The performances of the go-back-N ARQ scheme with selective repeat in intra-block (GBN-SR) and revised GBN-SR (Rev-GBN-SR) schemes are analyzed. Using the derived generating function of the service time of a packet, the throughputs of the proposed schemes are compared with that of Benelli's (1986) scheme. It is shown that the proposed Rev-GBN-SR scheme shows good performance in a wider range from low to high packet error rate regions.

In this paper, we determine an exact formula for the throughput performance of the GBN SR scheme, in which the go-back-N ARQ scheme is used for inter-block transmission and the selective repeat scheme is used for intra-block transmission. Using that formula, the throughput performance of GBN SR scheme is compared with that of Benelli's (1986, 1993) scheme in a numerical manner. As a result, it is shown that the throughput performance of the GBN SR scheme is superior to that of Benelli's scheme in a wider range of packet error rate from low to medium, but is inferior at high packet error rates.

The DQDB MAC Protocol standardized by the IEEE 802.6 Committee is the single segment bandwidth reservation scheme that reserves the bandwidth for only one segment in the distributed queue. Recently, the multi-segment bandwidth reservation schemes that reserve the bandwidth for not only one segment in the distributed queue but also a part of or all segments in the local node queue have been proposed. In this paper, we propose a new multi-segment bandwidth reservation protocol that can quickly react to changes of node's traffic and can quickly allocate bandwidth fairly and waste-free. We also evaluate the mean message transmission delay and the throughput convergence performance by simulation. As a result, the mean message transmission delay can be decreased and the throughput can be quickly converged to the fair bandwidth allocation.