Wednesday, March 12, 2008 Scalable Adaptive Multicast (SAM) Research Group (RG) Minutes of the Meeting at IETF-71, Philadelphia, PA (All times are local) John Buford opened the meeting at 1307 with a review of the agenda. Following that, we briefly reviewed the charter -- specifically, its direct mention of the Global Information Grid. Observing that there has not been much recent RG activity on this front, it was proposed that this portion of the charter should be de-emphasized for now. In particular, the charter refers to multicast over concatenated virtual private networks. In view of recent IETF activity in this area, we asked whether there are still research questions of interest to the RG. Mark Pullen, of George Mason University, commented that the class of solutions being explored by SAM RG is of interest to his C4I study group at GMU. He said that the GIG is a very complex environment, and that no one solution is likely to address its needs entirely. Jeremy Mineweaser encouraged increased participation on the mailing list from GIG researchers. In the first technical presentation of the meeting, John Buford presented the Hybrid Overlay Multicast Framework. This product, which is required by the charter, has been discussed previously; several issues raised at the last RG meeting have been addressed in the 02 revision of draft-irtf-sam-hybrid-overlay-framework. The overarching purpose is to facilitate SAM-related research collaboration. To this end, the framework enables hybrid topologies by allowing application layer multicast trees to use native multicast where available. AMT, currently a draft in the MBONED WG, provides the interface to these native multicast regions. The framework also defines several basic operations for group management. When complete, the framework should identify all of the necessary components of a scalable adaptive multicast solution, but it is not intended to specify "optimal" implementations for components. One open question is whether separate native multicast regions should be able to use different addresses for the same multicast group. Another open issue concerns the (temporary) split of a region. In the second technical presentation, John Buford presented the SAM Overlay Protocol, which is currently defined in draft-buford-irtf-sam-overlay-protocol. The draft assumes the existence of a structured overlay which connects a set of peers into an ALM connection. Buford explained that the P2PSIP WG is defining an "algorithm agnostic" protocol which may provide a foundation for SAM. The draft defines a set of operations which SAM overlay protocols must support, and illustrates the use of these operations for several scenarios. These include GET/PUT operations on a distributed hash table. Finally, the draft specifies the message format. During this presentation, the group discussed the need for clear requirements, prior to specifying the protocol. To encourage additional input and feedback on the existing (now expired) requirements draft, it will be reissued in a new revision. In discussion, it was said that the requirements document would be especially helpful, given the broad scope of the RG charter. It was proposed that the framework should be published as an Informational RFC. Experimental validation of the overlay protocol document is an open question, but the group would like to get feedback from implementers. In the third technical presentation, Nobuo Kawaguchi described the current status of the Scalable Adaptive Multicast Tool Kit, or SAM-TK, a middleware product intended to bridge the gap between protocol researchers and application developers. The toolkit allows developers to experiment with and evaluate a variety of SAM-relevant algorithms and technologies. He described several improvements made since IETF-69. In a live demonstration of the toolkit, he illustrated its features for group management, along with plugin support for multicast protocols. Next, Lim Boon Ping introduced a new draft describing an Application Layer Multicast Application Programming Interface. The draft focuses on topology management and network layer transport. The proposal is to use the API from SAM-TK to enable application developers to evaluate the performance of a variety of multicast protocols without making protocol-specific modifications to their own software. This would be implemented as a "wrapper API". Initially, the API would provide for protocol-independent topology management; later, traffic management functionality would be added. During this presentation, there was a question about the definition of a "path", and it was argued that its meaning depends on the specific multicast scheme in use. The speaker agreed that this was a difficult aspect of defining a wrapper API. In the final presentation of the meeting, Thomas Schmidt revisited the topic of an Application Programming Interface for SAM, with the goal of making the API transparent to whether the multicast protocol is native IP multicast or application layer multicast. With this API as "middleware," application developers would be freed from the implementation details of the underlying multicast mechanisms. This API can serve as a common view of the Application Layer Multicast stack, or of the Peer-to-Peer stack, more generally. To avoid duplicating functionality in SAM middleware, Schmidt proposed a "wrapper API" that would normalize the interface to group join/leave and other common functions. Due to time limitations, the RG had to postpone a planned presentation by Yuji Imai on version 2.0 of XCAST6.