Is the Internet Ready for Multimedia?
Ahmed Helmy
University of Southern California
Is the Internet Ready for Multimedia?
• What is Multimedia?– Common: Real-time/streaming (audio/video) – More general: Multiple requirements (QoS) by
different applications– What are these requirements?
• End-to-end metrics: Reliability, Delay, Jitter, least cost!
• Aggregate metrics: Least congestion, maximum utilization, fair, minimum energy!
– Is Internet-supported emergency service MM?• Need very high priority service
• What is the Internet?– Wired– Last hop wireless (cellular, access points, space
optics and pico cells)– Mobile (continuous vs. intermittent)– Multi-hop wireless (Ad hoc)– other… !
• Different network characteristics provide different challenges for various requirements– Need to understand characteristics of these
networks• Need models for: topology, traffic, channel, mobility
…
Supply vs. Demand
• Supply– Infrastructure (e.g., ISPs,
content providers)
– Architecture/protocols• IntServ
• DiffServ
• Hybrid!
– Business and pricing models???
• Demand (those who need QoS)– Audio, video, collaborative
virtual env.– Multicast
• Dense-mode, Sparse-model, source-specific
• Inter-domain? (BGMP)• Current: PIM-SM, MDSP,
SSM !
– P2P: effect on network– New paradigms
• Ad hoc/Sensor networks• Active networks
Demand is a moving target.We need to be flexible and think ahead.What’s the next killer app.?
Are we ready? (Questions)• What does QoS mean without end-to-end
deployment of IntServ or DiffServ?
• Can we over-provide resources?– What resources? Bandwidth, computation
(servers), energy ! (sensor nets)– Can we over-provide everywhere?
• Will the solution be a hybrid approach– Over-provisioning when/where possible– QoS architecture otherwise
• What about DoS attacks?– Need fairness and isolation of misbehavior
Are We Ready? (more Qs)
• We talk about network-side QoS
• What about protocol-side QoS?
• If the network is not congested, can we guarantee protocol behavior?
• We need better approaches to design predictable protocols– Can we provide a parallel of circuit-design
CAD tools?
What about emerging networks?
• Example:– Multicast in large-scale ad hoc networks– How to rendezvous in an infrastructure-less
network?• Need to re-think bootstrap and resource discovery
• Need to re-think routing in a highly dynamic env.
• Should we think about it when we do not know how to do it in the wired network?– Good intellectual exercise that has potential
Are we ready? (Answer)
• Of course not
• We have a lot more work to do!