Meeting Notes Scalable Adaptive Multicast RG Interim Meeting Jan 11, 2007. 1. John Buford (Panasonic) Overview of problem statement draft and the existing technologies available (ALM, Overlay multicast, Proxied overlay multicast, tunneling with native multicast islands). An example architecture and some operational scenarios. 1. Jeremy Mineweaser (MIT Lincoln Labs) Adaptativity depends on suitable metrics and ability to collect these metrics in real-time and feedback to the network. Discussion: KA: Do you think adaptive approach is too complex to be worth the benefit? JM: remains to be seen KA: We don’t really have a one-size-fits-all protocol 2. Mark Pullen (GMU) Unlike commercial ISPs,DoD ISPs have an incentive to use multicast to reduce congestion. But universal deployment in GIG may not work any better than in commercial sector. Gaming community can use multicast in different ways than DoD since they don’t require the real-world fidelity. In battlefield, reliability and robustness may not be fully achievable. We at GMU have a prototype to manage multicast using web services in cooperation with Naval Post-Graduate school. There is a need for a framework to interoperate different mechanisms, a pervasive interoperability should be targeted. Need the pieces to interoperate in order to mix and match technologies. Would like to see a standardized tunneling mechanism as well. Could use the framework for “relative” QoS not absolute QoS. Relative QoS is important for military applications. 3. Keith Ross (Polytechnic) P2P Multicast is an IPTV case study There are several success stories: PPLive, Coolstream, ppstream 200K users at 400-800 kbps for Chinese New Years What are the challenges? - Bandwidth intensive - peers dynamic - asymetric residential access - incentives for redistribution - lags among viewers - security (pollution, overtaking for DDoS) Comparison of two Architectures - Push, tree-based designs, eg. ESM from CMU - Pull, meshed based designs Inspired from BiTorrent But with live streaming More successful than ESM Overview of pull-mesh Locate other peers watching same channel Establish TCP connection Ask neighbors what chunks they have Request and receive chunks Start media player Upload chunks to others Presents some measurements of download vs upload rate for different users Popular and unpopular channels Pull-mesh / IPTV is killer application for multicast JM: what are limitations? KR: free-riding problem, susceptibility to pollution attack Audience: what is the performance comparison between pull and push KR: coolstreaming paper has analysis that shows the ESM is not good under churn JB: what about licensed content? KR: yes, content provides can see this as another distribution channel 4. Kurt Tutschku (U. Wuerzburg) Challenges on scalable multicast in overlay with mobile networks There are several challenges due to mobile nodes in an overlay which translate to problems with overlay multicast with mobile nodes These problems relate to roaming in the native network, node performance capability, variation in network connectivity etc. 5. Wenjun Zeng (U. Missouri) Do you plan to consider reliability, congestion control, etc., in the design? Pull vs push is an important distinction, as made by Keith Is there a hybrid of pull vs push? E.g., tree-bone based approach 6. General Discussion What about data aggregation for sensor data collection (ad hoc wireless networks), is there a multicast solution for this case, and how does it fit? The benefits of an overlay approach for launching SAM framework: Overlays provide a number of interesting functions including search, indexing, relaying, topology characterization, measurement. Integrating awareness of native multicast in to the overlay is another function like these others that a specific overlay can chose to use or not depending on application requirements. Solutions to the problems of overlays in terms of reliability and security can be used for multicast reliability and security perhaps.