I noticed that adding customers in an ePMP 75/25 2.5ms GPS synced AP I have a lower upload speed.
For example, enabling only 1 customer to connect, I have 10Mbps.
Enabling 30 customers, I have 5Mbps on the same user of the previous test, even if no user is making traffic (I did this test during the night).
Is that normal?
Is there a way to have a better upload with more customers?
I noticed that all my ePMP access point with about 30 or more customers have a max upload of 4-5Mbps in good wireless conditions, even with ePMP 2000 with beamforming.
What MCS levels are used on the uplink? Check Monitor -> Performance page and look at the MCS spread.
At night before you run the test, do you see any traffic on the uplink at the AP? Do you see the uplink packet count increasing? Or do you see downlink being used? If so, the other SMs could be using the uplink for acks although this will be low traffic. But it will depend on how many of the SMs are doing it.
If you’d like someone to look into it further, please contact our support team. They will be able to quickly determine if what you are seeing is normal or not.
I noticed that adding customers in an ePMP 75/25 2.5ms GPS synced AP I have a lower upload speed.
For example, enabling only 1 customer to connect, I have 10Mbps.
Enabling 30 customers, I have 5Mbps on the same user of the previous test, even if no user is making traffic (I did this test during the night).
Is that normal?
Is there a way to have a better upload with more customers?
I noticed that all my ePMP access point with about 30 or more customers have a max upload of 4-5Mbps in good wireless conditions, even with ePMP 2000 with beamforming.
Thank you
Lots of little things to check here:
1: are you using public IPs directly in your radios, or 1:1 Nat? or any other method that port scans can waste your traffic? if you can access the GUI yourself via those public addresses or nated IPs. that mean those radios are likely getting port scanned and that will have a domino effect on your devices, poorer and poorer performance as you add devices, add management VLANs if thats the cause and you'll see an immediate gain in performance (strongly advise to never have GUIs facing the public domain, your CPEs able to respond to pings/port scanns out to the web. if your not sure about this and want some help, PM me and i can see what shows up on port scans on your public facing side and help you bury all of it to get as much waste traffic out as possible.
2: SM isolation and port isolation on in your switch and AP? wasteful broadcasting stings more and more with that many CPEs.
3: is your uplink load expected to be high? the link test will not stop active traffic to give you feed back. it will only use available bandwidth.
4: in your AP, what are your %%% of your uplink MCS states? if your seeing the greatest % of uplink traffic at MCS10 for example, your gross uplink performance will be poor. better channel, or action to increase the MCS state of the poorest subs would be needed. more use the e2k with beamsteering if noise is a problem.
Hi Chris, Thank you very much for your useful answer!
1) That’s a great point! I have private IPs for most SMs, but for public IPs I have them assigned to the SM, so I can load the ePMP GUI from the internet and then 1:1 NAT from the CPE to customer’s router. Is there a way to make the management separate? Could you please post us some example? Anyway, I didn’t see so much traffic, but you’re right, it could be an issue. My SMs are doing PPPoE authentication.
2) SM isolation is disabled on the AP. I have a routed network so no switches, only routers, but could you please post here a sample of your configuration about that? Does SM isolation help?
3) No, during the test the upload load was very low and almost zero.
4) I’m already using an ePMP 2000, my upload is about 8% MCS13 45% MCS12 and 40% MCS11. The rest is on other modulations (very small part, 4%, on MCS15). Downlink is 15% MCS15, 33% MCS14, 36% MCS13, 10% MCS12. Unfortunately that’s the best channel available. Do you think having upload interferences on some users can make total performances worst?
Disabled: This is the default mode. When SM isolation is disabled, an SM is able to communicate with another SM, when both the SM’s are associated to the same Access Point (AP).
Enabled: When SM Isolation feature is “Enabled”, SM#1 will not be able to communicate with SM#2 (peer-to-peer traffic) when both the SM’s are associated with the same Access Point (AP). This feature essentially enables the AP to drop the packets to avoid peer-to-peer traffic scenario.
These threads pop up every once in awhile, where someone is concerned with upload speed in fixed 75/25 mode and I have yet to see a solution. Â I have yet to hear of anyone that has good upload performance in 75/25 mode with a decent amount of clients. Â The only solution which really isn't a solution is to switch to 50/50 mode.
Cambium always responds to these threads but it almost seems like a stall tactic and then eventually the thread dies and nothing ever gets resolved.
Maybe a 66/33 mode would be a better balance?
Or if anyone is having good upload performance please respond here. Â Let us know what firmware you are using.
These threads pop up every once in awhile, where someone is concerned with upload speed in fixed 75/25 mode and I have yet to see a solution. I have yet to hear of anyone that has good upload performance in 75/25 mode with a decent amount of clients. The only solution which really isn't a solution is to switch to 50/50 mode.
Cambium always responds to these threads but it almost seems like a stall tactic and then eventually the thread dies and nothing ever gets resolved.
Maybe a 66/33 mode would be a better balance?
Or if anyone is having good upload performance please respond here. Let us know what firmware you are using.
we have good performance on 75/25, our highest offering is 50x10, and don't have a problem getting that speed. most of the slow downs i've seen is wasted airtime that can be fixed. network secuity is more than just protecting your network from security realated problems. keeping port sweeps, broadcast storms, customer equipment that likes to discover things, like windows 10. etc. etc. etc. all of those near empty packets wont' show much bandwidth, but will trash your capacity.
@giuseppe4 wrote: Hi Chris, Thank you very much for your useful answer!
1) That's a great point! I have private IPs for most SMs, but for public IPs I have them assigned to the SM, so I can load the ePMP GUI from the internet and then 1:1 NAT from the CPE to customer's router. Is there a way to make the management separate? Could you please post us some example? Anyway, I didn't see so much traffic, but you're right, it could be an issue. My SMs are doing PPPoE authentication.
2) SM isolation is disabled on the AP. I have a routed network so no switches, only routers, but could you please post here a sample of your configuration about that? Does SM isolation help?
3) No, during the test the upload load was very low and almost zero.
4) I'm already using an ePMP 2000, my upload is about 8% MCS13 45% MCS12 and 40% MCS11. The rest is on other modulations (very small part, 4%, on MCS15). Downlink is 15% MCS15, 33% MCS14, 36% MCS13, 10% MCS12. Unfortunately that's the best channel available. Do you think having upload interferences on some users can make total performances worst?
the bulk of your traffic is on a lower modulation state, if you look at the throughput chart, MCS11 isn't going to give you much throughput. I'd suggest taking steps to improve the link modulations.
as far as limiting waste traffic, we use management VLANS on all APs and CPESs so the interfaces for our use are not routeable. (these units can't get to the internet at all. no NAT, local routing only) the CPEs are all in NAT mode, very little exception to this so stop enduser waste traffic. the APs isolate CPEs. the switch that the APs are plugged into isolate the APs from each other so the end setup is the AP thinks its the only AP attached to the router, and the CPE looks like its the only CPE thats attached to the only AP. with that degree of isolation, your risk of wasted airtime is reduced. when you clear up your waste traffic, you;ll get max performance from lower modulations. removing waste traffic will help your latency too. it will help keep your scheduler clean and give more airtime to actual customer traffic.
I have to work on VLAN to separate management from user data. How do you manage that exactly?
What performances do you have on your AP?
Unfortunately I have some APs in very populated areas with very high noise floor, so even beamforming couldn't help so much. You're right about MCS11, but that's the max I can achieve there. I have lots of interferences, and even if users are not so far (less than 3km, so less than 2 miles).
Competitors like to use 40/80MHz channels without even care about their performances...
How much users do you have on the ap with 50/10 profiles? How much distance from the AP?
