Having some trouble diagnosing a problem AP. ePMP 1000 GPS unit into an RF Elements 90 degree sector. SM (for testing) is a Force 200. Did an RF analysis prior to choosing frequency and numbers seem good:
RSSI: -49/-56 (as reported in AP)
SNR: 36/29 (as reported in AP)
MCS: 15/11
Not bad for a suburban environment. Link distance is 2 miles.
Speed test from SM using Cambiums tool in the radio: 87 / 30 (20 Mhz channel) with medium packets and 10 second duration.
The issue now is when I speedtest.net this link. It instantly jumps to 40 Mbps and goes no higher on the download and upload stops around 9 Mbps. My backhaul is 600 mbps and Internet connection is no where near saturated.
If this is interference, would I not see that reflected in the SNR? Noise floor on the RF scan was around -80.
I'm not clear why on this AP, a radio speed test is twice a real world speed test. If it was interference, would that not also affect the Cambium speed test?
On our other AP, Cambium speed tests and real world match exactly.
No Ethernet errors or drops being reported by the Netonix switch feeding this AP. On the client side, it is hitting a Mikrotik router and then into a computer via 1000BT.
do you see pause frames in your switch? if you plug in directly to your switch in that port, and the far end of the AP cable (if its possible) what kind of speeds do you get?
No on pause frames. SNMP reports one discard in the last month, no errors on the Ethernet port. The Netonix switch itself reports a handful of errors but those are cumulative over the last 30 days. If I plug into the tower switch, I get full bandwidth (or what I should see with my network connection). On my other AP on this same tower, it gives me full speeds (radio throughput = real world throughput).
I am co-located with FM. Cable is well shielded and everything is grounded. FM is low power (100's of watts, not thousands).
Can't really drag my laptop 100' up the scary *ss tower. :-)
If I am having interference issues that affect throughput, would I see that in real world speeds tests as well as Cambium radio to radio speed tests? I would think so but whenver I assume....
the next thing to check is the amount of re-trasmission happening in the AP. if the APs is contantly adjusting for noise it will impact speed sometimes, it usually does a decent job avoiding the problem and gets the info through.
if you can, post a screen shot of your performance stats.
your getting about a 10% retransmission rate, not good... UDP won't care and will blow through, but end to end TCP is get jumbled from that, the exact imact i'm not sure, but pretty big. if you are able, use another channel, or attempt to cut your modulation step to 13 downlink, and 10 uplink and see what happens. if you can move channels ore boost gain on your CPEs would be best. the AP seems to see the noise and is keeping the modulations down well. the downlink looks like very intermitten noise. (those are guesses based on these numbers, spectum analzyer would tell more)
I think you are spot on with intermittent noise. This is not happening all the time which is what is making it so hard to troubleshoot. I did not think about retransmission rates... I'll be looking for another channel but with intermittent noise, might have to scan for a while.
you might be able to cheat, use a 5ghz wifi device (i use the wifi analyzer app from google play) you might be able to see the router (if it is one) and get a diffrent channel on it.
My thinking is the intermittent interference is at the AP side - is that your thought? It's on a major transmission tower with various other 5Ghz devices on nearby towers pointed in various directions... But, when I E-Detect, it's all green lights for APs and SMs.
I'm starting by dropping max MCS from 15 to 13 for now and letting it run a while to see what percent retransmission I get. Obviously 0 is best but is there a good, bad, ugly percentage?
i exactly tried to resolve the same problem. Wireless Link Test showed me 85/25 Mbit (20MHz channel), when i used a speedtest (ookla) it gives me only 40/10Mbit/s result. I think, it is probably som kind of limitation of Cambium device per 1 stream. Later i gave it up.
This was a recoomended steps for me:
We tried to reproduce the issue locally, but unfortunately have no success.
Anyway, we have several ideas and assumptions along with recommendations for such scenario: 1. Check Ethernet port speed configuration between AP and host below: - in case AP is configured with 100 mbps port speed and host below with 1000 mbps - please change port speed to 1000 mbps on AP - in caseAP is configured with 1000 mbps port speed but host/switch below with 100 mbps - please change port speed to 1000 mbps on host/switch
2. Since speedtest.net server that is used by user is a worldwide, we can have some unexpected losses/high latency on the route to it that cause degradation of TCP tput result.
So, our recommendation is to use local speedtest.net server right behind the AP (it can be used for trial time).
3. Run the same tests using Iperf utility just to confirm our assumption that issue is related to unexpected losses on the route to speedtest.net server.
I've dropped my max download MCS to 13 in the AP (since 50 Mbps is the fastest package we offer). That has helped. Download re-transmitted packets have fallen to under 1% and speeds are pretty contstant at 50 Mbps download now. I'll keep testing since it seems to be intermittent interference but lowering the MCS may be helping. I had already lowered the max upload MCS in each radio to 11.
hopefully we can eventually control max downlink MCS per subscriber :)
On another brand of equipment that we've used, they had a ''Rate Aggressiveness Algorithm'' which we could set on each CPE to +1 or -3 or whatever was necessary for that CPE's environment. That information was communicated by the CPE with the AP and they would step up or down by that number of steps from what it would otherwise choose with t's normal rate algorithm based on SNR.
SO - they still had a separate MAX rate settings - but the 'aggressiveness' was a great way to tune a customers down a notch or two, and yet if the RF conditions improved or got worse, it was able to automatically step up or down automatically - just 1 or 2 steps below (or even above) what the normal rate algorithm would normally pick - and we could virtually elimiate 'rate flapping'.
hopefully we can eventually control max downlink MCS per subscriber :)
On another brand of equipment that we've used, they had a ''Rate Aggressiveness Algorithm'' which we could set on each CPE to +1 or -3 or whatever was necessary for that CPE's environment. That information was communicated by the CPE with the AP and they would step up or down by that number of steps from what it would otherwise choose with t's normal rate algorithm based on SNR.
SO - they still had a separate MAX rate settings - but the 'aggressiveness' was a great way to tune a customers down a notch or two, and yet if the RF conditions improved or got worse, it was able to automatically step up or down automatically - just 1 or 2 steps below (or even above) what the normal rate algorithm would normally pick - and we could virtually elimiate 'rate flapping'.
I like! we don't run into this often, but would be a great tool to have.
Hey Giuseppe, maybe you should start a new thread and try to provide more information and clues. For instance, what firmware revision are you running on your ePMP radios? What Mikrotik routers are you using? Have you confirmed that you're not running into a CPU utilization issue while running PPPoE and running these TCP bandwidth tests?
Anyway, create a new post with more information and I think you'll have better luck of getting some assistance. Thanks!
second... i have the same problem .. i'll show my screenshot .
I have try to set my ap at 40mhz, but my bw is ever max 35mbit (with 20tcp session ... with 2 tcp session i have only about 20 mbit )
test is done from a pppoe server > ptp to the bts > mikrotik switch > ap cambium > epmp force 200 > mikrotik ap (for btest)
how u can see from the screen, the ap have only 40mbit of bandwidth in use ( 40sm registred ) .
Ap is in flexible mode and 5ms frame
you've got a lot of hardware in this link, have you verified expected performance from everyhing inbetween? 100 meg switch giving you 90 ish meg for example. the stats on your cambium radio looks good. something as simple and turning on FC in your switch maybe all you need to smooth out the delivery. or you maybe getting paused to death from a gig inbound link, all 100 meg links out that are getting busy... many many factors can show the reduction in speed you've pointed out, but the air side of the cambiumradio, from what you've shown looks good.
