Message info
 
To:Hampel, K Georg (K Georg) From:Hampel, K Georg (K Georg) Subject:Re: [multipathtcp] Proposal for revised charter: MPTCP MOBILITY Date:Thu, 29 Mar 2012 02:18:07 -0500
 

All,

One important issue to be addressed for MPTCP MOBILITY is the overhead optimization when only one path is available. In wireless access environments, this applies to a large fraction of traffic. For MPTCP MOBILITY use cases, only 1 path is used for almost all the time.

When only 1 path is available there is no principal reason to require full-fledged 2-layer operation, i.e. insert/process DSS options or to require subflow-level buffering.

This item needs more work since such optimization should not jeopardize middle-box compliance.

Regards,

Georg

-----Original Message-----
From: multipathtcp-bounces@ietf.org [mailto:multipathtcp-bounces@ietf.org] On Behalf Of Hampel, K Georg (K Georg)
Sent: Wednesday, March 28, 2012 1:44 PM
To: Olivier.Bonaventure@uclouvain.be
Cc: multipathtcp@ietf.org
Subject: Re: [multipathtcp] Proposal for revised charter: MPTCP MOBILITY

Olivier,

I agree with all of your points. I also agree that we should NOT limit MPTCP to the mobile environment. But that shouldn't prohibit us from adding Mobility as one focus point to the charter.

The same applies to proxies: They are not needed for data centers but we still seem to agree that they deserve a spot on the revised charter.

Regards,

Georg


-----Original Message-----
From: Olivier Bonaventure [mailto:Olivier.Bonaventure@uclouvain.be]
Sent: Wednesday, March 28, 2012 11:46 AM
To: Hampel, K Georg (K Georg)
Cc: multipathtcp@ietf.org; philip.eardley@bt.com; Yoshifumi Nishida
Subject: Re: [multipathtcp] Proposal for revised charter: MPTCP MOBILITY

Georg,

>> 1) Strong use case:
>>
>> MPTCP allows 3G/4G mobility operators to dynamically offload traffic
>> from their air interface to unlicensed spectrum. When deployed as
>> end-to-end solution, it also permits dynamically offloading the 3G/4G
>> core. Together with MPTCP's middlebox compliance, there is no other
>> solution that offers such features.
>>
>
> This already works with the current specification. We have an
> implementation that is already able to perfom WiFi/3G offload and we are
> performing measurements in real networks.
>
> We already have some experience to prepare a report describing how MPTCP
> can be used in mobile
> environments. If the WG thinks that such a report would be useful, we
> can prepare a first draft.
>
> Based on our previous work, we also have some ideas on the datacenter
> usecase for MPTCP and could also contribute to an applicability draft in
> this domain.
>
> [GH] That's all fine, but it doesn't affect the validity of point 1).

We should document use cases of MPTCP. We are ready to contribute by
preparing a draft on :
- mptcp use case for mobile offload
- mptcp use case for datacenters

MPTCP could also play a role in the IPv4/IPv6 transition and we would be
ready to contribute to a draft to discuss this use case as well.

>>
>> There is a large group of people, e.g. mobile operators, interested to
>> find a solution to the above use case and they NEVER think of MPTCP
>> unless it gets the term "MOBILITY" explicitly spelled out in its charter!
>
> Mobile network operators willing to see a demonstration of MPTCP in a
> mobile Wifi/3G offload scenario can contact us offline...
>
> [GH] ...if they realize that MPTCP offers a "mobility solution for traffic offload". Mobility-seeking people don't trigger when you say "multipath". I don't understand why you want to miss out on this opportunity.


I agreed that MPTCP can play a role in this domain and we should
probably improve the marketing of MPTCP. Documenting use cases would be
a good way to improve the visibility within the IETF, but not all
operators follow IETF.

Concerning the charter, we should not limit ourselves to the mobile
environment. MPTCP has the same use cases as TCP and thus has broad
applicability.


Olivier

--
INL, ICTEAM, UCLouvain, Belgium, http://inl.info.ucl.ac.be
_______________________________________________
multipathtcp mailing list
multipathtcp@ietf.org
https://www.ietf.org/mailman/listinfo/multipathtcp
_______________________________________________
multipathtcp mailing list
multipathtcp@ietf.org
https://www.ietf.org/mailman/listinfo/multipathtcp