Fibre-Ribbon Pipeline Ring Network with Distributed Global Deadline Scheduling

This paper introduces a novel, fair medium access protocol for a pipelined optical ring network. The protocol provides global optimisation of deadline constraints on a packet basis. Requests for sending packets are sent by the nodes in the network to the current master node. The master uses deadline information in the requests to determine which request is most urgent. Arbitration is done in two steps, collection and distribution phases. The protocol is therefore called two-cycle medium access (TCMA). The network is best suited for LANs and SANs (system area networks) such as a high speed network in a cluster of computers or interconnection networks in embedded parallel computers. Services possible in this network include best effort messages, real-time virtual channels, functions used in parallel processing such as barriersynchronisation, and functions for reliable transmission. These are possible without additional higher level protocol layers. A simulation analysis of the network with the novel protocol is presented. Further analysis shows worst case latency, minimum slot length, and fairness of the protocol.