[ppml] Legacy /24s

William Herrin arin-contact at dirtside.com
Sun Sep 2 00:29:31 EDT 2007

On 9/1/07, Leo Bicknell <bicknell at ufp.org> wrote:
> In a message written on Sat, Sep 01, 2007 at 02:40:25PM -0400, William Herrin wrote:
> > 10%-30% of the routers in the DFZ today have a hard limit between
> > 244,000 and 260,000 IPv4 routes or half that number of IPv6 routes.
> You've nicely bounded the problem of upgrading Sup2+MSFC boxes to
> Sup720 boxes.  (I'm guessing, because I've seen you talking about
> the problem on the nanog mailing list and all the numbers fit.)

That includes most of the Sup720 boxes too; they're bounded at 260k.
In the 6500/7600 product line, only the Sup720-3BXL and RSP720-3CXL
accommodate more than 260k routes. And as you point out there are no
in-place upgrades for the remaining 7200 and 7500 routers.

> upgrades (some of which you want, some of which you don't), and can
> easily be in the range of $500,000-$2,000,000 per box.

Relatively speaking, there are few routers in the DFZ whose street
price exceeds $150k. More, once you're past $150k the major cost
driver is traffic capacity rather than route capacity.

> Your numbers also only account for the hardware cost. [...]
> In short, I'm afraid your estimate is WAY low on cost per device.

They don't account for the electricity cost for running the
power-hungry TCAMs and the required air conditioning capacity either.
Nor do they exclude the cost attribution associated with the routers'
traffic capacity and age-related upgrades rather than route capacity.

Like I said, its not a thorough analysis. I selected only the most
significant factors in order to produce a ballpark figure.

> > The number of routes and ASes in the DFZ implies that there are
> > somewhere between 200,000 and 300,000 routers in the DFZ.
> I can't come up with anyway to estimate, but my gut tells me this is a
> little high.

Could be. I'm open to better grounded numbers than the ones I offered.

> Ok, so, we keep talking about a 50-100 year lifetime for IPv6.
> Using your numbers, assuming it's a one time cost, and there's no
> inflation means if ARIN makes an allocation it will cost $900,000
> to $5,000,000.

You really can't project a lifetime cost that way. For one thing, the
cost of one route in one router is falling faster than the number of
routers in the DFZ is growing. For another, there are a variety of
proposals on RRG which on a 5-year horizon would adjust routing
technology in a way that radically alters the cost structure
associated with PI space. The more PI space there is, the higher the
probability of one of these technologies being adopted.

You can only really project the annual cost for the next few years
with any degree of accuracy.

> IPv6 was supposed to be different.  The concept is rooted in really
> good principals.  Unfortunately the execution has been lacking.

I pulled out and leafed through my old "Internet Protocol Next
Generation" book from 1997. Do you realize just how few of the design
goals have panned out operationally? Its appalling.

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