2

I have a networking problem. I have a network that runs of a POE switch that is routed to a relatively deep chain of POE powered switches that also passthrough POE. The problem becomes that the load on the POE becomes too large.

So to alleviate this issue I would like to put POE injectors at certain locations to "refresh" the power so to speak. But I have no clue if that is possible. I really can't use externally powered POE switches everywhere due to power routing constraints.

Any help is appreciated!

Below is an architecture image what I would like to achieve.

networking

3
  • Pass-through PoE is normally only useful for a single switching layer: PSE/PoE switch -> pass-through switch -> PD, due to power restrictions and the consumption on the PT switch itself. Could you add a meaningful diagram to your question for us to see where your pain points are? Otherwise I'd just say "use bigger or more PoE switches".
    – Zac67
    Commented Nov 14, 2022 at 10:35
  • @Zac67 Yes! See the diagram I have attached
    – John Smith
    Commented Nov 14, 2022 at 10:40
  • This has some XY to it. Some more realworld details might help explain how this monster came to be. I think you might just have to get AC power to more locations, and star-out from there with POE over UTP, and have a backhaul line from each new POE switch back up to the core. Sorry- but boring is often best when it comes to robustness of network design.
    – Criggie
    Commented Nov 14, 2022 at 23:42

2 Answers 2

6

That looks pretty horrible - I don't think any reasonable network engineer would build something like that. Chaining switches is generally considered bad design[*1] and it's aggravated by using PoE.

A (much) better design would be to use a larger PoE switch and run dedicated cabling to each powered device. Even using two or perhaps three PoE switches would definitely improve things.

But if that works you, intermediate injection should work as long as it provides enough power. Depending on the cabling (length and AWG) and the equipment at hand, simple pass-through extenders may also decrease the available voltage to a point that it doesn't meet the specs any more.

[*1] One problem with chaining switches is the large number of single points of failure. Another is that chaining multiplies the switching delay to an amount that it may become noticeable. Also, if you run an STP variant (which is somewhat standard in a professional network), the chain may exceed the protocol's design envelope and the spanning tree may never converge.

7
  • What are the issues you are foreseeing? This would save me a lot of headaches since it would be very hard and costly to do it properly. The biggest issue I see is that if one switch is down the entire network after it is gone.
    – John Smith
    Commented Nov 14, 2022 at 11:09
  • Yes, that's one aspect. I've added a footnote to my answer on that.
    – Zac67
    Commented Nov 14, 2022 at 12:33
  • Classic "spaghetti". Sometimes you do what you gotta do.(TM)
    – Ricky
    Commented Nov 14, 2022 at 22:22
  • 2
    Even just lowering the tree depth by increasing the fanout would significantly improve things if using one large switch is truly not an option. Commented Nov 15, 2022 at 2:48
  • It's not that a larger switch is not an option there simply is not enough room for cables.
    – John Smith
    Commented Nov 15, 2022 at 8:30
2

(Zac has already covered the insanity of such a design, so I won't repeat it, but I do agree.)

What you have built looks like any telco/cable line power system. When you've gone farther than your ability to transmit power (do to usage and/or distance), you simply put another power inserter. (and a DC block)

In your PoE monster, it doesn't matter where the power comes from. The Powered Devices (PDs) always get power from the directly connected system. (The PSE is always the thing at the other end of the ethernet cable. Where the PSE gets it's power is irrelevant.) You can put a stand-alone "midspan" anywhere you want - just as you can put one anywhere in your switch and patch landscape. The limitations are (1) where there's power to run the PSE ("midspan"), (2) where there's access to the ethernet cable, and (3) where you're close enough to get the necessary power to the PD. (i.e. 60W over 100m is not ideal.) [#1 has brought you to #3, which is going to require fixing #1.]

2
  • is it safe for the input to a PoE injector to have PoE? Commented Nov 15, 2022 at 5:11
  • @user253751 Generally, yes. A PoE injector isn't powered from its uplink, so whether there's a PSE connected or not doesn't matter.
    – Zac67
    Commented Nov 15, 2022 at 10:44

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.