Any recent predictions for routing table growth?

Mailman List Mirror (Read Only)
1

Hi,

I am looking for some recent estimates of future IPv4 & IPv6 routing table
growth, the most recent reliable estimate I can find was done by Vince
Fuller in his presentation in March 2007, is there any newer or alternative
figures out?

Thanks

Bradley

2

Geoff Huston's has http://www.potaroo.net/tools/ipv4/ which goes up to
the present.

3

Thanks for that link Bradley (& Joe who replied off list), but IPv4 address
depletion has been discussed to exhaustion and I was looking more for the
speculative sizes of the routing table in 5 to 10+ years time such as on
page 19 of this presentation www.vaf.net/prezos/*r*rg-prague.pdf is there
anything similar available?

Thanks

4

It doesn't provide you with the breakdown on page 19, but here is the
predictions he provides.

    http://www.potaroo.net/tools/ipv4/fig20.png

You can ask him for a breakdown if you like.

5

In order to double on schedule from the point where it hit 250k routes
the rate of prefix growth needs to be on the order of 2k prefixes a week...

I'm operating under the assumption that I'm going to need 500k dfz fib
entries around mid 2010 which oddly is about inline with where we
thought we'd be when we did the fib bof at nanog 39.

Brad Freeman wrote:

6

Hey, Brad - the latest I know of are ours, but I'm possibly out of date:

http://www.cs.cmu.edu/~dga/papers/aip-sigcomm2008-abstract.html

Look in section 4.1. The #s were from routeviews, June 30, 2008. The gist:

June 2008: 247K entries
Growth rate: 17% per year

So - June 2009: 288k

There's an embarrassing typo in the formula in the paper - it says "2.07 * 10^4" as the base, when it's obvious that it means 2.47 * 10^5. Sigh. I'll get that corrected. :slight_smile:

Also note that our #s differ a bit from, say, CIDR report since we used routeviews as our baseline. If you use the june 6, 2008 CIDR report as your starting point, which starts at 267k, the 17% exponential growth would predict that the October 31, 2008 CIDR report would report 284k prefixes; in reality, it reported 286. So, reasonably close. But you want to start with the # of prefixes that YOU observe, since that's going to be a little different depending on your vantage point.

Plug in:

STARTING_NUM_PREFIXES * e^(NUM_DAYS_ELAPSED * 0.0004253)

e.g., 267000 * e^(147 * 0.0004253)

and you'll have a pretty decent prediction unless things change course. :slight_smile:

7

Thank you very much David, the Routing Growth estimates is exactly the
research I was after.