Extremum Feedback for Very Large Multicast Groups

In multicast communication, it is often required that feedback is received from a potentially very large group of responders while at the same time a feedback implosion needs to be prevented. To this end, a number of feedback control mechanisms have been proposed, which rely either on tree-based feedback aggregation or timer-based feedback suppression. Usually, these mechanisms assume that it is not necessary to discriminate between feedback from different receivers. However, for many applications this is not the case and feedback from receivers with certain response values is preferred (e.g., highest loss or largest delay). In this paper, we present modifications to timerbased feedback suppression mechanisms that introduce such a preference scheme to differentiate between receivers. The modifications preserve the desirable characteristic of reliably preventing a feedback implosion.