Moving to IPV6

[Robert E. Seastrom]

Louis Mamakos <louie at transsys.com> writes:

> On Feb 7, 2011, at 8:15 AM, Robert E. Seastrom wrote:
>
>> 
>> Anyway, the *real* problem that IPv6 fails to fix is one that I'm
>> surprised that Mike didn't mention since he had an IETF draft or three
>> that would have fixed it.  That problem is one of routing table
>> growth.  In the middle of the Internet are large routers - not like
>> the one that you have on your cablemodem, but ranging in size from 2
>> rack units tall and weighing some 40 pounds to several adjacent 42u
>> racks and weighing a couple of tons.  They carry the routes for
>> Internet provider allocations.  This is called the default-free zone
>> or DFZ, because you don't carry a default route or gateway - you're
>> supposed to know how to get everywhere.  A full routing table in IPv4
>> land is 375k routes, approximately.  In IPv6 land it is 4500 routes.
>> Both are growing exponentially.  Processing power on the routers'
>> management systems to sort out the routing information is not.  You
>> can see where this is headed...  eventually...
>
> Processing power might be the easy part; it's the wicked fast,
> horrifically expensive memory that contains the computed forwarding
> tables that's also an issue.  You can dump route updates from your
> neighbors, and churn away doing computations in normal old DRAM,
> like in your computer.  But eventually the rubber must hit the
> road, and that's (depending on the router's architecture) on a
> line card that's got to do a route look-up for every packet that
> arrives in very high speed memory. 

Yep, TCAM and similar technologies are *expensive*.  Sucks to be us.
And by "us" I mean the people who are paying the bills, including
folks like my mom, etc.

> And you need to include enough memory bandwidth to be able to update
> the forwarding tables, too (or add twice as much and swap back and
> forth).

That's the real squeeze there.  Adding twice as much and swapping back
and forth is a nice stopgap, but an x2 lift when your requirement is
^2 is cold comfort.

> While the cost of all this exotic memory is one issue, it
> also tends to consume relatively more power and generates heat that
> must be removed, too.  Power and heat densities in the facilities
> that contain these sorts of network elements are a real, er, issue.

This part is less worrisome than the above, but still a big deal.  I'm
wondering how long till the liquid CO2-based cooling stuff takes off
in a big way.  Pipe it right into the router itself.  Return to the
days of the water-cooled mainframe.

-r