APUP Performance Testing Results
Hello OpenWrt community, my name is Javier From Argentina, I didn't want to create a huge post, here so feel free to ask more information if needed.
I'd like to share some exciting performance testing results for Access Point Micro Peering (APUP), which is a relatively new and promising approach to wireless mesh networking.
What is APUP?
For those unfamiliar, AP-UP (Access Point Micro Peering) is a simpler alternative to Ad Hoc, WDS, 802.11s mesh, Multi-AP, and EasyMesh. It was developed by Gio and allows APs to communicate between themselves while still functioning as regular APs that can serve client stations as described in this Freifunk blog post
My Testing Setup
I've been testing APUP with the following hardware:
- Two MT7621 based routers.
- Each router has a MT7915 radio in close proximity with small antennas (office environment).
- All the physical setup remains the same, and only network modes where changed.
- Tested TCP throughput using netperf from one router to the other.
Key Findings
AP-UP mode shows a significant performance loss compared to AP and Mesh modes.
It also generates more driver interrupts during the same test runs.
Packet captures shows the same number of beacons in AP-UP and classic AP mode (10 per second).
remote_capture_netperf.pcap equivalent files (ap-up,clasic AP) show some sequences witch differ from the same file in classic AP mode. APuP mode appears to have "[TCP ACKed unseen segment]" flags, also shows numerous acknowledgements in sequence from the same device, with occasional Clear-to-send frames meanwhile classic AP mode shows more structured traffic. You can find .pcap files in the corresponding folders.
It doesn't seem to be a cpu bottleneck, I relocated the drivers interrupt handling to a secondary cpu but got no performance improvement.
Cause/s ?
I think I'm experimenting what mwarning wrote in his article
"But be careful. If we would now set up a third device router C similar to router B (or even router A), then any packet will be send between routers A, B and C in a loop, due to the bridged interfaces. This will cause the CPU and bandwidth utilization to be pushed to the maximum and the network will be unusable."
But I cant get a clear diagnostics, nor a solution to achieve the same performance that I wold have in AP mode. I don't see repeated packets or packets whit the same src and dst.
Questions for the Community
Have you had a different experience?
Can you help me understand what is going on and debug this issue?
Are there any specific aspects of APUP performance you'd like me to test further?
Detailed Test Results
If you're interested in the complete test results, methodology, and configuration files, I've documented everything I thought is relevant in my GitHub repository:
The repository includes raw test output, configuration files, and summaries that you can use to replicate my tests or adapt for your own environment.
Looking forward to your thoughts and feedback!