Re: Forwarding Path & Signaling Management discussion

Alper,

Fair point on the complexity involved in getting this right. I agree with
that assessment.

Not having it as part of the baseline also carries the risk of getting
this interface tagged as an incomplete-interface and
not-ready-for-deployment.

Lets see how the interface shapes up, based on that we can decide.


Regards
Sri

 


On 10/16/14 10:32 PM, "Alper Yegin" <[email protected]> wrote:

>
>Sri,
>
>>> - Should QoS really be part of this work? I think notŠ
>> 
>> [Sri] There is QoS support in all mobility architecture. QoS is tied to
>> subscription and a key function of the mobility gateway. The CP entity
>> should be able to provide the QoS parameters to the DP entity
>
>I'd say let's not deal with it in DMM WG. Or at the least, let's not
>tackle it in the first phase of DMM work.
>Otherwise we are making our life an order of magnitude harder from very
>beginning.
>
>Alper
>
>
>
>
>
>
>
>
>
>On Oct 16, 2014, at 11:58 AM, Sri Gundavelli (sgundave) wrote:
>
>> Alper,
>> 
>> Inline Š
>> 
>> 
>> On 10/16/14 1:32 AM, "Alper Yegin" <[email protected]> wrote:
>> 
>>> Hello folks,
>>> 
>>> Here are few clarification questions and comments on the Forwarding
>>>Path
>>> & Signaling Management WT discussion material (oh, maybe it's time we
>>> start numbering these WTs :-)
>>> 
>>>http://www.ietf.org/proceedings/interim/2014/10/07/dmm/slides/slides-int
>>>er
>>> im-2014-dmm-3-1.pdf
>>> 
>>> 
>>> - Are we not going to support roaming? Or, even w/o roaming, I'd think
>>> there'd be inter-domain signaling, e.g., DPA and DPN belonging to two
>>> different networks.
>> 
>> 
>> [Sri] I agree, Roaming should be supported. That's a good point. That
>>is a
>> intra-controller interface.
>> 
>> 
>>> 
>>> - What is the difference between DPA and DPN? We need definitions for
>>> these terms.
>> 
>> [Sri] 
>> Data Plane Anchor
>> Data Plane Node
>> 
>> We had some definitions in the IETF slides that we discussed in the last
>> meeting 
>> 
>> 
>>> 
>>> - Slide 8: Is this simply showing how BGP-based solutions can be
>>>covered
>>> here? Or something else?
>> 
>> 
>> [Sri] The interface can be realized using multiple approaches, one being
>> the BGP-based approach, other explicit signaling based interface
>> 
>> 
>>> 
>>> - What is the relationship between Slides 6-7-8?
>> 
>> 
>> [Sri] I though we had some discussion on this in the last IETF. In the
>> context of Forwarding path, the key point is the interface between a
>> control plane entity and a data plane entity.
>> May be Marco has a better explanation
>> 
>> 
>> 
>>> 
>>> - What is a controller "type"?
>> 
>> [Sri] In deployments there could be controller entities on function
>>basis;
>> Ex: Mobility Controller, Routing controller ..etc. I agree with you we
>> should not decompose the controller, but should look at this as one
>>single
>> entity. 
>> 
>> 
>> 
>>> 
>>> - Should QoS really be part of this work? I think notŠ
>> 
>> [Sri] There is QoS support in all mobility architecture. QoS is tied to
>> subscription and a key function of the mobility gateway. The CP entity
>> should be able to provide the QoS parameters to the DP entity
>> 
>> 
>>> 
>>> - Why would DP be querying CP?
>> 
>> [Sri] Keeping the interface between CP and DP flexible allows us to
>> support all types of scenarios. Lets take a simple example of a DP node
>> loosing state, there should be mechanism to make a query to the CP
>>entity.
>> 
>>> 
>>> - Slide 13: It's not immediately clear why we need these identifiers,
>>>and
>>> how they are used. Maybe there's a solution behind them, but the reader
>>> cannot get the picture.
>> 
>> [Sri] These are described at a high-level; Subject to change. The draft
>> should have some details on the semantics and structure.
>> 
>> 
>>> 
>>> Thanks.
>>> 
>>> Alper
>>> 
>>> 
>>> 
>>> 
>>> 
>>> 
>>> 
>>> _______________________________________________
>>> dmm mailing list
>>> [email protected]
>>> https://www.ietf.org/mailman/listinfo/dmm
>> 
>