“…Thus, with a negligible workload, the optimal number of IS processes per system is 1, since this eliminates the need of a synchronisation (intra-system) message (i.e., that carrying the sequence number), as it can be seen in expression (3). However, when workload is considered, we could use a queueing model [17] in order to represent such system, but as it is already mentioned in [17,Chapter 5], in a queueing network we only need to identify its bottleneck centre in order to set upper bounds to the global system throughput. In our system, and with the assumptions given above, such bottleneck centre is the interconnecting channel.…”