WNDAP660 Connectivity Issues

I'll try to keep this story as short as possible.

I run a small office and about two years ago I upgraded an older netgear G only access point to a newer WN203. All was well for about a year. Then I started allowing smart phones to connect to this access point. Almost immediately I started noticing random periods of lagginess. Laptops sluggish to browse the internet or even access internal resources etc. I tried tracking down what device was causing it but the only thing that would ever work was to get everyone to restart their phones.

I figured that was just the nature of buying the then cheapest AP that netgear offered. I used it as an excuse to upgrade and I wanted something with 5GHZ.

So I bought the WNDAP660. I set it up and flashed to the latest firmware (3.5.20.0). I only have one access point for this office and it is located in the center of the room, rarely getting more than 10 clients connected. I decommissioned the WN203 and gave it away.

All was well for a few days with the WNDAP660 but then to my horror I started having problems again. This time instead of getting random periods of lag, I would get random periods of non-connectivity. All clients would be connected with a good signal but nothing could go in or out of the WLAN. I started hunting around again trying to see if removing a certain client would fix the problem but to no avail.

After a few weeks of dealing with this with no solution and having removed most of the other clients (I only had a few laptops and my cell phone still connected), it was still happening. The next time it happened again I pulled out my cell phone (Samsung Galaxy S6) and turned off the wifi on it. To my amazement and horror, everything on the entire WLAN regained connectivity. So apparently for the last year, all the issues were caused by my own phone in my pocket. I never suspected it since it is an expensive phone.

I verified this  several more times over the week. Each time clients would loose connectivity on the WLAN, I'd pull out my S6 and toggle the wifi from on to off, and back to on  again and the problems would go away.

Initially I just tried to accept the fact that when wifi issues popped up that I just needed to toggle the wifi on my phone. But I'm getting tired of this. I can accept that some mobile devices have flaky wifi cards, but what I cannot accept is one single device bringing down an entire wireless lan. And being able to do this on an enterprise grade access point is even worse.

So I've tried to be able to replicate the issue on demand. I don't know if its the exact same issue, but it certainly has the same symptoms.

I noticed if I ping flood the galaxy S6 from any device (whether on lan or wireless doesn't matter), it will eventually stop responding to the pings and at that same moment all clients on the WLAN will still show as connected but will not be able to access anything UNTIL I toggle the wifi on the galaxy S6. Then all goes back to normal. I've made a video of it here:  http://ow.ly/gvlh307oosh

It is a video of 4 consoles running ping on a desktop on the wired LAN. The top console is the galaxy S6. I'm using the linux ping command with a short interval (ping -i0.01, although ping -f will work too). To replicate the behavior, you have to kind of "pump" it a little bit (start and stop the flood a couple of times, but it always does the trick). The second console from the top is another mobile phone on the WLAN. The 3rd is a laptop on the WLAN. And the very bottom is the IP of the WNDAP660 itself on the wired LAN (I never lose connection to the LAN as the video shows).

As you can see, I have to "pump" the S6 a couple of times but within 20 seconds it has already stopped responding to the pings and then a few seconds later the entire WLAN stops responding. I let it sit there in that state for about a minute and then I toggle the wifi on the S6 and all is well and restored again. I let it run in that state for another minute and then "pump" the S6 again which causes it to stop responding and bring down the WLAN again.

I've tried to use the packet capture feature of the WNDAP660 but I really don't see anything interesting in the captures. I've tried flooding other clients and multiples at a time, but the only thing that brings down the WLAN is the S6. I've done packet captures of floods of other clients. They all seem very uniform, simply a steady stream of packets with each side requesting block acks. The S6 starts out fine returning block acks but then it stops. Looking at the captures it is hard to tell where the connectivity is lost, it seems packets are always flowing but they are just being ignored. It makes me think the AP is waiting for an acknowledgement and is ignoring everything until it gets it. With that idea I looked around online to see if mishandling of block acks in 802n would cause a denial of service. I did find some speculation from another forum ( http://ow.ly/qWU8307opEb ) but nothing really concrete. If anyone knows of a better way to capture traffic, I'd be interested to hear but honestly I think it should be in netgear's hands now. Subsequently, disabling AMPDU (and thus block acks) seems to get rid of the issue with the S6 but brings speeds to 802g levels.

So now I'm left to try and figure out who's fault this is. I'm sure Samsung as some to blame but I doubt they can do anything about their devices that are already out in the wild. That leaves netgear to fix it. But looking at this forum and all the complaints agains the WNDAP660, I don't think netgear will do anything either :(  Should I just file a bug anyway and hope for the best? I'm really tired of messing with the stuff, I just want something that works. I've even looked at buying a different brand but all the normal recommended brands are all cloud only solutions that require controllers or extra licensing. I just don't see the need in my case for a small office.

So in summary:
A few times a week, all devices connected to the WNDAP660 are unable to transmist data
Turning the wifi on a Samsung Galaxy S6 off and then back on again fixes the issue
Doing a ping flood against the S6 seems to recreate the issue on demand
Disabling AMPDU in the settings seems to prevent the issue but kills throughput.