Batch Throughput Efficiency of ADCCP/HDLC/SDLC Selective Reject Protocols

The batch throughput efficiency is studied for three variants on selective reject protocols that operate in a full duplex asynchronous response mode and that adhere to the common architectural features of ADCCP, HDLC, and SDLC. In these architectures, the receiver of information frames may use the REJ supervisory frame to request the sender of information frames to back up to an earlier point in the send sequence. Alternatively, the receiver of information frames may issue the SREJ supervisory frame, which requests resending of a single previous information frame followed by continuation of the ongoing send sequence. The first protocol discussed here uses SREJ as the principle method for recovering lost frames but reverts to REJ recovery if certain combinations of errors occur. The second protocol uses SREJ as the only means for recovery. The third protocol uses SREJ for "isolated" frame losses; if a second frame loss occurs within a specified time after the first, then recovery of the first loss is done by REJ. Results plotted for transmission via satellite at 1.544 Mbits/s show that the selective reject protocols permitted by the current versions of ADCCP, HDLC, and SDLC outperform the simpler REJ protocol only for bit error rates in a narrow range. Thus, it may be worthwhile to consider modifications to these architectures that will permit more flexible selective reject protocols.