Abstract: An input-buffered multipoint switch having input channels and output channels includes multilevel request buffers, a data path multiplexer, and a scheduler. The switch has a distinct multilevel request buffer associated with each input channel and each request buffer has multiple request registers of a different request buffer priority. The request registers store data cell transfer requests that have been assigned quality of service (QoS) priorities, where the QoS priorities are related to packet source, destination, and/or application type. The multilevel request registers are linked in parallel to the scheduler to allow arbitration among requests of different input channels and different request buffer priority levels. The preferred arbitration process involves generating QoS priority-specific masks that reflect the output channels required by higher QoS priority requests and arbitrating among requests of the same QoS priority in QoS priority-specific multilevel schedulers.

Abstract: A method and apparatus for supplying new requests to a scheduler in an input-buffered multiport switch involve selecting a request that does not target output channels that conflict with output channels targeted by requests that are already accessible to the scheduler. Specifically, target output channels of requests that are presently accessible to the scheduler are identified and compared to target output channels of requests that are included in a queue of next-in-line requests. The queue of next-in-line requests is reviewed and the highest priority request having no conflicting output channels is supplied to the scheduler. By supplying the scheduler with a new request that targets non-conflicting output channels, the scheduler is presented with a wider range of requested output channels from which to choose in each arbitration cycle. In a first embodiment, one, two, or eight ports are connected to each one of four input/output controllers in a switch having a four-channel switch fabric.

Abstract: A method and apparatus for forwarding packets from contending queues of a multiport switch to an output of a finite bandwidth involve first prioritizing the contending queues into different priorities that relate to priorities of the packets that are being forwarded in the network. Bandwidth of the output is then allocated among the prioritized contending queues and the bandwidth of the output is consumed by the queued packets according to the allocated proportions. Any unconsumed bandwidth is distributed to the queues on a priority basis such that the highest priority queue is offered the unconsumed bandwidth first and lower priority queues are offered the remaining unconsumed bandwidth in priority order. An advantage of the invention is that queues are not starved of bandwidth by higher priority queues and unconsumed bandwidth is not wasted when there are not enough packets to consume an allocated portion of the output bandwidth.

Abstract: A method and apparatus for forwarding variable-length packets between channel-specific packet processor in a crossbar of a multiport switch involve segmenting variable-length packets into fixed-length payload segments and multiplexing the payload segments with response or request data to form fixed-length switching blocks. The fixed-length switching blocks are transferred to and from the crossbar over respective input and output connections in order to minimize the number of connections between the packet processors and the crossbar. The input and output connections enable the transfer of a current packet through the crossbar while supplying the crossbar with request information necessary to schedule subsequent packets through the crossbar. In a preferred embodiment of the invention, packets are timed to pass through the crossbar one after another in order to utilize the maximum bandwidth of the crossbar.

Abstract: A method and apparatus for controlling the flow of variable-length packets to a multiport switch involve accessing forwarding information in a memory based at least partially on layer 4 information from a packet and then forwarding the packet only if the packet is within a bandwidth consumption limit that is specified in the forwarding information. In a preferred embodiment, a credit bucket algorithm is used to ensure that packet flows are within specified bandwidth consumption limits. The preferred method for implementing the credit bucket algorithm to control flows of packets involves first receiving a particular packet from a flow and then stripping the layer 2 header information from the packet. The layer 3 and layer 4 information from the packet is then used to look-up flow-specific forwarding and flow control information in a memory that stores a linked list of table entries that includes the fields necessary to implement the credit bucket algorithm.

Abstract: An input-buffered multipoint switch having input channels and output channels includes multi-level request buffers, a data path multiplexer, and a scheduler. The switch has a distinct multi-level request buffer associated with each input channel and each request buffer has multiple request registers for storing data cell transfer requests of different priorities. The multi-level request registers are linked in parallel to the scheduler to allow arbitration among requests of different input channels and different priority levels. The preferred arbitration process involves generating masks that reflect the output channels required by the same priority level requests. Utilizing masks to arbitrate between multiple requests in an input-buffered switch reduces arbitration cycle time and minimizes HOL blocking.

Abstract: A method and apparatus for routing packets through a multiport switch involves attaching indicators to packets before the packets are passed through a switch fabric, and then using the indicators to look-up output port vectors after the packets have been passed through the switch fabric. In a preferred embodiment of a 64.times.64 port switch, an 8-bit output channel vector and a 15-bit output port index are attached to a packet and passed through the switch fabric. The 8-bit output channel vector directs the packet to the proper output packet processor(s) and the 15-bit output port index is used to look-up an output port vector that identifies the output port(s) that will receive the packet. The method and system provide low packet overhead and flexible scaling.