[SAM] IETF-71 meeting for SAM RG

[Xiaoming Fu]

Hi John,

Thanks for your quick response.
John Buford wrote:
> Xiaoming,
>  
> Thanks for raising these points.  Please see the separate email on
> SAM RG 2007 report for a partial answer to the question,
> where are we now?

Maybe it could have been clearer, if action items were more visible to
the RG on what we intend to next. Some additional comments to your comments.

> 
>     http://www.samrg.org <http://www.samrg.org/> states that
>     "The Scalable Adaptive Multicast (SAM) Research Group invetigates
>     multicast protocols including application layer multicast (ALM),
>     overlay
>     multicast (OM), native multicast, and hybrid techniques with the goal of
>     greater scalabality and adaptability in terms of numbers of multicast
>     groups and network resources. It also provides input for
>     standardization
>     efforts within the IETF."
>     "The Scalable Adaptive Multicast (SAM) Research Group is chartered to
>     explore and research techniques which improve multicast performance with
>     respect to dimensions such as number of groups, dynamics of group
>     membership, dynamics of the network topology, and network resource
>     constraints. The RG will investigate approaches based on application
>     layer multicast (ALM), overlay multicast (OM), and native IP multicast,
>     as well as hybrid approaches."
> 
>     --> This indicates that, analysis of existing non-IP multicast
>     approaches would be a GOAL. I have somehow sensed this as possible RG
>     item but no team work being taken nor active draft at least from what I
>     know for now.
> 
> I agree it is it a goal.  We also know that a few research publications have
> compared several leading ALM/OM designs already.  Most of these
> comparisons have been done through simulations.  This requires
> a simulation environment that has a realistic network topology and can
> scale to large number of end systems.  What simulation environment do
> people think we should adopt in the RG?  The ones I could recommend
> from personal familiarity are SSFNet and OverSim.
>  
Performance comparison is one (good) thing, various system design
methodologies and considerations (including deployment aspects) may be
another thing would be of interest. It would be cool if some noteable
publications could be a starting point for generating towards new SAM
framework space, but we may want to check which set of representative
specific approaches would be of interest for the RG and help
understanding the SAM needs.

>     "A key design consideration is the placement of multicast state
>     information along the multicast path, including packet headers, end
>     hosts, and network nodes, where placement may be determined adaptively."
> 
>     --> This implies that the placement of MC state information could be
>     dynamic, which does not tell what we should do.
> 
>     "In SAM architectures, new protocols are expected to coexist and
>     integrate with native IP multicast protocols while offering more
>     flexible deployment options and scaling to support a greater number of
>     simultaneous multicast groups. Alternative technologies such as
>     end-system multicast and overlay multicast have been demonstrated, but
>     these mechanisms must be integrated into a unified architecture and
>     operational design."
> 
>     --> this implies a new (or called "unified") architecture will be
>     defined within the RG. Maybe I am missing something, but where is an
>     architecture or prototypical idea?
> 
>  
> Please see the SAM framework draft.  It was discussed in the IETF-69
> meeting and the slides are on-line.  The basic idea is to partition the
> Internet into domains of different categories of native multicast and 
> ALM support,
> and use an extension to AMT gateways to bridge these domains.  This has the
> advantage of leveraging the AMT work which is already being supported
> by several vendors and for which open source implementations exist.
> However AMT doesn't deal with ALM, only native multicast.  Hence the
> extensions.
Probably I missed it, but I can't remember that there was  a RG
consensus or decision making on moving any existing document as RG item(s).
Concerning this topic itself, my understanding was that, if there are
multiple MC (including IP and various non-IP MC approaches) deployed in
parallel in a single infrastructure, this maybe an interoperability or
interworking issue. Is this what you want to emphasis as the objective
of SAM framework?
> 
>     --> This looks to me we aim to define evaluation metrics and make a
>     systematic evaluation for various existing (non-IP) multicast proposals?
> 
>  
> I think this is saying that the SAM Framework should be evaluated under
> a variety of interesting dimensions.  Are new metrics needed?  Perhaps.
> There are already various metrics in use, but these haven't been used
> in the context of all these dimensions listed above.

It seems we converge here - Maybe we need some document identifying the
metric set as a work item...
>  
> Also, I don't expect to see VPN/GIG multicast work happening in the near 
> term
> since it is not clear how much progress is being made on the GIG itself.

I am not much familiar with GIG; for VPN: at least L3VPN is discussing
MC (RFC4834)
> 
>     "Methods will be explored for group formation and discovery that scale
>     to large numbers of groups, accommodate highly dynamic group membership,
>     and support user-initiated small-group multicast in which the group is
>     defined as a set of explicitly addressed endpoints. Further challenges
>     include efficient multicast for limited-resource nodes and access links,
>     control mechanisms for hybrid systems, approaches to optimization in the
>     network, including routing, and operation in mobile networks, including
>     Mobile Ad-hoc Networks (MANETs)."
> 
>     --> My interpretation would be that SAMRG intends to take group
>     management into account and consider multicast over specific scenarios.
>     For the former, I didn't recall much ML discussion on the design space;
>     For the later if I remember it correctly there is an ID (actually a
>     group of IDs) on mobile multicast recently and even discussed in last
>     IETF, would people here be interested on this and like to explore more
>     details?
> 
>  
> There have been no proposals in the RG that I can remember on ALM in 
> mobile P2P,
> I don't think there has been much work on this in general.
>  
> 
> 
> 
>     "Deploying, diagnosing, debugging, and managing multicast services is
>     complex, particularly for services which span multiple administrative
>     domains. The RG will propose and evaluate tools and strategies for
>     deployment, operations, and management of SAM services."
> 
>     --> I would rather see some document(s) and more discussions on
>     operational experiences on IPTV and  multicast deployments in the
>     Internet. This may allow better understanding of the customers' need and
>     what are the real operational requirements/constraints. Again question
>     is: would people here be interested in contributing and discussing
>     their
>     related work/experiences to the RG?
> 
>  
> Sure, all operational experience is useful.
>  
> 
> 
> 
>     "The RG will select candidates for analysis and evaluation from existing
>     research results. Researchers are invited to submit new approaches for
>     further investigation. Through experimentation, the RG seeks to deepen
>     its understanding of the solution space and to enable the identification
>     of preferred solutions as a function of dynamic network characteristics
>     and the number of multicast sources, receivers, and groups."
>     --> Well, we probably have some good starts in turns of evaluation of
>     existing works, but not much extended and reviewed in the RG - or I am
>     missing any active editors?
> 
>  
> More contributions in all areas of the SAM work areas including evaluations
> would be a good thing.
>  
> 
> 
> 
>     "The expected findings of the RG include characterizing the problem
>     space, including driving scenarios, comparisons and analysis of
>     existing
>     approaches, a SAM framework that supports multiple ALM/OM/ native/hybrid
>     protocols, analysis of network infrastructure impact when multicast
>     traffic becomes a dominant flow in a network, and deployment scenarios
>     which are independent of but can support and evolve with network
>     infrastructure support for native multicast. The findings are expected
>     to be published in technical reports, academic papers, and/or RFCs."
>     --> This sounds reasonable scoping, but I cannot resist ask myself:
>     where are we now? Maybe it would be helpful if we could make the future
>     work agenda more clearly expressed, by defining RG (at least some
>     mid-term) milestones and soliciting active contributors (esp.
>     corresponding editors) for each of the work items, e.g., (roughly
>     sequentially in time but may overlay):
> 
>     * A survey of existing non-IP multicast approaches (btw - did any
>     RG/IAB/IETF WG publish a survey on existing IP multicast protocols? 
> 
>  
> There are a few publications on surveying IP multicast protocols, but 
> not clear
> if these were organized by an WG/RG.

Again, if this is helpful for understanding SAM needs, maybe it would
not hurt to take a closer look first if some individuals would volunteer
to help and contribute...
> 
> 
>     RFC3569? Otherwise, I would suggest also to look at IP multicast
>     protocols too as another potential work item) - Informational
>     * Evaluation metrics of multicast approaches - Informational
>     * An evaluation of Internet multicast approaches - Informational
>     * A SAM framework for interworking (?) between different multicast
>     approaches - Informational
>     * Operational experience of IPTV services (or IPTV operational
>     practices) - Informational
>     * Experience with the xxx Protocol - Informational (or BCP?)
>     * (Suggestions for) Multicast in mobile IP(v6) environments (or some
>     other scenarios) - Informational
>     etc.

> I think there are enough WGs doing native multicast work already,
> but hybrid schemes are in scope for the RG.

Maybe the issue is where the RG wants to investigate. AFAIK native MC is
being deployed in IPTV systems and some other constraint environments,
but not a global Internet, but its implication may be interesting to
know before studying hybrid schemes.

>  
> Thanks for the comments.  Please attend the IETF 71 meeting and
> participate in the agenda.

Thanks - I would be glad to join the meeting and discuss open issues
both face-to-face and in the ML. My impression is that the RG probably
wants to fix a limited set of work items and raise more technical
discussions here esp. contributing and reviewing more related documents
in order to move on... :)

Cheers,
Xiaoming