In random linear network coding, rate control is an important strategy for limiting the decoding delay of a system. In broadcast systems where the channel rate is either unknown or varies over time, we demonstrate that a target delay can be achieved using the queue threshold scheme that we introduce. At throughputs approaching the channel rate, the queue threshold rate control scheme is shown to achieve improved throughput delay performance compared with existing schemes. We demonstrate that it is possible to modify this rate control scheme to greatly reduce the amount of feedback required, in exchange for a slight degradation of the throughput delay performance. Furthermore, this rate control scheme is shown to perform reasonably well, even under lossy and delayed feedback.

In this paper, we provide theoretical and simulation-based study of the delivery delay performance of a number of existing throughput-optimal coding schemes and use the results to design a new dynamic rate adaptation scheme that achieves improved overall throughput-delay performance. Under a baseline rate control scheme, the receivers' delay performance is examined. Based on their Markov states, the knowledge difference between the sender and receiver, three distinct methods for packet delivery are identified: zero state, leader state, and coefficient-based delivery. We provide analyses of each of these and show that, in many cases, zero state delivery alone presents a tractable approximation of the expected packet delivery behavior. Interestingly, while coefficient-based delivery has so far been treated as a secondary effect in the literature, we find that the choice of coefficients is extremely important in determining the delay, and a well-chosen encoding scheme can, in fact, contribute a significant improvement to the delivery delay. Based on our delivery delay model, we develop a dynamic rate adaptation scheme that uses performance prediction models to determine the sender transmission rate. Surprisingly, taking this approach leads us to the simple conclusion that the sender should regulate its addition rate based on the total number of undelivered packets stored at the receivers. We show that despite its simplicity, our proposed dynamic rate adaptation scheme results in noticeably improved throughput-delay performance over existing schemes in the literature.

In this paper we study in-order packet delivery delay of two recently proposed network coded transmission schemes with applications in wireless broadcast. Unlike previous works where asymptotic behaviour of decoding or delivery delay was presented, we provide a general analysis of the three conditions under which in-order packet delivery is possible at a receiver: by 1) catching up with the sender, 2) receiving while a leader, and 3) chance decoding. We use a Markov model to represent the difference between the knowledge space of the sender and a receiver. For the first condition, we calculate the expected distribution of decoding cycle lengths under the Markov model. For the second condition, we propose to use a simplifying independent Markov model among receivers to shed light on the factors that determine the probability of receiving while a leader. Finally, we compare the chance decoding probabilities of two transmission schemes and a baseline random transmission algorithm to show that surprisingly (and fortunately) the probability of chance decoding is significant in one of the transmission schemes. We verify our analysis by extensive simulations and discuss the usefulness of our study for understanding and design of better transmission algorithms.

We introduce a homogeneous r-regular network model for a peer-to-peer (P2P) video broadcast network. Such networks are simple to construct and allow the implementation of fairness strategies. We use our model to show why the greedy and rarest first push-based strategies give the suboptimal performance often observed in the literature. We propose a novel network coding based transmission strategy and prove that it results in optimal playback delay and throughput performance.