In the area of error control, there is the well-known NACK implosion problem, which happens when multiple clients simultaneously send negative acknowledgements to the source, causing source and network overload. Another problem lies in the recovery process when lost packets are retransmitted to the receivers. There is the question of who should perform the retransmissions? There is also the question of whether the retransmitted packets should be multicast to the entire group or just to a subset of receivers?
Flow/congestion control in reliable multicast is another difficult challenge. The key issue is: how does a sender adjust its sending rate to accommodate all receivers (with different bandwidth and buffer requirements) in the group? Does the sender transmit according to the rate of the slowest receivers? How can the sender handle flow control messages from multiple clients without being overloaded?
We intend to investigate these issues in the context of an Active Networks environment. In Active Networks, intermediate routers provide a programmable network service interface. Applications can inject program fragments in the form of capsules into these routers and perform application-specific processing. We intend to develop a scalable reliable multicast mechanism in which intermediate routers play an active role in providing error and flow control.