Re: [RRG] LISP and IP Interworking - Anycast PTRs == Ivip

[Robin Whittle <rw@firstpr.com.au>]

Hi Darrel,

Quoting my earlier message, you wrote:

>> Normal BGP practice is to have one router advertising a
>> particular prefix.  If two or more do, I think it is
>> reasonably described as "anycast".  The principles are the
>> same whether it is 2 or 20,000 routers advertising the same
>> prefix.
>
> One short point here.  You are incorrect above.  In the Internet,
> almost every prefex/aggregate is originated from more than one
> router - you'd not want an entire aggregate disappearing if a
> single router died.

Can anyone else comment on this this?  I can imagine it is done for
some prefixes, but "almost every" prefix?


> If multiple routers originate the same prefix, then the next hop
> ip address (usually the routerID/loopback of the originating
> router) is different.  This causes the BGP path selection to
> prefer/propagate one over the other.  This is normal.

Yes.


> In anycast, multiple devices are configured with the same ip
> address, nothing more.  Its quite different than having the same
> prefix announced from different sources.

I disagree with your second sentence.  AFAIK, in the context of BGP
"anycast" involves multiple routers (perhaps "many" rather than just
2 or a few) advertising the same prefix.


> Selective route announcements and anycasting are two very
> different ways to control traffic.

Google finds only 3 pages for "selective route announcements" and
none for the singular. Googling:

   "selective announcement(s)" BGP

finds about 34 pages.  I assume you mean this term to encompass
NOEXPORT though none of these pages mentions NOEXPORT.

Yes, but NOEXPORT and anycasting could be combined - as I think
Eliot did.

Eliot proposed multiple announcements, all over the Net, which I
interpreted as "anycast" - but with NOEXPORT (a selective route
announcement), which is arguably just anycast with limited
propagation.  I don't know what purpose NOEXPORT has, since it
limits the capacity of the system to provide connectivity from hosts
in non-upgraded networks.

In the ivip-arch-00 ID, I stated that Ivip's approach may be an
unusual form of anycasting:

   1.5 Anycast ITRs

   Multiple routers, usually each with an associated server,
   advertising the same prefix is known as "anycasting" [RFC1546]
   [ISC-Anycast http://www.isc.org/pubs/tn/isc-tn-2003-1.html].
   Ivip's use of multiple anycast routers may be novel: tunneling
   packets to a single tunnel endpoint, which forwards the packets
   to a single host.

With this proviso, I still think "anycast" is an appropriate term,
given that this new arrangement is reasonably similar to other
scenarios to which the term is applied.

In http://en.wikipedia.org/wiki/Anycast I find:

   On the Internet, anycast is usually implemented by using BGP to
   simultaneously announce the same destination IP address range
   from many different places on the Internet. This results in
   packets addressed to destination addresses in this range being
   routed to the "nearest" point on the net announcing the given
   destination IP address.

This uses "many", which is different from my statement about "2 or
more" - but I believe the technical principles are the same whether
2, 3 or "many" BGP routers advertise the same prefix.

RFC 1546 (1993) doesn't mention BGP.  It discusses the problems of
using a stateful protocol such as TCP with anycast, when there are
multiple destination hosts, with the packet arriving at any one of
them, according to the current state of the routing system.  This is
on the reasonable assumption that the separate destination hosts
can't coordinate their activities in order to allow TCP
communications to continue when the packets in one session are
forwarded to different destination hosts.

RFC 1546 then goes on to state:

   Stateful protocols will have to employ some additional
   mechanism to ensure that later datagrams are sent to
   the same host.  Suggestions for how to accomplish this
   for TCP are discussed below.

   After considering the two examples, it seems clear that
   the correct definition of IP anycasting is a service
   which provides a stateless best effort delivery of an
   anycast datagram to at least one host, and preferably
   only one host, which serves the anycast address.  This
   definition makes clear that anycast datagrams receive
   the same basic type of service as IP datagrams.  And
   while the definition permits delivery to multiple hosts,
   it makes clear that the goal is delivery to just one host.

I believe that in all important principles, LISP's Proxy Tunnel
Router scheme is identical to Ivip's "anycast ITRs in the core" scheme.

For instance, both schemes honour RFC 1546s preferred goal of
delivery to just one host - the destination host whose address is
mapped by Ivip or LISP.

The example in RFC 1546 involves just two hosts, X and Y,
advertising the one address.  I see no mention that the number of
advertising routers (or hosts) needs to be "many" - that is, more
than two.  RFC 3258 references RFC 1546 for its definition of anycast.

Can you point to any definition or widespread use of "anycast" which
precludes having only two, or some similarly small number, of
simultaneously advertising routers?

The semantics of anycast is a side issue to the question of how
similar or identical LISP Proxy Tunnel Routers are to Ivip's
"anycast ITRs in the core", or perhaps "anycast ITRs in the DFZ".

By the way, I don't see the "Proxy Tunnel Router" as a proxy for
anything.


> P.s. I'll make a reply to the list re: your comments on the draft
> next week when I have some more time.

OK!

  - Robin

--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg