[信息与通信]科大智能电力EPON产品技术交流.ppt

* * * * * * Cross-bar When a packet is received on a line card in the cross-bar architecture router, the first check determines if the destination is reachable via an output interface on the same line card. If so, the packet is copied to the local packet memory and switched by the line card itself. If the destination is reachable on another line card, the packet is transferred via the cross-bar fabric to the destination line card, which in turn transmits it on its medium. The cross-bar interconnection can be treated as a private high-bandwidth point-to-point connection between line card pairs. 分布式 When a packet is received on a line card in the distributed architecture, the first check determines if the destination is reachable via an output interface on the same line card. If so, the packet is copied to the local packet memory and switched by the line card itself. If the destination is reachable on another line card, the packet is copied to the fast shared memory (usually implemented with SRAM). The central CPU switches the packet to the destination line card, which in turn transmits it on its medium. This procedure involves the shared bus and presents a performance penalty, as multiple shared resources are used. Therefore it is imperative to implement distributed architectures by analyzing traffic flows in the network, and try to configure the chassis for its maximum performance. 集中式 When a packet is received in a shared memory architecture router, the receiving interface copies the packet into shared memory and invokes the CPU to perform packet switching. The CPU reads the shared memory, makes a forwarding decision, and transfers the packet to the output interface queue. The output interface copies the packet from the shared memory and transmits it on its medium. It is important to remember that the central CPU always performs all switching decisions. In the basic shared memory architecture router, all switching is performed using three shar