| 摘要 |
A N input and N output, scalable switching network (80) is implemented with an array of log base x of N rows and N/x columns of routers (40-68); where each router has 2x ports and where the rows of routers are connected to each other via a blocking compensated cyclic group mapping interconnect (49 and 59). A N input and N output, bi-directional scalable switching network (90) is implemented with an array of log base x of (N/2) rows and (N/2)/x columns of routers (40-68); where each router has 2x ports, and where the rows of routers are connected to each other via a blocking compensated cyclic group mapping interconnect (49 and 59). The routers provide destination routing. The levels of blocking compensated cyclic group mapping interconnect provide connectivity between the rows of routers. The internal connectivity of the scalable switching network (80) or the bi-directional scalable switching network (90) can be enhanced (82 and 92) by inserting at least one additional level of mapping interconnect (69) and row of routers (70-78). The external connectivity can be enhanced by providing alternate routes to (128) and from (124) various destinations. Fault tolerance (116) can be achieved by using dynamic routing to take advantage of the enhanced internal or external connectivity to route around faults. Input load balancing (124), internal load balancing (118), output load balancing (128), delay reduction (120), and jitter reduction (122) can be achieved by having the routers load balance between the equivalent routes provided by the enhanced internal or external connectivity. Shared distributed output buffering can be achieved by implementing flow control between the routers (115). Remote initialization, monitoring, and updating of the scalable switching network can be achieved by assigning each router (40-78) an IP address, and accessing the routers via a computer network such as the Internet.
|
