The Coexistence of Multicast and Unicast over a GPS Capable Network

Abstract

The last few years have seen a great number of new network applications. Many of these applications are characterized by their multipoint features and by their needs of high bandwidth. To allow the deployment of these applications at a large scale in the Internet without running the risk of network implosion, an adapted congestion control mechanism has to be used. In this paper we propose to use the Fair Scheduler paradigm for end-to-end congestion control for the design of a congestion control algorithm to be used with the Reliable Multicast Transport Protocol RMTP. The FS paradigm supposes that all the network routers use a fair scheduler queueing mechanism. This hypotheses allowed us to have worst case bounds on the queueing delay that a block of RMTP data packets experiences between the sender and any receiver in the multicast group. Based on these bounds the sender will regulate its transmission rate in order to avoid a network congestion. Two different versions of this algorithm are presented in detail. The coexistence RMTP/TCP will be the issue of a suite of tests that intend to evaluate the performance of the new algorithm over a Fair Queueing (FQ) scheduled network with the presence of TCP flows sharing the same network resource as the RMTP flow. The same tests will be rerun over a FIFO scheduled network in order to have an idea about the advantages of using the scheduling discipline as a network support for congestion control in the Internet.