I’m using a HD4 and Fusion Hub Solo - what is the difference between using SYSTEM>WAN ANALYSIS vs STATUS>SPEEDFUSION>PEPVPN Test Configuration?
I’m seeing much higher speed when testing using WAN Analysis with the HD4 as Client and the FusionHub as Server vs what I see if I do the test using the PepVPN Test Configuration.
Which one is more accurate to help determine throughput/speed to the FusionHub?
WAN analysis is point to point raw throughput between the HD4 and FusionHub.
PepVPN Test is throughput over the VPN between the two devices.
WAN analysis will always show higher throughput than the PepVPN test. We use WAN analysis to identify raw throughput potential, for each link, then we fine tune SpeedFusion/PepVPN to give us the most stable throughput possible using those links.
What isn’t immediately obvious is that just because a level of throughput is achievable it isn’t necessarily of good enough quality for SpeedFusion to use when combining links.
To see link quality we look at the SpeedFusion graphs whilst we pummel the SpeedFusion VPN with traffic. The graphs show us bufferbloat, jitter and packetloss metrics detected by SpeedFusion whilst data is flowing and that gives us the most accurate view of individual WAN link health and quality.
WAN Analysis:
Point to point between a given WAN (or WANs) and the FusionHub. This shows the expected raw capacity outside of the PepVPN tunnel so no overheads etc.
Typically I will test connections in isolation here and then together to get an idea of the capacity of each link. Testing a single thread/stream should be representative of what a PepVPN tunnel could achieve, testing multiple parallel connections is a bit more like what you’d see with multiple users behind the connection directly.
It is normal to see higher numbers here than via the PepVPN test as there are no overehads and no bonding taking place.
PepVPN Test:
This is a test inside of the PepVPN tunnel, so should give you an idea of how the tunnel will perform, the test will run against all enabled WANs according to your config, you can also temporarily turn on/off connection in the bundle to see what effect that would have in a failure scenario or to see if running fewer connections and using hot failover may get better results than bonding. I tend to have the detailed graphs out whilst letting this test run for minutes at a time and flipping connections on and off to see what will yield the best results.