Multicast Tree Construction in Directed Networks

Significant interest exists within the military in moving towards an integrated services environment where traditional network services such as ftp, telnet, and e-mail can co-exist with real-time services such as voice, video, and satellite imagery. Multicast routing is an effective means of providing the efficient utilization of network resources required to realize such an environment. Traditional multicast routing algorithms assume a symmetric network topology. Many military communication assets are either asymmetric in their load or asymmetric in capacity (a good example is Direct Broadcast Satellite). In addition, many military communication assets are bandwidth constrained, and routing symmetrically may further contribute to congestion. Therefore, multicast tree construction which tolerates network asymmetry is desirable for many military communication environments. This paper proposes an algorithm for constructing shared multicast distribution trees in networks with asymmetric link capacities or loads. The algorithm tolerates asymmetry by building distinct, loop-free, sender and receiver paths onto a shared delivery tree. Additionally, the algorithm exhibits desirable security properties. Simulation results are presented that demonstrate the lower tree cost and better load balancing characteristics of the resultant trees over shortest path trees, with only a modest increase in path length. Introduction Current multicast routing protocols were developed under a paradigm that assumes network link symmetry. Many fixed networks and satellite or wireless networks exhibit asymmetry in capacity or load. In addition, traditional multicast routing algorithms utilize "reverse-path" routing which in asymmetric networks may lead to poor routes. Several studies [10, 3] have been made that confirm the existence of network asymmetry in the fixed networks that make up the internet There are two basic approaches to multicast tree construction. The first is a shared multicast tree [1] and the other is a source rooted tree [13, 9]. The shared tree approach uses a single tree rooted at some center that is shared by all participants. In the source rooted approach, each sender builds a separate tree rooted at itself. The algorithm presented in this paper builds a shared tree that can tolerate network asymmetry. To build this shared tree, the algorithm creates forwarding state in multicast capable routers. This state represents disjoint, loop-free, paths for senders and receivers. The results of simulations conducted over varying topologies show that the resultant trees provide a lower tree cost than source rooted trees with only a modest increase in end-to-end delay. Tree Construction The join process is similar to the CBT approach [1] with some slight differences due to the asymmetry. A sender joins a tree by propagating a join-request message along the shortest path to the tree center. When the join-request reaches the center, a join-ACK message is sent back along the same path. A receiver joins in a similar manner by propagating a join-request to the center along the shortest path. The joinACK is then sent back to the receiver along the shortest return path (which may be different than the path taken by the joinrequest). Since senders must explicitly join the tree (and thus the group) this violates the traditional multicast model of "senders just send"1. However, this has some advantageous security properties in that sources can be authenticated before being admitted to the tree. The disjoint paths could result in routing loops if only group state is stored (as in CBT). Thus, to prevent routing loops the protocol must store source state as well, which has some scaling implications [7]. Thus, each router maintains a sender list (SL) per active group. The SL contains the following information [Sender ID, incoming interface, {set of outgoing interfaces (OGI)}]. The algorithm for adding a new sender is given below: