multihoming

[Rik van Riel]

On Sun, 28 Apr 2002, Michel Py wrote:

> > Bill Manning wrote:
> > tainting/overloading the routing table for IPv6 is
> > a non-issue at this point in time.  Some 400 prefixes
> > are active.  What ought to be of more concern is the
> > need to get a /48, -per provider- to deal w/ multihoming.
> > This does not bode well for effective address conservation.
>
> You are missing the point, IMHO. Granted, a /48 per provider is not the
> most efficient allocation, but please keep in mind that there is an
> almost unlimited number of /48s.

> On the other end, there are practical limits to the size of the routing
> table, end even if we had unlimited CPU, memory and bandwidth resources
> it would become a manageability issue anyway.

I guess this means for multihoming people just need multiple
netblocks (/48s or /56es) and the next higher layer of the
network stack needs to know about a host having multiple
addresses.

Good thing the SCTP people are already working on this.

Note that something like SCTP will always scale ... if your
server can handle 1000 connections, it can handle those 1000
connections with the remote ends having 4 possible addresses.

Figuring out why routing tables don't grow linearly this
nicely is left as an exercise for the reader ;)

regards,

Rik
-- 
Bravely reimplemented by the knights who say "NIH".

http://www.surriel.com/		http://distro.conectiva.com/