QoS Mechanism Triggers

Quality of Service mechanisms, and in particular, queueing mechanisms, that are applied on Cisco devices will trigger or "kick in" depending upon several factors.

Queuing

Cisco Routers

For Cisco routers, any configured software queueing will only activate if there is congestion on the particular interface on which queuing is enabled. The interface has a hardware queue. When that queue is full, we'll use the software queue. When traffic arrives at a port at a rate less than the port's speed, the traffic is served immediately and never queued. The software queue won't be used.

Once the arriving traffic exceeds the speed of the port, queuing occurs.

Cisco Switches

For switches, the behavior becomes somewhat more complicated. Depending on the platform and the operating system you may see various behaviors. There are cases where even if you haven't configured QoS at all, behind the scenes, the switch allocates a certain threshold of its buffers and queue thresholds to unmarked traffic, which will affect throughput rates even if the maximum interface throughput has not been reached.

This is highly platform dependent, and you should examine this closely before configuring anything. You can read more about it here:

QoS Hardware Queues vs Software Queues QoS Hardware Queues Scheduling

WRED

Mechanisms such as WRED for example, are called congestion avoidance mechanisms, and this term can become confusing. However, strictly speaking, WRED is a queue exhaustion avoidance mechanism. More on that can be found at QoS - is WRED really a congestion avoidance mechanism.

Policing and Shaping

Policing and shaping are applied whether there is congestion or not. When you hit configured limits, the traffic will be shaped or policed.

https://networklessons.com/quality-of-service/introduction-qos-quality-service

https://networklessons.com/cisco/ccie-routing-switching-written/wred-weighted-random-early-detection