Re: [Ila] [E] Fwd: New Version Notification for draft-herbert-ila-motivation-00.txt

[Tom Herbert <[e-mail]>]

On Sun, Jan 28, 2018 at 10:01 AM, Bogineni, Kalyani
<[e-mail]> wrote:
> Tom:
>
>
>
> Service function chaining is on the SGi interface in 4G and on N6 interface
> in 5G. Each operator decides whether to
>
> deploy mobile core network as an overlay on IP/MPLS network or integrated.
> Standards should allow both deployment
>
> options. Similarly operators could choose to implement several of the SGI/N6
> functions into PGW/UPF.
>
Okay, we'll mention that in the doc.

>
>
> In going through your document again, some comments:
>
>
>
> Section 2.2: It appears that you are using ‘nodes’ for network
> elements/functions as well as for end device like UE in
>
> 3GPP terminology. Is your sentence “Mobility is likely a rate event….”
> related to network functions or to UEs?
>
>
>
> Section 2.5: This section seems to be about mobility of UEs, but ‘node’ is
> used. Some clarification between the two sections
>
> is needed.
>
Yes. I think mobile node should be the generic term for something that
can move around a network. UE and mobile devices a real devices that
move around.

>
>
> Section 2.6.2: Relative to mobile traffic there is Uplink (UL) traffic and
> Downlink (DL) traffic. I think you need to address
>
> caching relative to both UL and DL traffic. With anchor-less mobility, ILA-R
> will need to send the DL traffic to the correct
>
> ILA-N/gNB.
>
Yes, I believe that describes the Downlink traffic and an ILA router
in the path will do the mapping. For Uplink traffic destined to
Internet, ILA is not needed. For Uplink traffic destined to another
node within the ILA  domain ILA transformation is done by either and
ILA router or forwarding node with a cache entry. Does this
description address it?

> In this section you describe a Pub/Sub model. Can it be implemented as
> RESTful APIs? 5G services based architecture
>
> uses APIs from network functions. ILA can be incorporated in 5G services
> based architecture in one of two ways.
>
Good point. I think we should consider if ILAMP and an protocols into
ILA-M (ILA management) can be RESTful. The protocol between the ILA
routers might also be considered, but I tend to think this more like
either a routing protocol (like BGP) or would implement a distributed
database that already has its own protocol.

Tom