BT10 Ethernet Backhaul Latency Speeds

[nMoose]

Hello,

I'm experiencing issues achieving expected speeds with the 10GbE backhaul on my new ASUS ZenWiFi BT10 mesh system. I've researched similar problems on this forum and elsewhere, but my specific scenario seems to be different. I'm relatively new to networking, so I'm hoping someone can help me identify the problem or correct my testing methodology.

Setup and Configuration:

• Hardware: 2 x ASUS ZenWiFi BT10 nodes
• Connection: Wired 10GbE backhaul between the two nodes
• Distance: Nodes are approximately 3 meters apart (for testing purposes)
• Cables: Tested with both the ASUS-provided Cat5e cable and my own Cat6 cable
• Configuration:
  • Both 10GbE ports on each node show a 10Gbps connection.
  • "Ethernet Backhaul Mode" is enabled in the ASUS router admin interface.
  • Backhaul Connection Priority is set to "10 Gbps WAN First", although only the 10GbE ports are connected for this test.

Problem Description:
Despite confirming a 10Gbps link between the nodes, iperf3 tests conducted over SSH are yielding speeds far below expectations. I'm consistently seeing bandwidth results under 300 Mbps.

Testing Methodology:

1. Enabled SSH access on the BT10 system.
2. Ran iperf3 tests between the two nodes.

Router Settings:

Questions:

1. Are my iperf3 test results indicative of a genuine performance bottleneck, or am I misinterpreting the results or methodology?
2. Are there other configurations with the BT10 that could cause such low speeds on the 10GbE backhaul?
3. What further troubleshooting steps would you recommend to isolate the cause of the issue?

Any help or guidance would be greatly appreciated. Thank you in advance!

 

[degrub]

run the iperf v3 tests with "parallel" 10-20 streams . 10 should saturate a 1 Gb/s link easily
Note that CPU loading can have an effect.

 

[nMoose]

run the iperf v3 tests with "parallel" 10-20 streams . 10 should saturate a 1 Gb/s link easily
Note that CPU loading can have an effect.

I will give this a try. So from what you are saying is that the 10GB bandwidth (or even 1gb) is not being reflected because I am not passing enough data. Let me know if I am not undersanding that correctly.

Thinking a bit more on this myself, I have a 10GB NAS I could direct connect to the port off the secondary node and measure the transfer of a large file from the 10GB NIC on my mac mini. That would saturate.

 

[Tech9]

Don't run it on the router, you're testing the CPU.

 

[nMoose]

Don't run it on the router, you're testing the CPU.

Hi Tech9. Meaning that the CPU on the router will max out before I am able to create enough steams to saturate the bandwidth?

So I should be running the iperf3 test through something like my mini with the 10gb nic so that the iperf is using those compute resources and the routers are just passing the traffic?

Just want to make sure I am understanding correctly. Ty.

 

[Tech9]

The CPU in home routers is a limitation for anything passing through it. You have RPi with 10GbE ports. Switching is okay since there is no CPU processing, WAN-LAN relies heavily on NAT acceleration hacks. Run iperf on something external or just test real life scenarios you are going to be using in real life.