Research On Large-scale Routes Management In Cluster Router

The human society is increasingly inseparable from the Internet. Routersplay a important role in building the Internet. With the development of Internetand the changes in network operation requirements, the cluster router architecturegains much attention in the next generation networks. Currently, lots of researchesextensively study the cluster router and address the scalability issue.However, with the development of the Internet, core routers need to providemore capacity for routing forwarding entries. In particular, a cluster router has lotsof line cards. Thus, the large-scale routing system management meet someproblems. This paper address the key issues: routing table distribution,decomposition of forwarding table, and multicast forwarding model, whichimprove the performance of cluster routers and the efficiency of large-scalerouting system management.This thesis systematically analyzes existing route distribution methods andthe cluster router architecture, and then proposes a Tree-based Parallel RouteDistribution (TPRD) model. In TPRD, all cards of cluster router are constructed toan unbalanced distribution tree and routes are transferred from the root node toleaf nodes. This thesis presents the design and algorithms to implement TPRD.The simulation results demonstrate that TPRD achieves an expected performance.We propose a novel approach called DSF to decompose FIB, which canreduce the unnecessary redundancy of FIB storage across multiple line cards.Moreover, an enhanced mechanism: EDSF is proposed to further minimize thediversity of the storage of line-cards. Storage decomposition model can alleviatethe pressure from the fast increasing size of FIB can save lots of hardwareresources.By means of optimizing the processing of data layer and control layer, wedesign optimized storage model of multicast forwarding table. Meanwhile, wedesign two-stage forwarding model of multicast. The proposed models reduce theforwarding delay and bandwidth consumption with a small amount of storagecosts and extra processing overhead. Thus, the performance of multicastforwarding is effectively improved.We design a supporting platform for cluster router, and construct a prototypeof cluster router using router BitEngine12K. Based on the prototype, we conduct experimental study to demonstrate the results of this thesis.