> The problem, as I understand is not really the size
> of routing tables but the frequency with which they
> are updated. 

At this level the type of routing we're talking about is BGP (Border
Gateway Protocol). Whole netblocks (i.e. 195.112/16) are aggregated
together in single routing objects called autonomous systems
(AS). Typically a single ISP will have a single AS containing all their
networks. Sometimes an ISP will have more than one if seperate parts of
their network have incompatible routing policies.

Whenever there's a change in the routing information of a particular AS,
each AS it talks to has to recalculate its routing tables. Then it also
has to tell each of its other peers about any changes. So it propogates
out - a ripple that goes through every BGP router on the Internet. It
takes a lot of bandwidth.

The main factor driving the increase in BGP traffic is the number of
routers rather than the number of routes, but obviously that has an affect
too.

> I imagine that this may be partly to
> do with the number of people on dialup connections
> with dynamic IP addresses.

BGP doesn't work at this level[1]. Dynamic IP doesn't tend to involve any
actual routing updates at all - a single Cisco dialup box with say 60 ISDN
channels has 60 IP addresses permanantly allocated to it. It assigns one
of its pool to you when you dial in. If your call arrives on a different
box, the IP address gets allocated from that box's seperate pool.

Static IP addresses, or network dialups with more than one address, on the
other hand do cause routing updates. Every router within the AS has to
know about which device has received the call. But even in this
circumstance, the actual netblocks within the AS and the routing policy of
the AS doesn't change. Indeed, BGP routing is usually set up manually and
staticly. Having it dynamicaly reflect the routing inside the AS is a
recipe for disaster - say, accidently announcing to all your peers that
your router is the best route to the whole Internet. Say goodbye to your
E3 :)

-- 
Andrew Crawford
Brainbench MVP Linux
http://www.brainbench.com/


[1] normally

--------------------------------------------------------------------
http://www.lug.org.uk                   http://www.linuxportal.co.uk
http://www.linuxjob.co.uk               http://www.linuxshop.co.uk
--------------------------------------------------------------------

Reply via email to