I have a simple tcp speedtest server on one machine; and web browser on another one.
Machine to machine, connected to a switch get about 900 megabits/second. Perfect.
When the machines are connected to a B11 in same vlan, host to host over same models of switches. I see maximum always about 120 megabits/second. Kind of got my head scratching.
Tinkering around with iperf, I’ve discovered that if I select 1 TCP connection (-P 1) the max I can obtain machine to machine is about 85 megabits/second. If I do 20 connections (-P 20) I can get about 640 megabits/second. From some coworkers it seems that mimosa always recommends -P 10 etc, to get more connections.
Reading over the Mimosa datasheet, it speaks about the aggregate throughput; on a per connection basis is the B11 about 85 max? Is that some sort of QoS/fair scheduling etc? Is there any way to change that behavior on the B11?
Customers use speedtest.net which seems to only use 1 connection… so no matter if they have full open channel the max they believe they get is this 85. Some of the tests seem to open multiple connections… trying to tinker with that but for now…
On other radios, I can all the time get +400 mbits on a single connection – (-P 1)
Any else have any ideas?