When i set the outbound routing policy to directly use the cellular1 5G connection and i run a speedtest i get numbers of around 500MBit/s download and 100MBit/s upload speed.
When i use the SpeedFusionCloud for the routing (wan smoothing and fec all turned off) i get speed around 180MBit/s download and 70-100MBit/s upload speed.
When i use our SpeedFusionHub-VM (Running as a VM on Proxmox with 1Gbit/s up/down) and i also just use no wan smooting and no fec i get numbers of around 350MBit/s download but ONLY 25MBit/s upload speed from the Balance 310 5G router to the Internet.
All SpeedFusion Tunnels are unencrypted.
Running the speedtest, the CPU usage is pretty low on the FusionHubVM. What could be the problem? Can anyone help me to trace the error here?
I tried to change the networkadapter in Proxmox from VirtIO to the Intel E1000. Looks like that improved the performance a bit. But nothing compared to using the SpeedFusionCloud. Hmm…
What networkadapter is the ideal one for the FusionHubVm?
Speedtests run direct on a WAN link will always show the fastest result - but those results vary all the time. I am on a business cable connection. I just ran three speed tests, Speedtest.net, Fast.com and speed.cloudflare.com I got 216Mbps, 490Mbps and 360Mbps respectively.
Why? Because we’re testing against different severs at slightly different times across a inter network that is more or less busy each time. Latency varies, packet loss happens, servers and services get busy.
When looking at the performance of our WANs and of SpeedFusion I do two things repeatedly.
I run WAN Analysis between my device and my FusionHub. This shows me how much bandwidth is available between me and the cloud hosting environment / datacentre. It’s a point to point iperf based pummelling of data from one place to the other. This shows me my maximum throughput possible over SpeedFusion.
Then I do a bunch of speed tests like you have done - sending my traffic over the bonded tunnel and I turn on the SpeedFusion Graphs to see what is happening.
Then I take the graphs and try and work out what I’m seeing and adjust how SpeedFusion uses the WAN links to compensate for detected issues. More often than not now that process involves me turning on Dynamic Weighted Bonding, enabling DSL/Cable Optimization in QoS, and then measuring the bandwidth and deciding on whether I need or want to use WAN smoothing or not.
Finally I’ll decide which traffic should and shouldn’t go over SpeedFusion.
Sounds like you need to deep dive on your SpeedFusion Graphs to understand what’s going on.
