Performance Evaluation

After describing the network at length in the previous chapter, we investigate the impact of various network architecture and protocol parameters on the network performance by means of analysis and/or simulation in this chapter. For tractability reasons we assume Bernoulli packet arrival processes in our analyses. We note that Bernoulli traffic is not an appropriate traffic model for LANs and WANs, as shown in [LTWW94] and [PF95], respectively. But at the time of writing it is an open question which traffic model applies in MANs. While the simulations in this chapter make also use of Bernoulli traffic, in Chapter 8 we provide additional simulations which are driven by packet header trace files that were recorded at the metro level. The performance evaluation proceeds in two steps. First, we consider different network aspects separately from each other in order to provide insight in how they influence the network performance. In Section 6.1, we focus on switching of fixed-size packets and show the positive impact of using multiple FSRs of the underlying AWG on the throughput-delay performance of the network. Next, in Section 6.2 we consider variable-size packets and demonstrate the benefit of spatial wavelength reuse for improving the network flexibility and efficiency. Section 6.3 takes also multicasting into consideration. We show that our network is able to support multicasting efficiently and we examine the interplay between multicast and unicast packet switched traffic.