[ English | Japanese ]
This dissertation addresses the multicast forwarding state problems in multicast network deployment and the feedback message implosion problems in multicast communications. The objective of this dissertation is to realize large scale multicast communications, which will exist when multicast network is deployed globally. To solve the above problems, this dissertation first proposes to use satellite unidirectional links as multicast paths between international backbone networks and regional or stub networks of the Internet, and then proposes two methods to scale multicast communications by minimizing the number of feedback messages, called two-step random timers and Tracked Sender List.
Multicast network deployment has not been widely deployed, which is largely attributed to the IP multicast architecture, where all routers along the path of multicast traffic have to maintain forwarding states for each corresponding multicast flow. This dissertation proposes to deploy a global multicast network using satellite unidirectional links in order to bypass terrestrial links, hence reducing the number of routers that have to maintain the forwarding states. In this proposal, the Internet backbone networks become the transmitters for satellite unidirectional links, called UDL feeds, while regional and stub networks become UDL receivers, which receive data transmitted by the UDL feeds. Simulations were done on several deployment scenarios, where each scenario consists of the selection of UDL feeds and UDL receivers, and the selection of UDL receivers depends on the network and geographical distances of the receivers to their UDL feeds. The results showed that such a deployment of a global multicast network benefits all networks that are connected to satellite unidirectional links.
On scaling multicast communications, the main question being addressed is how to minimize the number of feedback messages if only a single feedback message is needed, while the feedback message has to come as soon as possible and the multicast group has a large number of nodes. This dissertation proposes a class of probability density functions (p.d.fs) for random timers to delay sending feedback messages, called two-step random timers. Two-step random timers use the knowledge of network delays to minimize the number of feedback messages by making a density function transition based on the network delays. Analysis and simulation results showed that two-step random timers can give better results than the widely-used uniform and exponential p.d.fs.
The second proposal on scaling multicast communications is each receiver stores the receivers that recently sent feedback messages in a list, called Tracked Sender List (TSL), and allows the receivers in the list to send feedback messages early at the next feedback round. In this proposal, each receiver maintains its TSL autonomously and determines its random timer to delay sending feedback messages based on whether it finds itself within its own TSL. The message suppression performance of TSL were evaluated using analysis and simulations; and the results showed that TSL gives a very good performance for a wide range of receivers.
The proposals in this dissertation, which are to use satellite unidirectional links for global multicast network deployment and to minimize feedback messages using two-step random timers and Tracked Sender List, contribute to support multicast services on the Internet.
Keio University, Graduate School of Media and Governance
MAUI Project
Ph.D. Dissertation
Back to Index Page
ACADEMIC YEAR
2005
NAME
Achmad Husni Thamrin
TITLE
Global Multicast Network and Scaling Multicast Communications
ABSTRACT
CONTACT
To obtain the whole paper, please contact;
Achmad Husni Thamrin (husni@sfc.wide.ad.jp)