Method for active queue management with asymmetric congestion control
View Patent ↗A scalable method for efficient dynamic allocation of buffer resources in a store-and-forward device, such that high utilization can be maintained with small average buffer occupancy by providing asymmetric congestion control with opportune random detection. Also provided is tolerance of transient onset of congestion and fairness toward bursty traffic with ready reaction to declines in congestion.
1. A method in a store and forward device for actively managing a data queue in a buffer having one or more data packets, the method comprising the steps of:
assigning the data queue an initial state;
discarding an incoming data packet if the data queue is currently full;
determining a level of congestion currently experienced by the data queue if the data queue is not full;
reassigning the data queue to one of a plurality of states based on the determined level of congestion and the current state of the data queue, wherein the plurality of states comprise a series of successive states that are each associated with a different probability of dropping an incoming packet that is higher than that of a previous state and that are not derived from a length of the data queue;
deciding whether to discard the incoming data packet based on the current state of the data queue;
reassigning the data queue to one of a plurality of states based on the determined level of congestion and the current state of the data queue if the incoming data packet is discarded; and
upon arrival of another data packet, recycling back to determining a level of congestion currently experienced by the data queue.
2. The method of claim 1 wherein the initial state is designated a non-dropping state wherein the incoming data packet is allowed to enter the data queue without dropping if the data queue is currently in the initial state.
3. The method of claim 1 wherein the incoming data packet is discarded probabilistically in accordance with the probability of dropping an incoming packet associated with the current state of the data queue, unless the current state of the data queue is the initial state.
4. The method of claim 3 wherein the data queue is reassigned to the initial state if the incoming data packet is discarded.
5. The method of claim 1 wherein the step of determining a level of congestion in the data queue is based on a measure of instantaneous queue length.
6. The method of claim 4 wherein the step of reassigning the data queue to one of a plurality of states based on the determined level of congestion and the current state of the data queue further comprising the steps of:
assigning a probability to each possible transition from the current state to a next state based on the validity of a predetermined congestion condition; and
making a transition from the current state to the next state probabilistically in accordance with the assigned probability.
7. A method in a store and forward device for actively managing a data queue in a buffer having one or more data packets, the method comprising:
assigning the data queue an initial state;
discarding an incoming data packet if the data queue is currently full;
determining a level of congestion currently experienced by the data queue if the data queue is not full;
probabilistically reassigning the data queue to one of a plurality of states based on the determined level of congestion, wherein the plurality of states comprise a series of successive states associated with a probability of marking an incoming packet higher than that of a previous state;
deciding whether to mark the incoming data packet based on the current state of the data queue, wherein the marking of the incoming data packet is used for explicit congestion notification to the source of the data packet;
reassigning the data queue from the current state to the initial state if the incoming data packet is marked or discarded; and
upon arrival of another data packet, recycling back to determining a level of congestion currently experienced by the data queue.
8. The method of claim 7 wherein the initial state is designated a non-marking state wherein the incoming data packet is allowed to enter the data queue without marking if the data queue is currently in the initial state.
9. The method of claim 8 wherein the incoming data packet is marked probabilistically in accordance with the probability of marking an incoming packet associated with the current state of the data queue, unless the current state of the data queue is the initial state.
10. The method of claim 7 wherein the step of determining a level of congestion in the data queue is based on a measure of instantaneous queue length.
11. The method of claim 10 wherein the step of probabilistically reassigning the data queue to one of a plurality of states based on the determined level of congestion further comprises the steps of:
assigning a probability to each possible transition from the current state to a next state based on the validity of a predetermined congestion condition; and
making a transition from the current state to the next state probabilistically in accordance with the assigned probability.