@Felixmatex wrote:
Hi,this is Felix from Lagos Nigeria
Hi Felix,
How can we help you ?
best regards,
Andy.
Hi,
Now 16.0 is officially released it may be time to see if there are any improvements seen.
We can also do UL MUMIMO in this release, so aggregate throughput should see a boost if you are anywhere near the limits of the system.
I've got a few tests from our 'perfect' system that forms groups of 7 most of the time.
This is 20MHz, 2.5ms frame, 85% DL, 1 mile and 2 contention slots (AES-256 enabled).
The link test to multiple LUIDs now allows a choice of Bi-directional, downlink only or uplink only.
I'll attach a picture of each in turn.
best regards,
Andy.
3 Likes
The short range ideal conditions results in a 40 Mhz channel are cool, but I thought I would post what I consider to be some impressive performance for a real world rural WISP deployment in a challenging environment.
This 450M AP has 133 subs on it currently, some of which are as far out as 12.5 miles, and some of which have marginal at best signal levels.
It is on a tower where we have a total of ten 450/450i/450M APs, and there are also multiple competitor towers within a few miles, as well as many of our own towers that are also within line of sight, so the spectrum is crowded to say the least. We want to keep our 450M cluster in the high-power upper 5 Ghz band, so the best I could manage is a 30 Mhz wide channel
The short range ideal conditions results in a 40 Mhz channel are cool, but I thought I would post what I consider to be some impressive performance for a real world rural WISP deployment in a challenging environment.
This 450M AP has 133 subs on it currently, some of which are as far out as 12.5 miles, and some of which have marginal at best signal levels.
It is on a tower where we have a total of ten 450/450i/450M APs, and there are also multiple competitor towers within a few miles, as well as many of our own towers that are also within line of sight, so the spectrum is crowded to say the least. We want to keep our 450M cluster in the high-power upper 5 Ghz band, so the best I could manage is a 30 Mhz wide channel
4 Likes
What software are you running on this one? Just wondering if R16.0 is already implemented.
No version 16 in the field yet. We plan on waiting for the "ethernt tx stalled" fix which will hopefully come with 16.0.1.
Actually, that's great, since I think there may be further benefits to be had when Uplink MU-MIMO starts working in 16.0.1...
Replaced a 450i with an omni (75 subs), maxing out at about 150mb/s in the evenings with 2 450m. I'd say this is pretty successful. Couple weird subs we have to look at but otherwise pretty impressed with the numbers.
This is a 40mhz channel on both. on 15.2.1 as 16 seems to still have some stuff to work thru.
2 Likes
Can someone from Cambium confirm we should be setting the QOS settings in the AP to 0 to get the full thruput?
The only time these values are used are if all of the following apply:
-You are configured for RADIUS AAA or Authentication Server (Under Security > Authentication Server)
-The Configuration Source on the AP is configured for "Authentication Server" not "Authentication Server + SM" (Under General > Configuration Source)
-The authentication server does NOT have QoS values for the SM that is connecting
If all of these apply then the SM has to get its QoS values from somewhere. And in this very specific case it gets them from the AP that it is connecting to.
I hope that helps. It is a confusing setting.
Yeah we don't use radius so that won't work for us.
Will there be a software release to raise the QOS values? Technically they're too low for the thruput we are supposed to get.
Sorry. Allow me to clarify. The AP QoS values are the fallback settings used on the SM if the SM has no other viable source for QoS values.
If you are configuring QoS in the SM then the fallback settings in the AP are not ever used. Setting the value to max or min will not have any effect.
To be clear this is not a setting that throttles the AP throughput in any way.
The use case is when you have a server(i.e. RADIUS) controlling QoS for each SM. If for some reason a SM is misconfigured in the server (e.g. having no Sustained Downlink Rate defined) then you can have the AP fill in the hole in the configuration for this one SM by giving it say a default of 3Mbps that is configured on the AP.
Hi,
Here is a 3GHz test in the chamber for comparison to the 5GHz results I posted earlier.
Note 3GHz is a 8x8 MUMIMO system compared to the 14x14 on 5GHz, so the maximum group size is 4 instead of 7.
3GHz does however have more advanced rf chain hardware and much higher maximum transmit power.
20MHz, 12 SMs, 2.5ms frame, 85% DL, 1 mile, 2 contention slots.
Andy.
3 Likes
Playing with 16.0.0.1 in the chambers today looking for max aggregate throughputs at 20MHz.
20MHz bandwidth, 5ms frame, 64% DL
3GHz
5GHz
Not too bad.
Andy.
7 Likes
We seem to be maxing at 200Mbps on 30 mhz channel of real throughput during peak hour with 4.5 groups and 8X8X modulation on 90% of SU is this normal capacity ? our tests using the medusa throughput test show 420 mbps.
Just wondering if it could really hit 420Mb of real through put.
Any answer to this?
I am connected to this AP and get speed tests of 120Mbps down consistently yet can never reach 100Mbps at any point, is this actual possible throughput or an airspeed estimate?
Wibernet,
Can you attach a screenshot of how you are running your speed tests? As you may have noticed there are several options that all provide different results.
Some options will prioritize your traffic over all other traffic as you run the speed test...which isn't real-world unless you have some extreme QoS settings. Others will run the speed test alongside all of the other traffic currently being passed through the AP.
There are different purposes for each speed test.
Thanks,
Seth
Which link test mode should I choose for real world results?
All tests produce the same result, except MIR which produces 39Mbps down, which seems to be more accurate.
Hi russ007 and Wibernet,
Sorry for long wait for reply and long reply ...
The Multi-LUID Linktest works by effectively filling the queues for all SMs that are registered (and continually refilling them). The AP will then try and service these queues in as fair a way as possible. So, if all your users were fully loaded this is the throughput you should see. Of course, real world use won’t be like this, but it seemed a fair way to measure the max load on the AP.
You could get higher throughput if a subset of SMs were heavy users that grouped well (better than the average in the above test), but this is fairly unlikely.
In practice queue depths across the SMs attached to the AP will be varied and random. As more SMs require service each TDD then the ability to form groups will grow and you will see the MU-MIMO throughput gain. However, it is likely that some of the SMs in a group will not have as much data to transfer as the others. The MU-MIMO group will run for enough slots so that the SM with the most data will be serviced, while others may only fill a fraction of this. This means while MU-MIMO is still giving an improvement the overall efficiency gain is not as great as it could be. As the SM load increases further the load across the group should even out a bit more. If you have different packages for customers and multiple data channels this will also potentially unbalance the load in groups. Finally, if you have a high priority data channel enabled this will default to SU only and should be serviced before MU-MIMO groups are formed. A large number of high priority channels with high CIR would impact performance. Note, in the QoS page at the AP you can choose which priority data channels form groups (MU-MIMO Grouping QoS Level)
The radio can do the headline figure from the test but with the above limitations the figure will probably be lower. The numbers I have seen from looking at customer networks seem to be around 50-75% (but that’s a fairly limited sample). If we can get some more examples posted in this thread (good ones as well !) that would be great.
The Frame Utilization page can give you an idea of when you are hitting the limit. If the MU-MIMO utilization is high (>80%) and the average MU-MIMO group size is near that of the Linktest result then the total throughput will be reaching a plateau. There will be a bit more to be had as the MU-MIMO groups fill more evenly but not too much more. The Statistics->Spacial Utilization page can also be a guide to whether a certain section of the sector is overloaded. It splits the DL and UL %s over azimuth.
We have more improvements in the pipeline to optimise throughput, for example forming multiple groups for each TDD is scheduled for 16.1 release in the next couple of months. We are also working closely with a customer reporting a similar issue to see if we can optimise how we schedule which SMs need servicing in the most efficient manner.
I've just seen Wibernet's question
Which link test mode should I choose for real world results?
As every network will be different I'm not sure how we would go about this.
If you have gone through link planner to model what you have that is very conservative in the MU-MIMO efficiency gains. I would be interested to see if that matched the real world performance you are seeing.
Please feel free to ask more questions, it's hard to know if I've got over the details in an easy to understand way.
best regards,
Andy.
Thanks for your detailed response, based on what you have told me the AP is defnitely reaching max capacity.
A bit dissapointing I must say based on all the "500Mbps" claims, we would have been happy with 300 and the ability to at least provide packages in the 50Mbps bracket.
We have four of the 450m's up at the moment and the benefits over using a UBNT prismstation are relatively limited in our situation. Especially considering we only get double the performance over more than 20 times the price for the AP and 5 times the price for the SM.
Thanks, we await the new firmware. Are there plans for 80Mhz channels in the future?