1

I have been working a bit with POE powered devices and POE switches and I wish that the behavior of POE switches when it comes to POE-LLDP was more consistent.

For example, there is one type of switch (HP 1910 for example) that works like this:

  1. POE-PD device asks for 25.5W from a 25.5W (802.3at) POE PSE switch.
  2. POE-PSE gives 12.95W to the POE-PD.
  3. POE-PD connects to POE-PSE LLDP and sees that the device can provide 25.5W, but it shows that the "PD Requested" field is 0.0W and "PSE Allocated" is 0.0W.
  4. POE-PD asks for 25.5W in LLDP (in "PD Requested" field).
  5. After handshake the POE-PSE reports "PD Requested" field is 25.5W and "PSE Allocated" is 25.5W (which is OK).

Another type of switch (from Rubytech) do the following

  1. POE-PD device asks for 25.5W in HW negotiation.
  2. POE-PSE allocates 25.5W to POE-PD in HW neg.
  3. POE-PD starts LLDP to check if it can negotiate down the power.
  4. POE-PSE reports "PD Requested" field is 25.5W and "PSE Allocated" is 0.0W.
  5. Even after multiple handshakes, the above values stays (allocated=0.0W). The PSE refuses to tell the PD how much it has allocated in LLDP.

I would prefer if the values shown in LLDP POE-PSE reports "PD Requested" and "PSE Allocated" fields matched what the hardware negotiation has resulted in.

It would also make it easier to detect the case where a POE injector is connected between a POE-switch and a POE-PD device.

Are there any specifications or guidelines that ask POE switch manufacturers to handle these things (LLDP) consistently?

3
  • Without looking myself, have you checked the latest additions from 802.3cv, also incorporated into 802.3-2022? I get the HPE implementation where L1 negotiation is used for "takeoff", and L2 then follows and takes the helm but I don't get partial or ambiguous implementations.
    – Zac67
    Nov 17, 2023 at 14:29
  • Yes, I agree, the HPE is correct in a way, but it does not make it possible to detect the case when a POE injector is powering the PD under test, and a POE-switch is connected behind the injector . If the proposed handling was used, that could be possible. And I agree that the second handling is just plain wrong. I'll check the latest specs and see if I find any writings about these ethings. Nov 17, 2023 at 14:44
  • A combination of midspan and endspan PSE on a port isn't included in the specs. You likely need to deactivate LLDP on the switch port for negotiations to make sense - it's optional after all.
    – Zac67
    Nov 17, 2023 at 14:52

1 Answer 1

1

Interesting question - I've tried to find an explicit answer in IEEE 802.3 clauses 33 (2-pair) and 145 (4-pair) but couldn't.

However, since L2 negotiation is optional and 802.3 generally talks about requesting a new power value, I'd venture that repeating L1 power negotiation on L2 is entirely gratuitous and results can safely be ignored. After all,

The PD is not allowed to change its maximum power draw or the requested power value when it is not in sync with the PSE.

in 33.6.4.2 may even be read to mandate keeping the current power level (from L1) when L2 doesn't match.

The basic logic is that you need to request and receive a grant for any higher power level before you may actually draw that power. You can use either L1/physical layer or L2/LLDPDU for that purpose and either grant is equally valid.

Using an additional PoE injector on an already PoE-enabled switch port may be a valid use case (e.g. for increasing power from type 1/1st gen to type 2/PoE+, assuming the PoE injector supports L1 negotiation only and passes through LLDPDUs). With the same logic as above you can treat either grant for the requested power level as valid and simply use the power.

However, it just might be the other way around and a PD may be connected to a type-2 PoE switch port via a type-1 injector. In that case, L1 could only negotiate up to 13/15 W (with the injector), subsequent L2 negotiation could grant up to 25/30 W (with the switch) but actually drawing that power would fail. That case requires you to deactivate LLDP on that switch port or limit the negotiable power to avoid an invalid L2 negotiation between PD and endspan PSE/switch.

2
  • I have checked a bit more, and received the "PSE power control state diagram" from the state diagram in page 4 here grouper.ieee.org/groups///802/3/bt/public/jul17/… and I interpret it as it says that the LLDP fields for PSEAllocatedPowerValue and PDRequestedPowerValue should be set to the same as allocated by the hardware negotioation in the INITIALIZE box in figure 145-43. If all PO PSE's did that it would solve a few potential problems I think. Dec 11, 2023 at 14:31
  • 1
    Absolutely, yes - L2 should match L1. Your question seemed to be what to do if it doesn't.
    – Zac67
    Dec 12, 2023 at 4:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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