I am trying to model queuing delays in ethernet switches. Can anyone guide which queuing model can be used to model the queuing delays in a switch ?

  • 1
    Modern ethernet switches are wirespeed switches. Any buffers in switches are tiny, and oversubscription causes frames to get dropped.
    – Ron Maupin
    Feb 14, 2021 at 16:18
  • Did any answer help you? if so, you should accept the answer so that the question does not keep popping up forever, looking for an answer. Alternatively, you could post and accept your own answer.
    – Ron Maupin
    Dec 23, 2021 at 16:27

1 Answer 1


Most switches use multiple buffer queues for each port.

Each traffic class for QoS (frequently controlled by DiffServ) is mapped to one of the buffers. Buffers are then served in some priority pattern, e.g. 4:2:1 for a three-stage queue, aka weighted round robin.

Many switches allow you to configure the number of queues or even the priority pattern.

Minimum latency for non-congested forwarding is often provided in the device datasheet. Note that there's a distinction between last-bit-in-first-bit-out (LIFO) primarily for store-and-forward switching (with minimum frame size), and first-bit-in-first-bit-out (FIFO) for cut-through forwarding. When these are mixed, you can calculate the respetive counterpart by adding/subtracting the serialization delay for a minimum frame (72*8=576 bits).

As Ron has pointed out, buffering with common network speeds is necessarily shallow. A congesting 10 Gbit/s flow exhausts a fairly sized 12 MB (port-global!) buffer in 10 ms.

  • Thank you for your responses. I am trying to model the latency in switches that agregate ethernet traffic. For this i am running simulations in NetSim. I am observing that sources which transmit data at a higher speed or rate suffer lower queuing delays than the ones that transmitt lower data rates. Here i would mention that switches are operating in store and forward mode. Feb 16, 2021 at 4:57
  • 2
    You cannot simulate that in software. Switches use hardware forwarding at wirespeed. What you're seeing is an artifact of the simulation.
    – Zac67
    Feb 16, 2021 at 6:54
  • Aggregated links behave very much like single links if you know which flows share the same physical link. How traffic distribution works is a function of the egress switch which you might need to find out.
    – Zac67
    Feb 16, 2021 at 7:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.