Research And Implementation Of Routing Mechanism For ForCES Router

With the growth of Internet, the scale of network is increasing rapidly. On the one hand, more and more functions and services are added in routers. One the other hand, the demands of high-performance processors are increasing sharply. Although the control and forwarding elements have been separated in conventional routers, the protocols used between the two are private, not open and hinder the interconnection among devices from different companies and lack of flexibility. All of these lead to the advent of ForCES(forwarding and control element separation).ForCES implements the interconnection among various devices from different companies and is considered as an important research. The IETF (Internet Engineering Task Force) has set up a ForCES working group devoting to the study of separation between forwarding and control elements. Present research releases that ForCES router has to support at least hundreds of FEs, so routing inside a router needs to be researched and resolved.This paper proposes a label-based routing mechanism and forwarding-table- update algorithm on the basis of topology discovery in order to solve the routing problem within a router. First, CE collects the topology information of FEs, calculates the shortest path, downloads the forwarding tables to all FEs with forwarding-table-update algorithm which comes from the Dijkstra algorithm added by forwarding-table-update operations. Then ingress FEs encapsulate the packets by labels and the entire forwarding table. After that internal FEs forward the packets hop by hop and eventually packets are decapsulated and forwarded outside in the exgress FEs.Backup paths are added to improve former mechanism to avoid black holes and errors, reduce the recovery time. The mechanism presented in this paper features shortest path and fast label forwarding, supports reliable transmission and dynamic topology changes and has good robustness.Experiments indicate that calculation cost of forwarding-table-update algorithm is better than that of OSPF and this algorithm has better capability and extensibility. If n denotes the number of FEs and e means the number of ingress FEs, the calculation cost reduces from O(n3) to O(mn2) and communication cost decreases from level of ne to level of n. In addition, former solution is optimized by the introduction of backup paths. As a result, the robustness of system is increased, repeated routing calculation is avoided and the recovery time is also decreased. Furthermore, time complexity is better than O2 and NSP.