Scalability in Xcast-Based Multicast
Date Issued
2004
Date
2004
Author(s)
Yang, De-Nian
DOI
en-US
Abstract
Scalability is one of the most important reasons that prevent multicast from universal deployment. The scalability problem of multicast is that the network can not support a large number of members in a multicast group and a large number of multicast groups in the network due to the following two reasons. First, in control plane, each router needs to store lots of multicast forwarding states, and the distribution of multicast forwarding states among routers is not balanced. Therefore, each packet suffers from a large forwarding delay in each router, and each router requires a large memory to store all forwarding states. Second, in data plane, the trees used in current multicast mechanisms consume much more bandwidth than the optimal multicast trees, namely, the Steiner trees. Current multicast mechanisms waste the network bandwidth such that the network can not support a large number of multicast trees.
In this dissertation, we solve the scalability problem of multicast from both the control-plane perspective and the data-plane perspective. In control plane, we propose a new multicast forwarding mechanism that is scalable in terms of both the number of members in a group and the number of groups in the network. Our mechanism reduces the number of forwarding states stored in each router and balances the distribution of forwarding states among routers. In data plane, we propose a new multicast delivery mechanism that uses much less bandwidth than current multicast mechanisms. Both proposed mechanisms adopt an emerging IP forwarding mechanism, Explicit Multicast (Xcast).
Since Xcast is a new IP forwarding mechanism, the forwarding and routing problem in our mechanisms is much different from current multicast mechanisms. In this dissertation, we model the forwarding and routing in our mechanisms as optimization problems. We formulate them as integer linear programming problems, propose approximation and heuristic algorithms for our problems, and design protocols for our mechanisms. We also conduct simulations for all proposed algorithms. Our results shows that our mechanisms requires much fewer multicast forwarding states than current multicast mechanisms, and the distribution of forwarding states among routers is more balanced. Beside, our mechanisms also use much less bandwidth than current multicast mechanisms. Therefore, we believe that our mechanism can solve the scalability problem of multicast.
Subjects
群播
路由
網路
multicast
network
routing
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-93-F88921033-1.pdf
Size
23.31 KB
Format
Adobe PDF
Checksum
(MD5):0f73e927b3bc0d0c5da95c86934bb999