Streaming transmission of poisson traffic

We study the potential benefits of stream codes for delay-constrained transmission of random arrivals of data. Stream codes operate in a continuous manner by allowing new data to be encoded and decoded at any time. We develop a random-coding formulation for streaming that generalizes the traditional block-coding model, and obtain exponentially tight expressions for optimal decoding error as a function of decoding delay. For a single-user channel with Poisson arrivals of fixed- length packets, we obtain bounds on the end-to-end delay of rate-adaptive stream codes and show that simple greedy rate control policies achieve nearly optimal performance.