Forgetting the Lessons of the Net: Routing Scaleability

An excellent article on ICANN, DNS, and Internet governance by Andy Oram[4] was forwarded to Farber's IP list recently. The bulk of the article was excellent, and we're going to see a lot more discussion of this type of topic in the near future-- even the US Congress is getting involved. However, I was slightly disturbed to see a big chunk of net.history overlooked in the debate, or worse, a very real problem being taken as crying 'wolf'. Here's the note I posted in response to Andy's article.

I agree with many of Andy's points, but I'm surprised at his description of the 'shortage' of IP addresses. The issue was not running out of numbers, the issue was 'how many independent routes can current routing tables service'. Back when folks were scurrying about consolidating IP networks into CIDR blocks, it was because of limitations on the amount of memory that then-current routers could usefully address (or providers could afford, or both) to hold the route tables.

Take a look at Geoff Huston's excellent article about historical BGP table scaling[0], take a trip to the distant past of 1996 to RFC 2008 [1] or earlier to the 'growth plans' section of RFC 1519 [2]. That last document states:

"As of Jan '92, a default-free routing table (for example, the routing tables maintained by the routers in the NSFNET backbone) contained approximately 4700 entries. This number reflects the current size of the NSFNET routing database. Historic data shows that this number, on average, has doubled every 10 months between 1988 and 1991. ...

It should be stressed that these projections do not consider that the current shortage of class B network numbers may increase the number of instances where many class C's are used rather than a class B. Using an assumption that new organizations which formerly obtained class B's will now obtain somewhere between 4 and 16 class C's, the rate of routing table growth can conservatively be expected to at least double and probably quadruple. This means the number of entries in a default-free routing table may well exceed 10,000 entries within six months and 20,000 entries in less than a year."

There's an excellent set of descriptions of the Routeviews project, and some shortcomings of BGP (which itself is still more scaleable than OSPF on today's network) at the APNIC meeting transcript of February 2005 [3].

cheers, Strata

Disclaimer: I are not a network eNgineer, I'm a systems person, but I've been around the block long enough to know a bit of history, and less than 5 minutes of Googling lets me share it with you folks in better detail than I'd be able to write up personally. I was also looking for specific NANOG traffic from 'the day the net broke', eg when they separated the NAP/MAE traffic and handed off to ARIN(? was it ARIN?) and everyone with a backbone router found that suddenly they needed *double the memory* in their routers. Felt very sorry for my net-eng buddies that day! Anyone got cites for that lying around at hand?

[0] http://www.cisco.com/en/US/about/ac123/ac147/ac174/ac176/about_cisco_ipj_archive_article09186a00800c83cc.html [1] http://www.faqs.org/rfcs/rfc2008.html [2] http://www.freesoft.org/CIE/RFC/1519/10.htm [3] http://www.apnic.net/meetings/19/docs/transcripts/routing-sig.txt [4] http://www.oreillynet.com/pub/wlg/8147