Re: [5gangip] FW: New Version Notification for draft-xyx-5gip-ps-00.txt

[Behcet Sarikaya <[e-mail]>]

On Wed, May 10, 2017 at 12:19 PM, Tom Herbert <[e-mail]> wrote:

> >> Extension headers are end to end constructs and there is no obvious
> >> way to limit their scope to a single network (there is a long ongoing
> >> discussion about whether they can be inserted by intermediate nodes
> >> but the camp that wants that doesn't seem likely prevail). Besides
> >> that there is a good chance of packets with extension headers being
> >> dropped. Encapsulation is straightforward, can be significantly
> >> contained, isn't that much more overhead, and allows the option of
> >> other features for the encapsulation (security, fragmentation,
> >> checksum help, etc. like in draft-herbert-gue-extensions).
> >>
> >
> > Why end to end is not OK?
>
> Because that would imply that for any host the Internet to talk to a
> mobile host it would have to implement the extension header and
> mobility protocol. Any solution that requires updating all the hosts
> on the Internet to support some functionality is a non-starter.
>
>
I think you are saying that we need proxies.



> > Encapsulation is to me defeats the purpose, you tunnel to where? Anchor
> > point? Then we have distributed anchors?
> >
>
> For a host talking to a UE I would implement encapsulation or ILA from
> a carrier's gateway to the eNodeB. Process would look something like:
>
> 1) Host sends packet to UE's (virtual) address.
> 2) Packet routed over Internet to a gateway of the carrier
> 3) Gateway looks destination in mapping table, returns a locator
> (physical address) of UE's current eNodeB
> 4) Packet is encapsulated (or ILA translated) with destination for eNodeB
> 5) Packet is routed over RAN to eNobeB
> 6) At eNodeB remove encpasulation (or reverse translate ILA)
> 7) Deliver packet to UE. This is now the same packet that the Internet
> host sent.
>
>
>                                                    ←----
> Encapsulation/ILA path---->
>                             _______                              _____
> +-------+               (             )     +-------------+       (
>     )     +-----------+     +------+
> | Host |--------------(   Internet  )---|  Gateway |------( RAN
> )------| Enodeb |-----|  UE  |
> +------+                (_______)     +-------------+       (____)
>   +-----------+    +-------+
>
> Packets in the reverse direction, from UE to Host should need any
> encapsulation or ILA. The can be sent as is and just following routing
> to get to appropriate egress gateway.
>
>
Yes


> When a UE moves to a new UE all the mapping tables need to be updated
> this will take time. To mitigate that hit, I would probably place a
> COA at the old eNodeB to forward packets (using encapsulation) to the
> new location for some period of time.
>
>
These are all fine.

I was talking about ILA case without proxy and that's where we can consider
 extension header with a destination option.

I think that ILA as in data center use case (draft-herbert-nvo3) can also
find its niche in 5G world. Marco Liebsch had some ideas about that and I
am expecting he is going to post them in the list.

Behcet

Tom
>
> >
> >
> > Behcet
> >>
> >> Tom
> >>
> >> >
> >> >
> >> >
> >> >
> >> > [Uma]:  Not sure if https://tools.ietf.org/html/draft-ietf-lisp-mn-00
> >> > has
> >> > been discussed. I see it has IPv6 support from beginning..
> >> >
> >> >
> >> > _______________________________________________
> >> > 5gangip mailing list
> >> > [e-mail]
> >> > https://www.ietf.org/mailman/listinfo/5gangip
> >> >
> >
> >
>