@Au Wireless wrote:
There is the old saying "fix something until you break it." I have a ePMP 1000 AP with 17 clients on it. It is perfoming pretty well with a mix of 25 and 50 Mbps connections (couple 75 Mbps). However, it is on a noisy tower and in a fairly noisy environment. Perfect for what the new ePMP 2000 claims to be able to fix. So I thought I'd run some tests.
I opened up the 1000 AP to full speed on all SMs and raised the MCS rates all the way. I then measured signal strength, SNR, MCS and speed test (radio test, not Internet) at various times during the day to get an average at all 17 SMs.
what do you mean your rasied the rates?
The ePMP 2000 with beam steering antenna was configured exactly the same as the 1000 AP. It was installed (the 1000 was removed) and I ran the same suite of tests again.
I found signal strength for the download remained about the same (small improvement), signal strength for the upload improved across the board and MCS rates stayed about the same. However, speed tests were very dissapointing. Download speeds slowed down across the board. In some cases, fairly significantly (20 - 30 Mbps). Upload speeds however increased in almost all cases. Looks like beam forming works.
the 2000 isn't designed to improve download rates, its ment to resolve uplink problems when used as an AP.
if you have a problem with really bad uplink performance, your real world TCP performance will improve from a clean uplink. the radio linktest will not show an improvement on the downlink. itd a UDP test. a test from speedtest.net would show improved downlink ONLY when the uplink has gotten so bad TCP ACKs are having a hardtime getting through.
I have run eDetect tests on the AP and other than 2 SMs, no reported bad interference.
edetect will only show you WIFI or EPMP APs that are exact on channel issues, try using spectrum analyser to check for other issues overlapping or co-channel issues.
I then moved the ePMP 2000 to a slightly cleaner 5 Ghz channel. Signal strength went up (from an average of -60 to an average of -53) and SNR improved by 5 to 10 points. MCS rates also went up. But speeds on the radios dropped even further. Again, eDetect from the AP does not show anything alarming.
What am I missing? What else should I be testing? I have a complete spreadsheet and am going to publish the full results but I need to get a better picture of what is going on before I start drawing conclusions...
the EPMP2000 is specificity driven to deal with noise, when an AP, just uplink issues. in your shoes I'd suggest using the spectrum analyser, pick a channel that looks clean overall. including your edges. or return to your previous channel. the over the air test from the radio shouldn't show much of a downlink improvement. but you should get a noticeable uplink improvement. (thats the goal of the e2k) in the real world tests, if your uplink was getting butchered from noise, you'll get a noticeable improvement in downloads, or speedtest.net. a UDP test wont expose that improvement as they don't wait for ACKs before sending the next packet, they just flood the link.
if you've got a CPE having downlink issues, if you can get a spectrum pull from it. if you see crazy co-band noise, you can use an e2k lite to improve it. i've never seen that happen, but it is a possibility. it would need to be crazy loud like -25 to bother the radio. across the band.
I'd suggest asking your customers about the performance changes for them. if any of them used speedtest.net much they will have a history you can see. TCP sessions is where you will see your results. and of course, thats where it matters most.
i'm hopefully going to complete my testing results by the end of the week on the e2k. i've really picked on this thing and frankly, i'm impressed.