Policing vs. Queueing

Does the ePMP  rate limiting use policing or queueing? Is it configurable?

It’s done in the Mac layer in the packet scheduler in the Aps air interface

So then when more data is coming in than the rate is set for, are excesses dropped or queued?

neither, the rate limit constricts the data flow for the information heading to the CPE. its causes the TCP ack to get delayed back to the source slowing the over all flow of information being sent to the radio.   if a specific UDP stream is setup and sent faster than the MIR is configured, it will result in discards.   normal activity over TCP will fall into a stead flow heading to your client nearly everything is wrapped in a TCP header sent over the internet. All information going into the AP is cached before being forward to the correct CPE in its assigned frame in the air interface.   

 policing and queuing, at least what i think your referring to happens in L3 control, the results are essentially the same.    the cambium gear does its rate limiting in L1, at the mac layer from my understanding of it.   it also reads IPToS and DIFFServ to influence what is happening in the mac layer of the radio as well.    

So for TCP it queues and UDP it discards? What about other protocols such as ICMP, GRE, etc?

"normal" is key here. What I'm researching isn't likely to be normal TCP. See my post in the general\business portion of the forum. Just needed to know how ePMP behaved.

all protocols will behave as they would in any other situation.  a deep understanding of each protocol will tell you what you need to know about them.   UDP for example, if you've got a steam of 120 mpbs getting to a 100 meg port, you'll have  a20% drop rate.    same will happen to  EPMP.  

ICMP, if common quality queue will behave just like it does on a jamped ethernet port.     same with every other protocol.      since the epmp does its rate limiting at L1 the same behavior you see from any other L1 connection rate being exceeded is what you will see from the EPMP.  

CDN,    a huge amount of if and thens

for example, if you use unicast TCP video, and UDP voice the situation will be handled diffrently than multicast UDP and UDP voice.   

QoS can be used to control a lot of whats going on, but what will happen to your content is a huge up in the air.  

if you can tell me more about the protocols you plan to use, how you plan on defining QOS and your distrubtion i can give you a better idea of what to expect.   but the biggest up and downside of something like this is the amount of options you have to do the same thing.  

We're not at all on the same page.

I am a WISP trying to help other WISPs with their problems. This thread was to learn how ePMP handled rate limiting and nothing more.

thought you were looking for more specific information after i noticed your thread about CDNs.   my applogies.  

In the CDN thread, yes, I am looking for specific information about problems people are having.

Sorry for the confusion.