Scalable multicast routing : protocol design and performance evaluation

Pao Yue-kong Library Electronic Theses Database

Scalable multicast routing : protocol design and performance evaluation

 

Author: Cheng, Y. Y. Filli
Title: Scalable multicast routing : protocol design and performance evaluation
Degree: M.Phil.
Year: 2000
Subject: Multicasting (Computer networks)
Computer network protocols -- Design
Hong Kong Polytechnic University -- Dissertations
Department: Dept. of Computing
Pages: ix, 121 leaves : ill. ; 30 cm
Language: English
InnoPac Record: http://library.polyu.edu.hk/record=b1535437
URI: http://theses.lib.polyu.edu.hk/handle/200/4431
Abstract: Existing multicast routing protocols create either source-specific delivery trees, e.g., Distance Vector Multicast Routing Protocol (DVMRP) and Protocol Independent Multicast-Dense Mode (PIM-DM), or a shared delivery tree for all sources, e.g., Protocol Independent Multicast-Sparse Mode (PIM-SM) and Core Based Tree (CBT). The source-specific routing protocols were designed to achieve efficient end-to-end delay by using a broadcast-and-prune approach. However, this approach does not scale to both the number of sources and the number of multicast groups. The shared-tree protocols, on the other hand, scale to the number of sources in a multicast group by building a delivery tree rooted at a core router that is shared by any sources. Since multicast packets no longer travel along the shortest path to each member in a multicast group, shared trees incur longer end-to-end delay in the delivery of the packets. The inappropriate location of core router can worsen the delay performance. Furthermore, the shared tree is not scalable to the number of multicast group. Although some protocols, such as PIM-SM, allow a co-existence of source-specific trees for some sources and a shared tree for other sources, the problems mentioned above remain for the respective delivery trees. This thesis addresses two problems to improve the delay and scalability performance of the current multicast routing protocols: 1. How can we obtain a good delay performance as well as a good scalability performance for multicast routing in an effective manner? 2. Given a set of multicast delivery trees, created by some routing protocols, how can we increase their scalability performance? Adaptive Source and Core Based Multicast Because all existing multicast routing protocols cannot address the first problem, we propose a novel multicast routing protocol, coined as Adaptive Source and CORe Based MutlicasT (ASCORT). ASCORT is neither a pure source-specific approach nor a pure shared tree approach. In fact, an ASCORT delivery tree can be considered a two-level structure. In the lower level, each member host is connected to only one source-specific tree, called a reduced tree; therefore, a reduced tree connects only a subnet of member hosts. In the upper level, the reduced trees are connected in the manner of a shared tree, and one of the source's first hop router is selected as a "core router" for this level. While the delivery tree on the lower level is uni-directional, the one on the upper level is bi-directional. The way that ASCORT establishes reduced trees ensures that the resultant reduced trees are disjoint. One simple way to achieving that is to use hop counts from the sources as a criterion for establishing the reduced trees. As a result, each memeber host connects a nearest source's first-hop router, thus minimizing the delay between the root of the reduced tree and the member host. Moreover, the routers on the reduced tree are required to store state information for that particular tree, independent of other reduced trees for the multicast group, thus minimizing the state requirement for each reduced tree. For the bi-directional tree, the root is selected from the sources' first hop routers; experimental results have showed that this selection method achieves better average end-to-end delay as compared with a random method. Extensive simulation studies indeed show that ASCORT achieves good delay as well as good scalability. Besides delay and scalability, ASCORT possesses many other desirable properties for multicast routing: - Unlike the shared tree approach, ASCORT does not require an additional protocol for electing a core router. - ASCORT is a distributed protocol and it incurs a very low overhead for operating the protocol. - ASCORT delivery trees are loop-free given that the underlying unicast routing does not contain loops. - ASCORT is independent of the choice of the underlying unicast routing protocol. - Unlike some shared-tree routing protocols, ASCORT does not use tunneling techniques, thus facilitating multicast routing with QoS requirement. Multicast Tree Switching The current multicast routing protocols fail to scale to the number of multicast groups. There are two possible approaches to this problem. The first one is for a multicast router to perform state aggregation for a number of groups. This approach assumes that there are significant "overlaps" among multicast trees, so that a router may keep states for different multicast groups at the same time. Even when this assumption is valid, multicast state aggregation is not trivial, because, unlike unicast addresses, multicast addressing space is flat. A second approach is to modify the paths of a multicast tree, such that the new path requires fewer states as compared with the original path. In this thesis, we have proposed a new Tree Switching Protocol (TSP) to implement the second approach, and at the same time to increase the tree overlapping, so that the first approach can be applied to further reduce the multicast state requirement. Specifically, given a forest of multicast trees created by the same multicast routing protocol, say PIM-SM, TSP selects a base multicast tree, which is based on some static information, such as the IP address, or others and it allows other trees to switch to the base tree. If a candidate tree cross-over with the base tree on some router, TSP will establish a new tree branch to the base tree, and the old branch will be pruned away. TSP is a distributed protocol and is initiated by a leaf router independently from others. Extensive simulation results have shown that TSP can reduce the amount of multicast state by as much as 30%; with additional state aggregation using leaky and nonleaky methods, the amount of state reduction can be further increased by a significant amount. In addition, the TSP possesses many attractive properties: - The process of creating new branches and pruning the old ones does not create routing loops. - TSP is independent of the underlying multicast routing protocols. - TSP is independent of state aggregation methods. - TSP is adaptive to the degree of overlapping existing in a forest of multicast trees.

Files in this item

Files Size Format
b15354374.pdf 4.510Mb PDF
Copyright Undertaking
As a bona fide Library user, I declare that:
  1. I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
  2. I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
  3. I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.

     

Quick Search

Browse

More Information