Re: [Softwires] Thoughts on new I-D?: Gateway Initiated Dual-stack lite Deployment

[Sri Gundavelli <sgundave at cisco.com>]

Hi Christian,



On 11/4/09 10:49 PM, "Christian Vogt" <christian.vogt at ericsson.com> wrote:

> Sri -
> 
> To clarify:  In addition to saying that an operator COULD decouple the
> NAT from the gateway, it would be good to identify the scenarios where
> the NAT MUST be decoupled from the gateway.  It is these and only these
> scenarios that motivate your proposal.
> 
> But then again, there is at least one motivating scenario:  higher
> efficiency of IPv4 address sharing in networks with multiple gateways.
> 

Sure, we can certainly add the motivation for that configuration. It is to
be noted that the spec is also supporting the allocation of the same IP
address for all UE's. Its solving this in a consistent way for all mobility
architectures (MIPv6, GTP & PMIPv6), by applying dual-stack lite approach in
the gateway initiated mode. Now, if the GGSN and CGN are collocated in a
given scenario, still there is some context sharing between those two
elements that allows the GGSN to route the packet to the UE correctly, it
may be internal to a given implementation. The resulting side-affect of this
is a simple migration approach, with minimal changes to the PDN.

- Solves the IPv4 exhaust issue
- UE can be a legacy IPv4-only node or a dual-stack node
- There is no need for IPv6-only bearer, or IPv6-only transport for UE
- CGN and PDN can be collocated or decoupled. When they are decoupled,
allows the network to be IPv4 or IPv6
- There is no change to the EPC, or to the address assignment procedures
- No changes to the UE architecture

This in essence provides a IPv6 migration approach for mobile operators and
the solution is valid for other architectures as well.


Sri







> - Christian
> 
> 
> 
> On Nov 5, 2009, Sri Gundavelli wrote:
> 
>> Hi Christian,
>> 
>> Thanks for the review. We can surely add some text on the placement
>> of the
>> CGN. Based on the topology and network design considerations,
>> operators may
>> choose to place the CGN in the PDN gateway, or at the edge of the IPv4
>> network with the fixed and mobile networks within the CGN scope. As
>> you
>> pointed out, the collocation of the both the functions on the PDN, do
>> require management of the shared address space. However, the spec
>> does not
>> mandate the placement of the CGN function, as that is a typical
>> deployment
>> and network topology consideration and its up to the operator. We
>> can surely
>> add some text on this aspect.
>> 
>> Regards
>> Sri
> 
> 
>