It's a bird, it's a plane, it's

Everything here is my opinion. I do not speak for your employer.
August 2016
September 2016

2016-08-03 »

I've been testing our aforementioned wifi latency reduction patches in a few more conditions to see how they fare. Here's the latest.

These are clips of http://6-dot-gfblip.appspot.com/ on my laptop while my iPhone generates traffic. tl;dr our patches to the AP greatly reduce latency impact (down to about 10ms) when the AP is sending stuff downstream. Not-so-surprisingly, we don't fare as well when the station (which we haven't patched and thus can't control) is sending things upstream as fast as it can.

Nevertheless, the upstream situation seems much worse than it should be. I'd expect, with only one AP and two stations ever transmitting (and only one of them transmitting a lot), that my gfblip ought to get at least 1/2 or 1/3 of the airtime. If stations get scheduled in a mostly-round-robin fashion, and each frame takes at most 3ms of airtime, then that should be only 3-6ms of added latency when only one station is blasting away. But instead we're seeing ~100ms of added latency, which is just nuts. Oh well, more debugging...

I'm CEO at Tailscale, where we make network problems disappear.

Why would you follow me on twitter? Use RSS.

apenwarr on gmail.com