[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lisp] LISP Interworking: Proxy Egress Tunnel Routers



Jari, 

> Of course, I understand that pretty much any tunneling or translation 
> technique can be used for IPv6 deployment as well. I do not 
> mind if such 
> use case gets mentioned in a list of other potential uses of Lisp. 
> However, the main use case, deployment rationale, etc. needs 
> to stand on 
> its own and talk about routing scalability, traffic engineering, and 
> other things central to Lisp. 


Thanks for your note, I have a few comments that follow about the
intended scope of LISP Interworking.  I don't see a conflict with what
you've written, but I want to try to clarify my position on LISP
Interworking for the WG. 

A central design goal of LISP is that it be practical, that is, work
within the existing, deployed, Internet.  Today that Internet has a
mixture of IPv4 only, dual stack, and IPv6 only hosts and networks.  It
also is comprised of a myriad of access technologies, security
mechanisms, that LISP xTR must integrate with in order to be deployable.
LISP attempts to solve the routing scalability problem for both IPv4 and
IPv6, and therefore LISP sites must interwork with non-LISP sites of
both IP protocols.

A central benefit of network based Interworking is how it enables the
benefits of LISP for the site which has deployed LISP:

On ingress, using Proxy Ingress Tunnel Routers, this is primarily the
enablement of LISP's Ingress Traffic Engineering features.  One result
of this is that LISP sites relying on Proxy Ingress Tunnel Routers will
see all ingress traffic LISP encapsulated - and therefore see this
benefit of LISP immediately.  I might add, a second order benefit is
that if the Proxy Ingress Tunnel Router's RLOCs are connected the a Dual
Stack IPv4/IPv6 network, the connectivity from the Proxy Ingress Tunnel
Router to the site's ETR does not have to be Dual Stack.  Proxy Ingress
Tunnel Routers also are a primary tool to encourage sites not to Inject
EIDs into the Internet routing system.  They in effect make EID's
routable without direct injection (and inevitable degradation) of the
EID space by individual sites.

For a site using Proxy Egress Tunnel Routers on egress there are fewer
LISP specific benefits.  This is because LISP does not attempt to
enhance Egress Traffic Engineering (instead utilizing the site's
existing egress IP routing mechanics).  The primary use cases for Proxy
Egress Tunnel routers are actually only to address practical deployment
concerns. These secondary issues include working around limitations to
the installed access networks' implementation of BCP 38 filtering, and
working around the lack of support found in access networks for dual
stack IPv4 and IPv6.  So this means that not every LISP-Site will need a
Proxy Egress Tunnel Router, but we know from our experimentations to
date that it will help some of LISP sites, and this instigated the ideas
and proposed text found in this thread.

Note that both of these network based approaches are not required if
LISP-NAT (currently specified only for IPv4)is used.  LISP NAT instead
uses routable EIDs that fall within the Provider Assigned address space.
But a LISP site using LISP-NAT misses out in the advantages of ingress
TE above, and will obviously not be IPv6 capable.  


>I also understand that what we develop 
> technology X for may not be what it eventually gets used for; 
> we've been 
> surprised before. Lets see what happens with Lisp. But for 
> the RFCs that 
> come out of this working group, lets focus on the initial main reason 
> why Lisp was developed.
> 

Understood, and I think we are.  My opinion is that Proxy Egress Tunnel
routers are an interesting option that I think will assist in LISP sites
to communicate with non-LISP sites.

-Darrel

Note Well: Messages sent to this mailing list are the opinions of the senders and do not imply endorsement by the IETF.