The thesis investigates the reliable transport from one sender to a group of receivers (multicast). It is often stated that network support is required to allow for efficient reliable multicast. This thesis shows that this is not the case. Efficient reliable multicast is possible on a pure end-to-end basis even for a very large number of receivers. In the case of data loss in the network, reliability is efficiently provided, if receivers signal the loss to the sender and the sender in turn performs a retransmission of the missing data. The complexity of such a system increases with the number of receivers. Both, the number of retransmissions to be performed by the sender and the amount of feedback returned to the sender increases with the number of receivers. The thesis makes three major contributions to the field of reliable multicast: For the first time the scalability of protocol mechanisms is analyzed for groups of size 1 up to 1 million receivers. For loss recovery a single parity packet can repair different losses at different receivers. Therefore, loss recovery by retransmission of parities outperforms retransmitting originals. The comparison of loss recovery by parity and loss recovery by originals is investigated for burst loss, for correlated loss among receivers and for a limited processing capability at sender and receivers. Feedback from very large groups is dangerous, since the sender may get overwhelmed by feedback messages. The thesis contains an end to end solution for feedback from an unknown number of receivers. The feedback method allows to control the average number of feedback messages and to estimate the number of receivers. It is shown that this feedback method results in faster feedback than provided by other end-to-end solutions. Finally, the proposed multicast protocol mechanisms are compared to protocol mechanisms that have support from the network.
Reliable multicast transport to large groups
© EPFL. Personal use of this material is permitted. The definitive version of this paper was published in Thesis and is available at : http://dx.doi.org/10.5075/epfl-thesis-1832
PERMALINK : https://www.eurecom.fr/publication/896