AF11FX Traffic Issue on Netonix Switches

[Mac_SPITwSPOTS]

We currently have two sites connected to each other using a pair of AirFiber 24HD's and a pair or AirFiber 11FX's. Using OSPF splitting the traffic across both links. All radios are plugged directly into Netonix switches and powered off of them using 48VH. OSPF Path cost is set to 10 based on 500Mbps expected capacity over each of the 2 links for an aggregate of 1Gbps.

While both links are live, and running traffic, we are seeing poor performance. For example sure peak hours, we tend to see an upper limit of around 240-280Mbps across each wireless link. If we run traffic over only the AirFiber 11FX we see aggregate traffic of around 300-340Mbps on the upper limit. If we run traffic over only the AirFiber 24HD we see aggregate traffic hit 600Mbps+.


So far for troubleshooting we have physical swapped the AirFiber 11FX to no improvement, tested multiple firmware builds on the AirFiber 11FX with no improvement, checked and verified that the cables are operating at full 1Gbps and pass cable diagnostics, there are little to no port errors on either side of the link being reported in the Netonix interface. At this point any further software testing on the AirFiber feels like .

Are there any reports of Netonix switches not working properly with AirFiber 11FX's specifically? Our next step after getting home from WISPAPALOOZA is to test the AirFiber 11FX radios plugged directly into our routers, bypassing the switches to see if there is any improvement.

[sirhc]

Please post up Screen grabs of the switch UI TABs, Status, Ports, VALN, STP, Device/Status.

I use OSPF in my WISP and I use 2 ports of the switch for each backhaul segmented off with VLANs to create a POE injector that I can control.

I outline it here in this thread: viewtopic.php?f=30&t=452#p2961

There is also a link in that thread to the 1.5 hour video I made showing my tower switches.

[Mac_SPITwSPOTS]

I will get screen posts here shortly, however please keep in mind that the airFiber 24HD link works as expected.

As far as I can tell from looking over your thread that you linked, we are set up in the same fashion. The only problem we are experiencing is poor capacity when the airFiber 11FX link is online either handling all traffic or splitting traffic across both links.

[Mac_SPITwSPOTS]

Hi Chris, sorry for the delay in getting these uploaded but here are the images you requested.

Status

Ports

VLANS

STP

Device_Status

[sirhc]

Please tell me what is on each port and its purpose.

From what I can see you are running all your radios through ports 23 and 24.

If you go look at my setup I separate each backhaul radios out into a separate 2 port midspan injector with each backhaul radios going to a single port on the router.

You do not want backhaul radios sharing ports in the router and you definitely do not want local tower radios sharing ports on the router with backhaul links.

Go back and look at my thread that shows my typical tower setup and also in the video I explain why each backhaul wants its own port in the router.

In my towers I make up to (4) midspan injectors with the following 2 ports for backhauls:
Ports 1 and 25 - Port 1 powers AF Port 25 with copper SFP goes to router port
Ports 2 and 26 - Port 2 powers AF Port 26 with copper SFP goes to router port
Ports 3 and 5 - Port 3 powers AF Port 5 goes to router port
Ports 4 and 6 - Port 4 powers AF Port 6 goes to router port

Then I create a STATIC LAG to my Cisco 2951 with ports 23 and 24 that service all local tower radios such as APs and private high end customer PTP links

This way Pause frames and other issues from local radios can not affect backhaul traffic coming into and out of tower and one backhaul can not affect another backhaul.

Backhaul radios should NOT share router ports

[Mac_SPITwSPOTS]

Port 1 is the AF11FX link that backhauls from this site (site B) to site A.
Ports 23 and 24 lag to another switch at this site.
The ospf ptp VLAN's are tagged on port 1 and port 23 and 24 to the other switch where they then are tagged on the router port that the ospf link exists on.
None of our router ports are shared between backhaul links and access point links. We do have a single port set up to handle both backhaul links from Site B to Site A. I will look into testing with each backhaul radio going to a dedicated router port though as well as watch through your video.

[sirhc]

Please list what is on each port, remember this all makes perfect sense to you as this is your network, I need to wrap my mind around it.

[sirhc]

Also I wonder why WISPs would run jumbo frames?

The worldwide web is standard MTU of 1500 + needed overhead for encapsulation and in wireless a large MTU is not always a good idea because wireless retries cause more repeat air time.

[Mac_SPITwSPOTS]

@Chris after ongoing testing we were able to get a load on our AirFibers that resolved the issues finally. At this point you are welcome to lock or delete this threat at your discretion. Re. the jumbo frame we have ospf operating between all sites at a higher mtu size. Is there any detriment to running at 9600 vs say 5000, or just slightly higher than we need to?

[wtm]

Get the new 4.0.1 firmware for the AF11x units. You will find that you can then get almost what the capability says it should be getting. We did a 4.0.1 dev test on the software, it has worked wonders. Now it is available as a stable release. AND we have them powered by Netonix switches on both ends. The stats in the Netonix look extremely good with the new firmware. And the AF11x Flow Control seems to work with the Netonix LIKE IT SHOULD HAVE DONE IN THE FIRST PLACE !