Abstract: A technique for managing a customer's traffic in a network node, such as a service provider edge device, involves dedicating a group of queues in the network node to the customer, performing queue-specific rate shaping on the customer's traffic according to queue-specific bandwidth limitations, and performing group-specific rate shaping on the customer's traffic as a whole according to a group-specific bandwidth limitation. In an embodiment, the queues in the group are associated with different types of customer traffic in order to provide type-specific rate shaping. Further, the queues may be prioritized among each other such that unused excess bandwidth is distributed among the different traffic types in priority order.

Abstract: A technique for managing the utilization of processing resources involves filtering packets that are sent to a CPU for learning before allowing the packets to reach the CPU. The filtering involves determining if related packets have already been allowed to reach the CPU for learning and using the knowledge about related packets to determine if a current packet should be allowed to reach the CPU. In one embodiment, the processing resources of the CPU are conserved by allowing only one packet per flow to reach the CPU for learning. The one packet is used by the CPU to generate the necessary forwarding information and to initiate programming of the hardware-based forwarding table so that subsequent packets of the same flow can be forwarded directly from the hardware-based forwarding engine.

Abstract: A method and system for controlling the flow of packets within a network node is disclosed. The method comprises receiving packets that are intended for a queue, classifying each of packets into at least one of a plurality of traffic classes and applying class-specific random early detection to each of said packets. The system comprises a classification engine configured to classify packets into one of a plurality of traffic classes, class-specific random early detection (RED) logic configured to apply class-specific RED to each of said packets that have been classified by said classification engine; and a queue for queuing any of said packets that are not dropped in response to application of said class-specific RED.

Abstract: A packet-based traffic forwarding system with a plurality of contending queues. Each queue has an allocated bandwidth and an associated borrow vector. When a first queue does not consume all of its allocated bandwidth, the borrow vector associated with a second queue is checked to determine whether the second queue is permitted to use the unconsumed allocated bandwidth of the first queue. If the second queue has packets to forward, the second queue forwards the packets using the unconsumed allocated bandwidth of the first queue.

Abstract: A packet-based traffic shaping system including a scheduler that forwards packets in accordance with a scheduling algorithm from queues that have at least a minimum good credit rating. The minimum good credit rating is not dependent on the size of a packet that is to be forwarded. Rate shapers associated with the queues reduce credit by transmission costs associated with the packets and restore credit incrementally over time. When a queue is empty, credit does not accumulate above an idle credit value at the associated rate shaper. When a packet is enqueued in a queue that has been idle for a period of time, the packet is forwarded without undue delay.

Abstract: Performing atomic lookup rule set transitions. A method involves performing lookups using a first set of rules, flagging a subset of the first set of rules as old rules, adding a second set of rules while continuing to perform lookups using the first set of rules, and atomically transitioning to perform lookups using the first set of rules plus the second set of rules minus the old rules. A system implementing the method includes a lookup table and a lookup table management module that is configured to make atomic lookup rule set transitions in the lookup table.

Abstract: A technique for managing the utilization of processing resources involves filtering packets that are sent to a CPU for learning before allowing the packets to reach the CPU. The filtering involves determining if related packets have already been allowed to reach the CPU for learning and using the knowledge about related packets to determine if a current packet should be allowed to reach the CPU. In one embodiment, the processing resources of the CPU are conserved by allowing only one packet per flow to reach the CPU for learning. The one packet is used by the CPU to generate the necessary forwarding information and to initiate programming of the hardware-based forwarding table so that subsequent packets of the same flow can be forwarded directly from the hardware-based forwarding engine.

Abstract: A technique for managing a customer's traffic in a network node, such as a service provider edge device, involves dedicating a group of queues in the network node to the customer, performing queue-specific rate shaping on the customer's traffic according to queue-specific bandwidth limitations, and performing group-specific rate shaping on the customer's traffic as a whole according to a group-specific bandwidth limitation. In an embodiment, the queues in the group are associated with different types of customer traffic in order to provide type-specific rate shaping. Further, the queues may be prioritized among each other such that unused excess bandwidth is distributed among the different traffic types in priority order.

Abstract: The flow of packet-based traffic is controlled to meet a desired rate by calculating, as a moving average, a current rate of packet-based traffic on a link, calculating the sum of the error between the calculated current rate and the desired rate, and determining whether or not packets can flow in response to the calculated sum of the error. When the sum of the error between the current rate and the desired rate indicates that the current rate is less than the desired rate, packets are allowed to flow and when the sum of the error indicates that the current rate is greater than the desired rate, packets are not allowed to flow. The magnitude of bursts can also be controlled by artificially controlling the minimum values of the current rate and the sum of the error. The flow control algorithm can be used for rate shaping or rate limiting.

Abstract: A method and system for controlling the flow of packets within a network node is disclosed. The method comprises receiving packets that are intended for a queue, classifying each of packets into at least one of a plurality of traffic classes and applying class-specific random early detection to each of said packets. The system comprises a classification engine configured to classify packets into one of a plurality of traffic classes, class-specific random early detection (RED) logic configured to apply class-specific RED to each of said packets that have been classified by said classification engine; and a queue for queuing any of said packets that are not dropped in response to application of said class-specific RED.

Abstract: A packet-based traffic shaping system including a scheduler that forwards packets in accordance with a scheduling algorithm from queues that have at least a minimum good credit rating. The minimum good credit rating is not dependent on the size of a packet that is to be forwarded. Rate shapers associated with the queues reduce credit by transmission costs associated with the packets and restore credit incrementally over time. When a queue is empty, credit does not accumulate above an idle credit value at the associated rate shaper. When a packet is enqueued in a queue that has been idle for a period of time, the packet is forwarded without undue delay.

Abstract: Performing atomic lookup rule set transitions. A method involves performing lookups using a first set of rules, flagging a subset of the first set of rules as old rules, adding a second set of rules while continuing to perform lookups using the first set of rules, and atomically transitioning to perform lookups using the first set of rules plus the second set of rules minus the old rules. A system implementing the method includes a lookup table and a lookup table management module that is configured to make atomic lookup rule set transitions in the lookup table.

Abstract: A packet-based traffic forwarding system with a plurality of contending queues. Each queue has an allocated bandwidth and an associated borrow vector. When a first queue does not consume all of its allocated bandwidth, the borrow vector associated with a second queue is checked to determine whether the second queue is permitted to use the unconsumed allocated bandwidth of the first queue. If the second queue has packets to forward, the second queue forwards the packets using the unconsumed allocated bandwidth of the first queue.

Abstract: The flow of packet-based traffic is controlled to meet a desired rate by calculating, as a moving average, a current rate of packet-based traffic on a link, calculating the sum of the error between the calculated current rate and the desired rate, and determining whether or not packets can flow in response to the calculated sum of the error. When the sum of the error between the current rate and the desired rate indicates that the current rate is less than the desired rate, packets are allowed to flow and when the sum of the error indicates that the current rate is greater than the desired rate, packets are not allowed to flow. The magnitude of bursts can also be controlled by artificially controlling the minimum values of the current rate and the sum of the error. The flow control algorithm can be used for rate shaping or rate limiting.