The Analysis of Some Selective-Repeat ARQ Schemes with Finite Receiver Buffer

In high bit rate data transmission systems with ARQ error control, the throughput efficiency is a function of bit error rate, block or packet size, and the effect of significant round trip delays such as may be experienced in satellite communication systems. The selective-repeat ARQ scheme is capable of providing superior throughput performance independent of round trip delay, but requires excessively large receiver buffers; as a result the inferior GoBack N procedure is commonly adopted. This paper analyzes a class of mixed-mode ARQ protocol models which incorporate a selectiverepeat mode with finite receiver buffer. The protocol models are shown to be amenable to exact throughput analysis, but do assume that the round trip delay is constant and known, blocks are of constant length, and the ACK/NAK signals are returned error free. These assumptions might create difficulties for practical implementation. However, the analytical model results highlight those aspects of ARQ protocols which affect throughput performance as round trip delays increase. The results show that it is desirable for best throughput performance in practical systems that at least the first retransmission of a block following an error should be in the selective-repeat mode to obtain superior performance over GoBack N schemes. Furthermore, alternative secondary retransmission modes are considered which ensure that reliable transmission can be achieved without receiver buffer overflow, even if the selective-repeat mode retransmissions fail. It is shown that the choice of secondary mode does not have a significant effect on the throughput efficiency but has a bearing on complexity.