Performance of a shuffle exchange network using Markov chains

Summary form only given. This paper presents an approximation of a performance evaluation of a single stage shuffle exchange network (SEN) topology. The proposed model is based on a homogenous and finite Markov chain (MC), where the state is discrete and time is continuous. The unbuffered and the buffered strategies of SEN have been considered. We show that the unbuffered strategy has a low blocking probabilities compared to the buffered one. We show that there is no linearity, on one hand, between the bandwidth (Bw) or the average service time (Avt) and the size of the network, and on the other hand, between the Bw or the Avt and the length of the message.