern, but if multihoming is going to move from the large/medium enterprise to the SOHO, or even residential, a unique AS per site may become a non-scalable solution.
What do others see as possible explanations for the phenomenon?
The other thing to bear in mind is that the "developing regions" still have to catch up with North America and in some case, Europe, when it comes to resource consumption. I'm thinking of China, India, other parts of Asia, Latin America and Africa - big growth potential here in the near future. So far I've counted a total of 94 ASes assigned to African Internet sites, and 488 ASes assigned to Latin American sites. This compares with 10000 for the US and Canada alone! cheers, philip -- -------------------------------------------------------- Philip Smith ph: +61 7 3238 8200 Consulting Engineering, Office of the CTO, Cisco Systems --------------------------------------------------------