We detect a problem where our costumers are using web sites like fast.com or beta.speedtest.net to check is bandwitdh, and results are awfull, only 1.5Mb or 2Mb, but even if we give him 10Mb from the server, the result is the same, we try giving the costumer a free bandwitdh limit plan, same result, only 1.5Mb or 2Mb, the only way we can show the client that he haves more then that, is opening some youtuve videos, and show him the SM Internal Throughput Chart where it shows that he haves more then just 1.5Mb or 2Mb, doing peaks of 8Mb or more.
We check the Frequency for interference, we try diff frequency, same result, it also affects direct downloads.
Any Ideas ?
Do you have any firewall rules over the AP or SM?
What results do you get testing at FAST.COM?
We find that SpeedTest and many others give strange results on Satellite and Wireless connections - I think they get fooled or affected by the latency. But we find that FAST.com is simpler and ''lighter weight'' and more accurately reflects the real ability to download data. Besides FAST.com is operated by NetFlix, so it also tells people exactly what they probably want to know... "how fast is my NetFlix?"
Sorry for the late answer, no, the AP have no Firewall rules
Fast.com gives as top a bandwitdh of 2Mb, but always below that, if we open a youtube video in HD, the Throughput Chart of the SM shows over 5Mb.
Is this with their device connected to their router or does it also get poor results if you connect their computer straight to the ePMP radio (bypass the wifi router) ?
Yes, we try without the router wifi, we even try using the SM as bridge, as router, with private IP, with public IP, we even try with the SM as bridge and router with a PPPoE Conection, same result. And is not the network, since we have ubiquiti radios that dont have this problem.
Hmmm. Do you have a Speed Test server locally? Meaning that the Ookla Speedtest people have a small/light speedtest you can download from them, put on a server in your shop, and test to that. It's not bad to setup - any server with php is good to go.
We do that, because sometimes we have users who are testing to some speedtest server half way around the world, or sometimes our connection upstream to the internet might be wonky or saturated or whatever... so telling our customers to test to speedtest.ourispname.com and sometimes see a better result.
That's probably NOT your problem by the sounds of it, but it is nice to be able to test to something locally and to take all the internet routing hops out of the equasion. :)
I should also mention that I have my ePMP SM set at 10 Megabit down, 1 Megabit up and that's pretty much exactly what I get from SpeedTest.net and Fast.com and SpeakEasy and other speed tests. So, I'm not sure what to say about your issue - it obviously doesn't seem to be a real speed problem, just a speed testing problem. :)
Another idea (probably my last idea) is that some speed tests are thrown off by latency. We also sell Satellite Internet, and it kinda sucks all round... but it's the only product that reaches some people. But, one of the side effects of the 1,000 ms latency is that some speed tests are really really low - way lower than it should be - while some other speed tests are OK.
So, that's also a possibility maybe? We don't see that here on our ePMP, but maybe for you somehow the latency is throwing off the speed tests? I didn't think FAST.com was affected by latency - it basically just cares about how fast you can stream from NetFlix's servers... since FAST.com is run on the NetFlix network.
But maybe TESTMY.net or some other test that is ''satellite internet latency aware'' may give a more actuate result for you? I don't know what to recommend really.
We have seen this exact problem but only when going through customer wifi routers, it doesn't happen when we connect straight to the radio. Still, this smells like a MTU issue somewhere on the network.
If it is a latency issue maybe try www.dslreports.com/speedtest they have options specifically for high latency connections like Satellite. It also tells you if the PC you are running the speed test from doesn't have enough power/resources to correctly run the speed test.
I have seen that bumping up the MTU on the ePMP STA from 1500 to 1564 has improved some subscribers' Speedtest results. I always explain that this is not truly a performance problem, and that it only impacts the kind of traffic typically generated by speedtests, not normal internet traffic.
I would start with the suggestions from ninedd and brubble1 and then throw in dshea's suggestion depending on the initial results. Also what do you get with the onboard Wireless Linktest between the AP and the SM?
Latency plays a major factor in speedtest results, especially when the speedtest server is out on the internet. The local WiFi router is something you should check. WiFi routers don't include airtime fairness by default and you may just be running into an issue where a client is under poor conditions in the WiFi network and latency is high. And in some cases where the WiFi router supports airtime fairness and is enabled, and you are the client under poor RF conditions, that will increase your latency too. Plug in your laptop directly to the SM bypassing the WiFi router and run a speedtest. If its still bad then you can eliminate the WiFi router as the issue. Start to look at the rest of the link from there on out. Installing a local speedtest server is a great idea as well. This will help troubleshoot various hops within your network and eliminate the unpredictability of the internet speedtest servers. ninedd's suggestions on alternate speedtest servers are great too.