Assessing the performance of packet retransmission schemes over satellite links

Satellite links are often characterized by scarce bandwidth, long delays, and fading. Next-generation, packet-switched satellite networks will increasingly use packet retransmission strategies because they can greatly improve the performance of packets sent over satellite links when errors are bursty due to fading. It is not clear, however, which packet retransmission scheme, if any, applies for a given application traffic type and link characteristics. This paper presents a methodology for assessing the performance of various packet retransmission schemes over satellite links that may include channel fading, turbo-coded forward error-correction (FEC) coding, and channel interleaving. First, a combined set of bit- and packet-level simulations must be run to quantify packet performance (i.e., error-rate, utilization, delay, and jitter) for a given link and packet retransmission scheme. Then, a set of trade-space plots can be generated to directly compare different packet retransmission strategies and determine which is optimal for a particular application traffic type and link characteristics. In order to illustrate and validate the approach, a study is conducted that simulates how three basic types of packet retransmission strategies (fixed-retransmission, go-back-N, selective-repeat (Lin, 1983)) perform over a satellite link that has Rayleigh fading channel statistics (with a short-decorrelation time) and a serial concatenated turbo-coder (with no channel interleaver). Results show that Selective-Repeat is optimal for non-real-time applications such as file transfer, while fixed-retransmission is optimal for realtime, streaming applications such as voice