Was there any major change in how frame utilization was reported from 15.1.3 to 16.0.1? I upgraded a couple PMP450M APs and now see that mu-mimo utilization kicks in a lot more and peaks around 80% but the peak usage is only 90mbps even though linktest shows I should have a capacity of 220mbps on the downlink.
It may be a change in the actual usage of MU-MIMO by the sector. I will try to have an engineer take a quick look at this...
Hello,
That does look odd.
I expect that you are using SNMP to gather the information displayed in your plots.Â
Would you mind letting me know which OIDs you are accessing then I'll look throughÂ
the code for those two release versions, see what's changed, and then report back.
Best regards,
simon
Im using 1.3.6.1.4.1.161.19.3.1.12.3.30.0 and 1.3.6.1.4.1.161.19.3.1.12.3.31.0 to gather data via snmp for the 15min Frame Utilization. I currently don't graph the the 5min for our sectors.
The Frame Utilization statistics have been revised for 16.0 and some corrections have been made.
I'll describe the SU-MIMO and MU-MIMO Sector Utilization as it is implemented now.
If we consider, say, the downlink then there will be a fixed number of slots per TDD and therefore a fixed number of slots for the measurement interval you are using (15 minutes in the graphs you've presented). Slots carry either SU-MIMO data, MU-MIMO data or may be unused - if the sector is lightly loaded.
The values you are displaying correspond to the following:
(SU-MIMO data slot count in interval / Total Slots in measurement interval) * 100%
And:
(MU-MIMO data slot count in interval / Total Slots in measurement interval) * 100%
These figures are also available in the GUI under Statistics -> Frame Utilization in the Sector Utilization table.
The Sector Utilization table also shows the Total Utilization which gives an immediate indication of whether there is more sector capacity. It may be worth displaying this on the same graph.
It is worth noting that with MU-MIMO operation there may be additional capacity available even when the Total Utilization is at 100%. For example, another suitably placed subscriber may be able to form groups and carry additional data alongside the existing traffic. We often see sectors running well at full utilization and with high multiplexing gain.
There are some aspects of your graph which I don't understand and I'd appreciate a little more information. In the lower, weekly graph I see a solid green area, a blue line and a purple line. In the key below that MU is written in green and SU is blue. Do those correspond to the colours on the graph too?
In the text, you mention that the mu-mimo utilization kicks in a lot more and peaks around 80%. On the graph, if I have it right, the blue SU-MIMO line seems to peak around 70%.
Thanks for the respsonse, in regards to your questions. The magenta and dark green links in the Weekly are just the Maximum 5Min data points that MRTG plots. Since the weekly graph is a condense day graph, it plots the high points, as well as average out the days to produce a readible graph.
The main data points I look at are the solid green, which is mu-mimo utilization, and the blue line, which is su-mimo utilization. To read the graph, the most recent data is on the left hand side and the older data goes to the right. So for the daily graph if you look at the right end of it you see that the SU-MIMO utitilzation was hovering around 60%, then when more data was requested and the sector kicked in MU-MIMO you would see the SU-MIMO decline and the MU-MIMO raise as I would expect.
After the update, it seems that the sector doesn't need to reach as high of a SU-MIMO utilization before MU-MIMO kicks in. The current graphs I have now show the MU-MIMO only reaching up to 85% but the throughput on the AP has not changed very much. I had originally limited our capacity to 70mbps in our QOS system so that customers speeds are shaped correctly and we could control congestion better and we would nearly reach that limit at nights. After the update I raised the capacity limit to 110mbps, but the highest level I have seen the AP reach is 90mbps.
From reading the explanation it seems that though MU-MIMO may be at 70% or 80% that doesn't necessarily mean I am at the peak capacity of the AP. Currently my AP shows that it is grouping about 3-4 at a time and occassionally at 5 when enough customers are online. My initial fear was that I was reaching the capacity limit of the AP because I had never seen the MU-MIMO utilization reach so high before and more constantly now.
Thanks for the clarification about the axes, that makes sense to me now.
Best regards,
simon
Thank you Fld Wrls for your post, very clarifiyng.
Have you done more test about the reaching capacity?
We have a Medusa that do 60% Mu-MIMO Utilization with 100mb passing (link capacity test with multiple VC show me 200 mb) and i'm little worried abaout that...
Hey Telemar,
I haven't had much time to really push this particular PMP450M. I am currently working to add graphing all the utilization numbers to give me an idea of what I really have. From what I have seen the MU-MIMO utilization can be a little misleading based on the explaination above and what I have seen in real life. There are days when MU-MIMO utilization will reach 60% and max capacity is 90mbps. There are other days when MU-MIMO utilization reached 80% capcity but only at 75mbps capacity. The key is making sure you have the customer base spread over over the sector to see its true potential.
Even with these numbes I do plan on increasing customer speeds to see how the AP reacts. We currently offer plans of 5mbps, 8mbps, and 12mbps and I plan on doubling those numbers to see if I can push the AP harder. I'll may start a new tread with the data so people can see what it does for me.
Thank you!
I'll waiting for your updates!
Well, I think this could also have to do with the nature of MU-MIMO and the signal quality variation to my knowledge....
MU-MIMO utilization also doesn't hint at the grouping if its 2,3,4,5,6,7 groups (does it count 1x group as mu-mimo?)
And on top of that its possible customers with your worst modulation are the one using RF resources, a 2 group mu-mimo with 2 customers at 16qam, for instance, downloading a ton on the panel.
Just a p
Good points Csup,
There is no MU-MIMO group of 1, that is always counted as SU-MIMO.
As you point out MU-MIMO Utilization % doesn't reflect the size of the group, a group of 2 will look the same as group of 7. There are also cases where not all SMs in a group are sent data for all of the MU-MIMO slots (depending on data to be sent to each).
You can look at the Average MU-MIMO Group Size and the dynamic Slot Grouping table to get an idea of how hard the MU-MIMO is working.
If you run a link test to multiple LUIDs this will give a group size number for when MU-MIMO is working it's hardest across the sector. Only when the average group size is nearing this test number will MU-MIMO be near maxing out. You can also compare Mux gain numbers from test->live in the same way.
best regards,
Andy.
AndyR, what you think about that screen? (attached file)
50-60mb passig traffic and 160 connected SM's.
With 100 MB Passing we have +90% sector utilization and 60% MU-MIMO Downlink utilization.
Can you "translate" the reading of the MU-MIMO avarage group size? 3.5 is good? what is the max?
Because the "Total Utilization" with 90% scare me (78% with 70 mb)
@Telemar wrote:AndyR, what you think about that screen? (attached file)
50-60mb passig traffic and 160 connected SM's.
Hi Telemar,
That (attached file) doesn't look as though the AP is nearing it's limit.
With 160 SMs you should have a good spread across the sector and so have high potential max group size.
The link is still running SU traffic for 57% of the frames. As traffic increases the Total % will go up as the SU & MU fill more of the slots of the frame, it will then start to lower the SU % and increase the MU % as the opportunities to form groups occur. As even more traffic comes in then, even as the total utilization seems to be nearing the maximum, the group size can and will increase so the max throughput shouldn't plateau.
The average group size is 3.4, this is an average of the values in the table above it discounting the 1(ungrouped) as that isn't counted as a group. The Multiplexing gain takes in to account the SU (ungrouped) data, so is smaller at the moment as most data is SU. The max values can be derived from a link test (or link planner will also give this type of info if you are planning new sectors or replacing APs)
Which leads on to ......
What does a link test (Multi-LUID) give you in terms of throughput, group size and multiplexing gain ?
It will depend on how demand varies across your sector but unless you are over 75% of the Link Test results I would think the AP should be coping fine.
best regards,
Andy.
I think cambium needs to work on the wording and maybe do some articles explaining utilization since the 450m came out, the fact is after years of watching total frame usage and suddenly that being basically irrelevant I think it scares most operators.Â
I think the other big issue is that there's no real way to say an MU-MIMO sector is "full" as technically unless your at 100% MU-MIMO and 7 grouping technically theirs still theoretical space.
Then theirs the issue that since not all your customers have the same signal quality, and customers don't use the traffic at the same time at the same areas it makes things even weirder to estimate congestion, as some days you might have a specific narrow area of your sector above average and maxing out performance drawing all the resource that it looks like the sector is full but in reality if users in the other azimuths were to use the internet it would operate fine because they could be grouped with the existing user frames on the other path.
Having total utilization and having total frame utilization i think is a bit wacky and confusing and seriously makes it difficult for me to tell my bosses "hey this carriers congested we need to add another 450m for additional capacity".Â
Like was said earlier, 160 people well spread out could be amazing and you could get 90% theoretical, but then again maybe 140 of those people live in a narrow 20degree area of your sector and then technically can't take advantage of higher groupings.
I was just thinking, maybe a good way to test throughput might also be to run a MU-Link test but narrow it down to your most active customers, not the full 160, (maybe pull a list from the throughput tab and test against the top 30%), that way you'll see what kind of sector throughput to expect when your heavy hitters are all actively punishing your sector.
@CSup wrote:I was just thinking, maybe a good way to test throughput might also be to run a MU-Link test but narrow it down to your most active customers, not the full 160, (maybe pull a list from the throughput tab and test against the top 30%), that way you'll see what kind of sector throughput to expect when your heavy hitters are all actively punishing your sector.
Yes, I agree, this is an excellent way to test.
I'm sorry but i've not understand very well this part:
"It will depend on how demand varies across your sector but unless you are over 75% of the Link Test results I would think the AP should be coping fine"
75% of what?
Another question, can you be more specific about the test you and Csup suggest?
Tools > Link Capacity Test
Set it to Link Test Multiple LUIDs and set the LUID List to test to 0 to test against the whole sector.
That will give you the bandwidth if everyone on your sector tried to max out their circuit at the same time, this shows how much the sector could push and be fully congested.
But since that's a perfect condition and that won't ever happen, you won't ever have perfect conditions so that value as the maximum throughput isn't precise for what you can really expect.
That's, why he says to take that number and at around 75%, that will more likely what you can really expect based on your real-world customers and their signals
