Token-based Switch Fabric Scheduling Algorithms Research

High performance routers are essential equipments in the communication networks. With the increasesing of valume of communication network services, the forwarding capability and scale is surging. At present, the carrier-class core routers, enterprise routers and other high-end large-scale routing equipment are mainly consisted by the physical layer interface chips, network processors, traffic managers, and switching modules. For example, traffic managers excute the functions of buffing data cells, shaping traffic, avoiding congestion. The switching modules switch the cells from the input ports to the output ports. The certain topology which interconnected by switching modules, input and output modules, and internal protocol constitutes the internal switching network in a router.The switching network is one of the core technologies of building routers, using different configurations, the switching network can scale from several Gbit/s to Tbit/s. With the enhancing of capacity, diversifing of services, improving the switching network's scheduling efficiency, resource utilization and quality of services are becoming central issue in the field of network research. The existing scheduling algorithms have room for improvement in supporting multi-service features and scheduling efficiency. For the two aspects, the main contribution of this article and innovation lies in designing two distributed token-based scheduling algorithm, CBTS (Credit-Based traffic state scheduling) and CBUM (Credit-Based Unicast and Multicast).This paper first introduces the technical background of switching network, researching status, the existing scheduling algorithm, its improved algorithms and their application results. Then take multi-stage multi-plane CLOS switching network as an example to build a common OPNET simulation platform, and implement the two scheduling algorithm based on the platform. For the two improved algorithm, and the general token based scheduling algorithm. Then we use the same service source model to validate their feasibility and effectiveness in muti-priority service environment. The simulation results show that the two algorithms can achieve the desired experimental effect in the business environment and improve the distributed token based packet scheduling algorithm's delay and multi-services supportting, with the practicality.