Change of behavior in PoE budget from SG3xx to CBS350

[kolping-mw]

Hello,

we have the problem that we switched from SG300-10MPP to CBS350-8P-2G. So far so nice but the PoE power budget is now calculated in real time and is blocked.

Before we could power our access points easily because the real power amount was counted and therefore much was left.

Real example from today:

- 60W PoE Budget

- 4 Lancom Access 

- Real power consumption all together is around 30W

- Lancom devices reserve up to 25W per device

- If I limit via port to 15W the Lancom devices come up but via LLDP they negotiate out only 12-13W. Seems to be good... but only at first glance

- If Lancom clients have less power than 17W negotiated via LLDP they disable the second Wifi module... although they don't need that much power

What I tried:

- Limit port to 15W -> working but limited wifi power on Lancom

- Disabling "power via mdi" in LLDP messages -> same effect like limiting the port

Someone has another idea to get 4 access points online on these devices?

[pieterh]

is there an extra feature like an USB-port or an LAN-AUX port  on these Lancom device ? (i.e. LAN-out + PoE for a camera or such)
can you disable this feature so the requested power budget is less than 25Watt ?

[kolping-mw]

Good idea. I could ask Lancom if there is a dynamic adaption if you deactivate some features on the devices. BTW: 25W is the maximum I have seen via LLDP. They come out with 17W without deactivating features. This means 20W on the port of the switch (there is some tolerance between this value and the negotiated one based on the 20W). I need 15W on the switch, this means 12-13W for the access point via LLDP as seen on console.

[pieterh]

btw,   there is another model  CBS350 that has 120W budget,

CBS350-8FP-2G		120W			8
CBS350-8FP-E-2G		120W			8

Cisco Business 350 Series Managed Switches Data Sheet - Cisco

if no other solution works, maybe you can ask your vendor for a swap?

[kolping-mw]

Yep I knew this one but wanted to try if we get a workaround without swapping. At last I changed the model as there was no way to get around the budget. Whether at Cisco nor at Lancom. So thanks for the answer and can be closed like your proposal.

[kolping-mw]

Today we ran into the next annoying problem with this one... I have six access points which consume around 8 Watt and 2 other PoE devices. Together they need around 70 Watt in real. As this one changed my switch with 195!!! Watt now tells me that there is no power for more devices as the PoE devices told the switch that they COULD need more. So we are talking about 120 unused Watts. That is ridiculous that I should now install an FP switch instead of a P even there is so much power unused. Please change the behavior back to like it was with SG!

[kolping-mw]

To make it clear for all, here are the pictures. I understand the idea behind this but there should be a switch for the user to enable overbooking...

[pieterh]

looking at the picture I guess you should change from class-limit to port-limit

[kolping-mw]

Also thought about that and did it yesterday. But this isn't really better. Access points are negotiating power to 30 Watt via LLDP. If I reduce the port power too much then they throttle their wifi rate. At the moment I reduced everyone to 25 Watt and so I had the power for one more of them but still there is more than half of the budget unused at all time...

[pieterh]

>>> Access points are negotiating power to 30 Watt via LLDP. If I reduce the port power too much then they throttle their wifi rate. <<<

There is your problem !
The AP requires 30Watt for full functionality / full-speed
in normal situation not all functionality /full-speed is not operational, so the actual used power is lower than max.
but the AP reserves full power to be able to switch to 30Watt when needed
to me this seems not a problem in the Cisco switch  it responses correctly to the requested power from the AP.

what you would like to see is an option to "oversubscribe" the power budget
 not on what the switch reserves on request from the device, but based on current power usage, 
I do not think that is possible it needs to act on the reserved power bij the connected devices

[kolping-mw]

Yeah you are not wrong. It is something between as it is kind of traceable why it is happening.

It is possible to handle it this way as the former series (SG300 for example) worked like that and we had no problems with it.

AP just reserves it but doesn't need it anytime. From my side I watched them even in heavier traffic situations and they never got a consumption with over 12W. Even more than 10W is more than rare...

[kolping-mw]

Found it now: Cisco DOES it like that in Meraki. So it is possible, Cisco itself demonstrates it:

https://documentation.meraki.com/MS/Other_Topics/PoE_Support_on_MS_Switches

PoE Power Budgeting

Meraki switches will budget based on the PoE device classification, the budget is allowed to exceed available power as it's used to gauge overall power that might be consumed on the switch. Devices will continue to be powered until total power consumption goes over the available amount of power.** In this case the lowest port numbers take precedence and power will be pulled from the highest ports thus denying them power.

[pieterh]

Valuable document!
a section below mentions: 
For MS390 switches, once total power requested goes over the available amount of power, new devices will be denied power.
so behavior can differ per switch model

[kolping-mw]

For now I "only" have to convince Cisco to do it similar in CBS350 like it was anyway before in Small Business...

[pieterh]

or maybe select a different switch model which behaves like you need
I don't think you will find that particular information in the datasheets , so it may be a long journey
-> contact Cisco or your Cisco partner directly about this matter