According to Mimosa, I should be telling my customers that if they're using the most popular metric in the world for testing internet speeds, they're doing it wrong (I concede that while this may be technically correct, my customers - and yours too - don't do technically correct very well."
When TDMA is set to 75/25, 8ms window, MAC Tx/Rx is 980/290. This gives me as much Tx bandwidth as I require for peak times, but no one client IP can download more than 20mbps of TCP traffic (from my speedtest.net at the edge, nor anyone else's beyond my edge). When TDMA is Auto, MAC Tx/Rx is 780/780 (lower Tx, which is undesirable as it's 100mbps shy of what I need during peak hours), but TCP throughput per client is greatly increased (150+mbps). So I'm in a pickle. Either my scrupulous customers can get those coveted speedtest.net results they love seeing as they run them every thirty seconds ad-nauseum at the cost of overall Tx capacity of the link. Or I give myself some headroom in link capacity but the fastest speeds my 100mbps clients can see is 20mbps. What's even stranger is that client upload seems unaffected. I can upload 150+mbps from my test on the link no matter what TDMA is configured. I hit up Mimosa's chat support was as chipper as they were unyielding in their idea that I should test in a way that caters to the B11's shortcomings. I've been a Mimosa fanboy for a while now but boy am I feeling burned right now. Chris Wright Network Administrator
