While researching a throughput issue, I am discovering another issue that is related to ePMP gear on our network.
We use Netonix swtiches everywhere. We keep flow control turned off on the ports feeding our APs (though for this discussion, it does not matter how flow control is set).
On every switch port facing an ePMP AP, we see huge amounts of Rx Pause Frames. And, the number of Rx Pause Frames = the number of Rx Multicast Frames. This does not seem right. For one, we have flow control turned off so there should not be any pause frames. It looks like ePMP APs ignore the switch setting for FC. I also see no way to turn off FC in the ePMP radio.
Two, why is every multicast frame getting paused by the switch. This happens regardless of throughput on the AP (busy or quiet). I don't think I really want any multicast traffic on my subscriber subnets. I see it there as well as my management VLAN. Turning off all VLANs but one (each one independently) still gives tons of multicast packets and they all get paused.
I don't think this is realted to my throughput issue but it made me pause (pun intended). I don't see this behavior on UBNT or Mimiosa AP switch ports.
We use a couple Force 180's as point to point radios and we see the same bahavior there as well so it is not isolated to PtMP APs.
Here is a switch port screen grab not long after we reset the counters on it: