Method And Apparatus For Managing Network Congestion

Abstract

A manner of managing congestion in a data-traffic network. In one embodiment a network node such as a bridge, switch, or router includes an AQM having a PI controller configured to calculate p′ using the difference between Q and a Target Q, wherein p′ is p 0.5 and p is the probability that a received packet will be dropped or marked, and some drop decision functions are configured to indicate that the node should drop a received packet by comparing p′ to two random values. A marking decision function may also be present and configured to indicate that the node should mark a received packet by comparing p′ to one random value. A congestion control classifier, which is in some embodiments an ECN classifier, is also present to classify a received packet and facilitate making the proper dropping or marking decision.

1 . A method of data traffic congestion management in a network node, comprising:
 receiving data packets;   enqueuing at least a portion of the received packets;   determining whether to apply a drop decision to a received packet;   determining, if applying a drop decision, whether to drop the received packet applying a squared drop decision, wherein the squared drop decision comprises generating at least one random value and making the drop determination when a drop probability is larger than the at least one generated random value and a second random value;   dropping a received packet for which a drop determination has been made; and   forwarding from the network node, packets that have been received and not dropped.   
     
     
         2 . The congestion-management method of  claim 1 , wherein one of the random values is generated for each received packet and the other random value is a random value applied to the previous received packet. 
     
     
         3 . The congestion-management method of  claim 1 , wherein a received packet is dropped prior to enqueuing the packet. 
     
     
         4 . The congestion-management method of  claim 1 , further comprising:
 measuring load of queue buffer to extract at least one queue parameter Q;   providing the at least at least one queue parameter to an AQM (active queue manager); and   calculating a mark or drop probability as a proportional and integral control function on the Q parameter and a target Q parameter.   
     
     
         5 . The congestion-management method of  claim 4 , wherein the at least one queue parameter comprises a plurality of queue parameters. 
     
     
         6 . The congestion-management method of  claim 4 , wherein the drop or mark probability is also calculated using at least one of gain parameter α and gain parameter β. 
     
     
         7 . The congestion-management method of  claim 6 , wherein at least one of α and β remain constant for a given traffic flow. 
     
     
         8 . The congestion-management method of  claim 4 , further comprising deciding, if not applying a drop decision, whether to mark a received packet using the calculated mark or drop probability. 
     
     
         9 . The congestion-management method of  claim 8 , wherein the marking decision comprises generating at least one random value and making the mark determination when the mark probability is greater than the at least one random value. 
     
     
         10 . The congestion-management method of  claim 9 , wherein a new random value is generated for each received packet for which a marking decision is to be made. 
     
     
         11 . The congestion-management method of  claim 9 , further comprising marking a packet for which a determination to mark has been made. 
     
     
         12 . The congestion-management method of  claim 1 , wherein the determination of whether to drop a packet is a function of the type of flow in which the packet is received. 
     
     
         13 . A non-transitory computer-readable storage medium that stores computer readable instructions, which when executed by at least one processor implement a method for congestion management, the method comprising:
 receiving data packets;   enqueuing at least a portion of received packets;   determining whether to apply a drop decision to a received packet;   determining, if applying a drop decision, whether to drop the received packet applying a squared drop decision, wherein the squared drop decision comprises generating at least one random value and making the drop determination when a drop probability is larger than the at least one generated random value and a second random value;   dropping a received packet for which a drop determination has been made; and   forwarding from the network node, packets that have been received and not dropped.   
     
     
         14 . The computer-readable storage medium of  claim 13 , the implemented method further comprising:
 measuring load of queue buffer to extract at least one queue parameter Q;   providing the at least at least one queue parameter to an AQM; and   calculating a mark or drop probability as a proportional and integral control function on the Q parameter and a target Q parameter.   
     
     
         15 . The computer-readable storage medium of  claim 14 , wherein the squared drop decision is applied using a drop probability that is proportional with the square of the calculated mark or drop probability.