Reliable Multicast Network Configuration

可靠组播作为下一代信息传输系统的技术，受到了广泛的关注。它保证了大规模组播中信息的无差错传输。然而，可靠的组播方法存在两个主要问题。一种是，当只有发送方同时向所有接收方重新传输丢失的分组时，随着接收方数量的增加，它们不能很好地工作。这种重传方法称为全局恢复。另一种是ACK内爆。网络拥塞是由大量返回的ACK集中引起的，本地恢复机制就是为了解决这些问题而提出的。它可以通过对节点分组来定位重传的影响。目前已经提出了几种重传方式，如使用单播、组播和子播，这些重传方法结合本地恢复系统将是更好的解决全球网络大规模组播的方法。然而，适当的重传方法可能取决于网络拓扑。因此，本文比较了基于本地恢复的重传方法，并找到了一种适合于典型网络拓扑结构的重传方法：星形、环形、增量-星形以及代表当前Internet拓扑结构的Waxman模型。从网络建设成本的角度讨论了适合可靠组播的网络拓扑结构。

Reliable multicast has got much attention as a next generation's technology on information delivery systems. It guarantees the information deliveries without errors on large-scale multicasting.However, reliable multicast methods have two major problems. One is that they cannot work well as the number of receivers increases when only the sender retransmits lost packets to all of receivers simultaneously. This retransmission method is called the global recovery. The other is ACK implosion. Network congestion occurs by concentration of many returned ACKs.Local recovery system has been presented to cope with these problems. It can localize the influence of retransmissions by grouping nodes. Several ways of retransmissions, for instance using unicast, multicast, and subcast, have already been proposed.This study focuses on these retransmission methods with local recovery systems since they must be the better solutions for large-scale multicasting on worldwide network. The appropriate retransmission method, however, may depend on network topologies. This paper therefore compares local-recovery-based retransmission methods and finds the suitable one for typical kinds of network topologies ; star, ring, delta-star, and Waxman's model which represents current Internet topology. The suitable network topology for the reliable multicasting is also discussed from the viewpoint of network construction cost.