Netonix Forums

Well, UPS finally showed up at the door with my new Netonix's 8 port switch! Took me a day to get the test bench set back up.

1 gig CPU <-> ERL3 router <-> switch ? <-> NanoBridge ~~~ NanoBridge <-> 100 Mbps CPU

Ran IPERF3 down it in TCP mode with about 8-12 streams

Current setup, resembling what our main backhaul is doing!

Used a Netgear dumb switch (which is what we currently replaced the UBNT TS5 unit with) ERL3 was put in "100 Half-Duplex" on the LAN port. Which was what we had to do to get any decent throughput out of the system when we originally found the TS problem a year ago, and did our tests. Signal strength on the link was set for @-60 Dbm and 216 Mbps on air connection rate with 80% Quality and 49% Capacity of the link! (Usually what we get at night on that long haul link)

IPERF posted @ 55 Mbps Download on the test link.
I changed the ERL3 back to "Auto" LAN port negotiation and test posted @ 7 Mbps Download (the dreaded 1 Gig to 100 Mbps mismatch)

I replaced the Netgear switch with the new Netonix switch, and ran the test ! Got @ 65 Mbps Download ! 65 Mbps Download
This was out of the box, NO firmware upgrades yet! Changed the setting on the FC controls - showed NO Change in throughput!

I am Impressed, an immediate 10 + Mbps improvement !

Not through doing more testing, we are going to be trying out some NanoBeams in place of the NanoBridges to use the 1 Gig ports on them. That should hopefully give us an idea of what the AirFiber X's will look like in the system with their 1 Gig LAN ports!

We will probably also crank up this link to higher output, and see what it does with a full 300 Mbps connection rate, and full Quality and Capacity readings on it!
Also want to try some different firmware on it, currently running 5.6 Beta 9 in it, and the Beta 6 might be better? Also want to try "Alternative Modulation" settings.

But so far, great improvement !

Thanks,
Wayne

I still recommend leaving Flow Control on the 1G up link interface and all the 100M airMAX radio interfaces. Our Flow Control uses "Back Pressure" whereas not all Flow Control implements this.

In fact I have found just leaving Flow Control on all ports does not hurt anything.

Read this thread if you have not already.
viewtopic.php?f=30&t=452

I agree on leaving the FC one, but wanted to give the tests with it on and off. I expected that there would, in fact, be a difference on the link setup.
But, it seems that the switch has sufficient buffers in it to overcome the back pressure, without having to trigger the need for FC !

Since it was late yesterday, and I was only able to do about 4 tests, this is a preliminary test.


I will be doing more today, with different configurations on the links, radio firmware, and such!

Wayne

Yea but when the 100M port is receiving 100's of streams from different sources different speeds with tons of small packets and the AP gets hung up on a poor link the Flow Control will kick in.

I know you recommend that Chris, and you and I have had the "talk", but we'll have to agree to disagree for now.

I have seen no difference doing tests with FC on or off in my networks. I have seen a huge difference swapping TS for Netonix switches. The buffers really do make a huge difference.

On some networks I'm pushing upwards of 300megabits/second to microwave links and Ubiquiti radios that range anywhere from 3-5megabits/second to 15megabits/second.

Well, did additional tests yesterday.

We changed the NanoBridges out for NanoBeams on the testbed, these had 1 Gig ports on them.
IPERF3 tests (8 to 12 streams) showed a 25 + Mbps improvement over the original setup!

Unfortunately the IPERF only does the testing in "Duplex" mode, Wish I could get a One direction mode test on it?

But I am really impressed, awaiting the additional Netonix switches to come in, and I will be able to then test having a switch in front of the Radios, and behind the radios to see if that improves the whole link? Suspect it will !

It will be really interesting to see what we can get out of this network when the Airfiber 5 X's get in and are installed on the links?