IP Multicast In The Next Generation Internet

IP multicast is an efficient multipoint-to-multipoint delivery mechanism in IP network, and accelerates the development of multi-point communications services such as Video on Demand(VoD), video/audio conferences and data distribution. It is widely agreed that Next Generation Internet(NGI) will be built on IPv6 protocol, which provides much larger address space than IPv4. As compared with IPv4 protocol, IPv6 significantly improves routing efficiency, and provides Quality of Service(QoS) guarantee and efficient mobility support by defining simple and flexible IPv6 packet header format and designing network auto configuration mechanism. However, it is rather difficult to guarantee the large-scale deployment of multicast in the Internet and NGI, due to some problems introduced by IP multicast technology such as poor multicast forwarding state scalability. In addition, existing QoS provisioning mechanisms for IP multicast are too complex to provide scalable QoS support for real-time media multicast services in the Internet/NGI even if QoS provision is embedded in NGI.Therefore, this dissertation investigates major issues appearing in the multicast deployment, and presents our approaches to these issues by using new characteristic introduced by IPv6 protocol- specification. The following are main contributions of this dissertation.(1) A multicast forwarding state reduction scheme based on interface format for sparse mode multicast, which aims to minimize storage requirement of multicast forwarding state, is proposed. Since one group member dynamically joins or leaves the current group, the location and number of branching points on the multicast delivery tree will change. The proposed scheme reduces forwarding state storage requirement for sparse mode multicast by dynamically modifying the format of each interface in the multicast forwarding table entry according to the location and number of branching points on the multicast tree. If the number of hops between two neighbor key routers is larger than a threshold, corresponding interfaces in the key routers adopt general reduction format with complex structure, Otherwise, standard multicast format with simple structure, which is used in the ordinary implementation of sparse mode multicast, is adopted. In addition, multicast-capable routers forward incomingmulticast data downstream according to the format of the outgoing interface. If the outgoing interface has standard multicast format, multicast data are sent downstream by ordinary multicast forwarding mechanisms. Otherwise, destination address of replicated multicast packet is replaced with IP address of downstream neighbor key router. At the same time, destination options header of IPv6 Extension Headers, which contains multicast address, is appended to the multicast packet, and then the packet is unicasted downstream. As compared with packet encapsulation, the usage of IPv6 destination options header lowers forwarding process and packet header overloads.(2) A representative-based multicast congestion control scheme, which is based on multiple modes General Additive Increase Multiplicative Decrease(GAIMD) algorithm, is proposed. In the proposed scheme, multicast source uses multiple modes GAIMD algorithm to regulate TCP-friendly transmission rate of multicast session according to feedback messages from Congestion Representative(CR). Multiple modes GAIMD algorithm estimates the change of network situations, which is inferred from Maximum Rate Envelop(MRE) maintained by the source and Interval between the Time when One-way delay is Maximum(ITOM) monitored at the CR, to adapt the increase parameter and decrease parameter, and improves the smoothness and the response to network congestion. At the same time, non-CR receivers assist multicast source in selecting and switching CR based on their congestion situation, which is determined by ITOM information and transmission rate of multicast session at congestion. There is no need to measure exact Round Trip Time(RTT) and suppress feedback messages based on timer at all the rece...