actual queue depth – the number of packets in the queue in a given moment
average queue depth – calculated based on the actual queue depth and the previous average. WRED uses this value to slowly adjust to what may be quick queue depth changes
class-based weighted fair queuing – reserves a minimum amount of pipe for each queue, as defined by a class
custom queuing – reserves a minimum amount of pipe for each queue
differentiated tail drop – in a LAN switch, this is where some queues discard frames more quickly than others
DSCP-to-CoS map – used by a switch to classify a frame and put it in the appropriate queue
DSCP-to-threshold map – used by a switch to classify a frame in regards to a WRED threshold
expedite queue – in a LAN switch, this is the queue that gets strict priority scheduling
exponential weighting constant – used by WRED to calculate the rate at which the average queue depth changes when compared with the current queue depth. The larger this constan, the slower the change in the average queue depth
finish time – used by WFQ. This is the number assigned to the packet when it goes into a WFQ queue. Lowest FT packet goes next.
full drop – the WRED equivalent of tail-drop
hardware queue – small FIFO queue on each router’s physical interface. Saves a CPU interrupt for every packet.
internal DSCP – in the context of LAN switching queuing frames, this may or may not be the DSCP value of the packet inside the frame
low-latency queuing – reserves a minimum of bandwidth for some queues, and high-priority for others; uses policing to avoid queue starvation.
mark probability denominator – WRED uses this value to determine the max percentage of packets to discard, while the average queue depth is in between min/max thresholds
maximum reserved bandwidth – defaults to 75%; this is configurable between 1 and 99 as a value defining how much of the pipe can be reserved for queues
maximum threshold – when average queue depth exceeds this value, WRED performs a full-drop.
minimum threshold – when average queue depth is below this value, WRED doesn’t drop any packets.
modified tail drop – how WFQ drops packets, similarly to tail-drop
no drop – what WRED calls not dropping any packets (People, people! Does EVERYTHING have to have a term associated with it?!?)
priority queue – a queue that receives better service than the other queues
priority queuing – what happens when the high priority queue is getting serviced to the exclusion of all others
queue starvation – when a queue does not get serviced because a higher priority queue has all the attention of the scheduler.
remaining bandwidth – bandwidth eligible for queuing that isn’t reserved as a “priority” queue.
scheduler – how a queuing methodology determines who the next packet is to get de-queued. Like being de-loused, only with electricity.
sequence number – in WFQ, this number is assigned to a packet as it gets queued up. The packet with the lowest SN in all the queues WFQ is managing will get forwarded next.
software queue – not a hardware queue, but rather one created within IOS. Because you’re smart, and you care.
strict priority – when a scheduler will always service a particular queue, assuming that queue has packets in it.
tail drop – when a packet shows up and needs to be queued, but the queue is full, thus the packet discarded
weighted fair queuing – when packets are automatically classified by flow and queued
weighted round-robin – where queues are serviced in order, but weighted to determine just how much service they get when it’s their turn