[arin-ppml] A challenge to the assumption that a big DFZ is a problem

William Herrin bill at herrin.us
Tue Dec 15 01:08:20 EST 2009

On Mon, Dec 14, 2009 at 2:07 PM, Ted Mittelstaedt <tedm at ipinc.net> wrote:
> One of the fundamental assumptions that we all seem to accept many
> times is the "Ballooning BGP table/DFZ"  Much policy discussion
> seems to be centered around the idea that if the DFZ gets bigger
> it's going to cost bazillions of dollars for every ISP to upgrade
> equipment, yadda yadda yadda.
> I have to ask, however, is this assumption really technically
> accurate?


As the DFZ gets bigger it will cost every ISP bazillions of dollars to
upgrade equipment. It isn't a question of "if." Exactly how many
bazillions depends on how fast the DFZ grows. At current growth rates,
few of the 200,000 or so deployed DFZ routers will be usable in any
DFZ application in 4 years time. Not even the ones that cost half a
million or more.

There's also an upper limit somewhere. We nearly hit it in the late
'90s but the routing table we're capable of building routers to handle
has been growing at about 40% per year since then while the actual
routing table has only been growing around 25% per year. This has
dramatically reduced the cost of low-end BGP routers (like the Cisco
2800 series) while allowing the high end router costs to stay
relatively steady despite the rapidly increasing packets per second

The technical short short version is:

Routing protocols and the code which processes them are
single-threaded while nearly all recent advances in processing speed
are in parallel processing, so growth beyond a certain point will
require radical software changes.

The packet forwarding process (FIB) in high-end routers is governed by
custom hardware (such as TCAMs) which are expensive to build and whose
cost varies linearly with the size of the routing table. Opex too --
TCAMs are very power hungry.

Bill Herrin

William D. Herrin ................ herrin at dirtside.com  bill at herrin.us
3005 Crane Dr. ...................... Web: <http://bill.herrin.us/>
Falls Church, VA 22042-3004

More information about the ARIN-PPML mailing list