@Michael Meluskeywrote:Data for the 120 customers went from 155Mbps to 180Mbps.
Lesson learned......and Cambium, 310Mbps is not enough total QoS capacity for 40Mbps channels.
Thanks Michael,
Great datarate and interesting observation on the QoS cap. I'll look in to it, especially with UL MU-MIMO and other DL MU-MIMO improvements on the way.
best regards,
Andy.
Hi Michael,
Sorry for the delay getting back on the QoS issue.
These settings are used by an SM if no other setting is available as the 'authenication server'
some more details are in this thread
An individual SM can never get more than 310Mbps total (even in 40MHz 5ms) so that is where that number is from.
It may have been better if this was automatically adjusted to reflect the DL percentage in the radio page settings. On 40MHz it is not unreasonable to go over 155Mbps DL to an individual SM (even at 64QAM) so yes well spotted that this needed changing too !
So in summary it's an setting used on a per SM basis (if no other QoS server setting is found) not throughput at the AP that is being set here and it should reflect the DL percentage. Link Test to a single SM will give you the theoretical max throughput in each direction for 6X and 8X performance.
Do you have some users with 'unlimited' packages on the sector doing speed tests ? :)
Hope that makes sense. Not a part of the user guide I had ever read before !
best regards,
Andy.
Hi,
I've just been chatting with a US WISP who has allowed me to share the Link Test results they have in a 20MHz channel, 75% DL, 3 mile range and 4 contention slots. They hope to get around 150 SMs (from 36 now) on here in the near future. This might just squeak in as the best on 20MHz so far. Good grouping being seen.
Thanks Simon,
Did you want a cambium T-Shirt ? :)
Andy.
We loaded up an AP with 200 customers in a 40Mhz channel (it usually has 120 customers, moved SM's from adjoining AP's for the test....a very busy tower).
233Mbps average throughput with spikes over 400Mbps.
80% CPU utilization (a feature of 15.1.5)
Steady ping times, still able to do 25Mbps speed tests.
Customers are mostly on a 5Mbps plan, but 25% on a 25Mbps plan.
Awesome Mike... it would be interesting to know if all of those "ungrouped" guys are having high priority traffic, or if they could be persuaded into groups to further improve your results.
@CambiumMatt wrote:Awesome Mike... it would be interesting to know if all of those "ungrouped" guys are having high priority traffic, or if they could be persuaded into groups to further improve your results.
Matt, I don't know how to persuade SM's into groups. We are very excited about these results and are going to keep the AP at 200 customers for further evaluation. Bumping up customer speeds is on our agenda in the next few weeks.
Hi Michael,
Thanks for showing us these results, very promising.
Matt, note the grouping picture that Michael shared is from the Frame Utilization page, NOT the Link test to Multiple VCs. This means that group of one is perfectly acceptable in this case as the AP is not running anywhere near full sector/MUMIMO capacity, so some VCs will be serviced in SUMIMO mode.
If a Link test to Mutiple VCs is done and still shows a significant proportion in groups of one that would be more worrying/interesting (unless they are assigned as HP and not allowed to form groups).
Not the first time these two tables have caused a bit of confusion (even in this thread !)
best regards,
Andy.
2.7% ungrouped on the multi-VC link test.
281Mbps with 92% efficiency, with 250Mbps of traffic flowing.
So I'm assuming 531Mbps total capacity.
Hi Michael,
We think it's possible the CIR settings may be hampering the link test result, you said most SMs were limited to 5Mbps. Upcoming releases will allow the link test to ignore CIR settings (via radio button).
Could you share the field_diags.cgi from the unit to help us understand the setup ?
I sent a PM last week with my details and how to collect them.
best regards,
Andy.
Hi Andy,
We have a PMP 450M with 86 clients off it. So far we are not seeing the throughput we would expect from a 450M. We are hoping someone at Cambium could assist us in figuring out the bottleneck.
I will send files from the AP later this afternoon once the peak usage hours arrive. Below is some information on the AP, the environment, configuration, and what we have observed so far:
All of the clients are within 1 mile of the AP. The AP is running on firmware version 15.1.3 at 5.1 GHz, 20MHz channel, 2.5 ms frame period. Frame configuration is at 75% downlink.
According to cnMaestro and our Solarwinds installation, we are seeing the AP top out at around 132 Mbps with 99% DL frame utilization. Sector utilization usually settles around 56-63%. The switch the AP plugs into is negotiating at 1Gbps so that is not a bottleneck.
We have received numerous complaints lately from clients off the AP. Each SM is set to 12Mbps down / 3Mbps up. Yesterday we ran a speed test on one of the troublesome customers. A capped speed test from the AP -> SM was only receiving 3.6 Mbps down whereas a speed test without MIR hit 60Mbps down. The AP does run through RADIUS but I have not seen any evidence of the server breaking a sweat. If RADIUS was the issue, I would expect to see complaints from clients off other APs, but that has not been the case.
We originally had 4 contention slots and tried bumping that up (4 ->5->8->12) but that did not make a difference in throughput. We tweaked the QoS settings on the AP, though it appears via reading the user manual that only impacts each SM throughput.
Thanks,
Rachael
Hi Rachel,
Thanks for the detailed background information.
field_diags.cgi for the 450m and engineering.cgi from a troublesome SM would be a good place to start, email direct to me.
I'll have a quick look but if it's down in CIR/MIR settings I'll probably have to get extra advice from other engineers on Monday.
best regards,
Andy.
Hi Andy,
The diagnostics have been emailed to you.
Hi Rachel,
Nothing has turned up in my inbox, although I think office365 was having problems yesterday.
I'll send a private forum message to confirm the details.
best regards,
Andy.
re: QoS settings, entering zero is the way to get around the 310Mbps aggregate capacity limit.
|
Brady Joseph Jun 18, 8:27 AM CDT Chat started: 2018-06-18 12:59 PM UTC (12:59:18 PM) Brady Joseph: QoS on AP questions.. (01:09:24 PM) *** Purnankh Dhankhar joined the chat *** |
Thanks Michael,
re: QoS settings, entering zero is the way to get around the 310Mbps aggregate capacity limit.
Did you get a change in throughput after doing this ?
best regards,
Andy.
Hi Andy,
We changed each field under the AP Bandwidth Settings header in the QoS tab to 0. However we have not seen any significant improvement this afternoon.
I ran a link test to an SM around 1PM and received our standard 12Mb down / 3 up. Then this afternoon around 4PM the SM was capping out at around 3Mb down / 3Mb up.
I have emailed the newest diagnostic files to you from this afternoon.
Would it be possible to have an engineer look at this tomorrow? This is becoming a major issue for us. We are having many, many complaints from customers off of the AP.
Hi Rachael,
I'm not in the office today to look at the files myself but I will forward them to several others who should be able to go through them.
best regards,
Andy.
