System and method for determining a cause of network congestion

A method to determine a cause of congestion in a network element, the method comprising:

• measuring a queue occupancy of a queue group for a port in the network element, wherein the queue group stores a plurality of packets to be communicated through that port and the measuring of the queue group occupancy is performed with a periodicity that is smaller than a poll drop periodicity used to poll drop counters;
• recording the queue group occupancy continuously based on the periodicity if the queue group occupancy is greater than a first threshold, wherein the queue group occupancy recording is turned off if the queue group occupancy is below the first threshold and the recording stores a plurality of queue group occupancy values from the continuous queue group occupancy recording;
• determining if the measurement indicates a potential congestion of the queue group by determining if the queue group occupancy is greater than a second threshold, wherein the second threshold is larger than the first threshold;
• and
• if the potential congestion exists on that queue group, gathering information regarding packets to be transmitted through that port.

A system that communicates network data between a network and a network element, the system comprising:

• the network;
• and
• the network element, coupled to the network, wherein the network element determines a cause of congestion in the network element, the network element including,
• a port, coupled to the network, wherein the port communicates a plurality of packets between the network and the network element,
• a queue group, coupled to the port, wherein the queue group stores the plurality of packets;
• a network congestion module, coupled to the queue group, wherein the network congestion module measures a queue occupancy of the queue group and the measuring of the queue group occupancy is performed with a periodicity that is smaller than a poll drop periodicity used to poll drop counters, records the queue group occupancy continuously based on the periodicity if the queue group occupancy is greater than a first threshold, determines if the measurement indicates a potential congestion of the queue group by determining if the queue group occupancy is greater than a second threshold, and if the potential congestion exists on that queue group, gathers information regarding packets to be transmitted through that port, wherein the second threshold is larger than the first threshold and the queue group occupancy recording is turned off if the queue group occupancy is below the first threshold and the recording stores a plurality of queue group occupancy values from the continuous queue group occupancy recording.

A network element to determine a cause of congestion in a network element, the network element comprising:

• a port to be coupled to the network, the port to communicate a plurality of packets between the network and network element;
• and
• a data plane to be coupled to the port, the data plane including,
• a queue group, coupled to the port, the queue group to store the plurality of packets;
• and
• a network congestion module, coupled to the queue group, the network congestion module to measure a queue occupancy of the queue group, record the queue group occupancy continuously based on a periodicity if the queue group occupancy is greater than a first threshold, determine if congestion exists on that queue group using the measurement by determining if the queue group occupancy is greater than a second threshold and the measuring of the queue group occupancy is performed with the periodicity that is smaller than a poll drop periodicity used to poll drop counters, wherein the congestion prevents a packet of the plurality of packets form being communicated within a time period, and if the congestion exists on that queue group, to further gather information regarding packets to be transmitted through that port, wherein the second threshold is larger than the first threshold and the queue group occupancy recording is turned off if the queue group occupancy is below the first threshold and the recording stores a plurality of queue group occupancy values from the continuous queue group occupancy recording.