@WingWand Do you observe the same behaviour if you plug one of the satellite Deco into the main Deco, bypassing the switch?

  @yves_b I have tested, even resetting the Decos and setting up as a new network, and can confirm that plugging the child nodes directly to the main node does not circumvent the wireless backhaul preference. Both child nodes show connection via ethernet in the Deco app. 

With child node 1 plugged directly via ethernet into the main node, pings between a wired device on the main node and one on the child node vary between 4 and 15ms. If it were actually using wired backhaul, this latency would be 1ms. Worse yet, when I try to ping between wired devices connected to the two child nodes it seems to be using wireless backhaul for both steps, and my latency is doubled. 

PLEASE allow us to disable wireless backhaul for these APs. There is no reason this should not be an option. 

  @WingWand 

When you look in the app at the detail of a satellite Deco, what does it show for the backhaul?

Does it not show a combination of Ethernet and wifi?

.

I expect that the Deco will use Ethernet + wifi for the backhaul.  You observe that the Deco makes a poor choice for pings.  Deco is perhaps doing a better choice when there is more traffic, spreading the traffic among all channels.

  @yves_b This comment is not helpful as you're basically saying "that's how its meant to behave," but if it's designed to use wired backhaul, then it should be using wired backhaul. The app shows that both child nodes are connected via Ethernet, and all 3 wireless channels. But there is no scenario in which WiFi signals half way across my house between X number of walls is faster than a CAT 6A cable plugged directly in. I can see that in the latency (4-15ms vs 1ms or less). The wiring has been tested thoroughly because of this problem, and I can push 10Gbps traffic with <1ms latency over the runs in question.

Deco's mishandling of backhaul traffic is the problem. 

  @WingWand When you do a network performance test between two devices connected on two Deco, you do not observe that it starts using the Ethernet backhaul?

  @yves_b That is correct. It actually behaves the opposite of what you'd expect. As I stated in my original post, the devices default to wired backhaul when they first boot up at which point traffic behaves as expected. But after 20-30 seconds (what I assume is the time it takes to establish the wireless connection), latency times increase and become sporadic, indicating a switch to the wireless backhaul. 

I know that the wired backhaul CAN work properly, because I have 2 other nodes (W7200s) managed by the same BE16000 located in a detached shop a few hundred feet away. Those nodes work over the wired connection without any issues. 

  @WingWand 

The W7200, indeed, exclusively uses the Ethernet backhaul if it is available, while the BE16000 is designed to combine all backhaul channels. 

It is expected to improve overall throughput and reduce latency.

You mentioned disappointing latency.  What throughout do you achieve via the combined backhaul?

I suggest testing the maximum throughput without Ethernet, then adding Ethernet to see if it improves.

  @yves_b I appreciate your attempts at resolution, but this problem is not user error or exhibiting any sort of preferred behavior for a network device. It's poor implementation of aggregating back haul over wired/2.4G/5G/6G. I am a network administrator in the business space, and idea that a piece of network hardware would prefer a wireless connection over a direct wired one is nonsensical to say the least. It's either a bug or poorly coded.

Max throughput above and beyond what the wired connection can push is irrelevant to me. I have no use cases that require even near the max throughput of the 2.5Gbps Ethernet. I just want to be able to disable wireless backhaul to shore up latency and network loop issues. 

  @WingWand 

Actually, I wanted to check if you could obtain a throughput above and beyond what the combined wifi can provide.

From your description, we do not know If the combined backhaul uses the wire at all.  If not, that would be a problem with a greater impact.

  @yves_b Ah, I see what you are getting at. However, use of aggregated backhaul is of no use to me if core functionality is not maintained. I want to disable the wireless backhaul completely and have the Decos pass my traffic on the wired connection no matter the scenario as to alleviate the issues present in their implementation of aggregation. 

Again, if the devices are configured as such: 

[Workstation 1]---CAT6A---[Child Node 1]---CAT6A---[Main Node]---CAT6A---[Child Node 2]---CAT6A---[Workstation 2]

Then there's no good reason it should be routing traffic between Workstation 1 and Workstation 2, or either workstation and the internet, over a wireless connection, no matter what. This defeats the purpose of hard wiring devices to the network to begin with.

We have not even discussed the issue of network loops that the wireless backhaul is causing, which would also be solved with the ability to disable the wireless backhaul. I have seen several other users of this forum with the same problem on various network setups. This particular problem has been solved for me by having the child nodes plugged directly into the main one, but this is not a permanent solution for my use case.

I purchased this mesh system because it allegedly supported Ethernet backhaul with a switch in between. This appears not to be the case with their current firmware. I've submitted a feature request.