Hello,
Been running in circles trying to figure out a massive performance degradation using Speedfusion between a Max Transit Cat18 (8.2.1 build 5372) and 2 Fusion Hubs (8.2.1 build 5093). For diagnostic’s sake, I’m just using a single ethernet WAN on the Max Transit that is capable of 250Mbps down/50Mbps up.
When not routed through Speedfusion, I get full link speed.
When routed through Speedfusion, I get 30Mbps down/30Mbps up to either Fusion Hub. Tested both with speedtest.net and with internal speedfusion link test.
Fusion Hub 1 has symmetrical 1Gbps
Fusion Hub 2 has symmetrical 80Mbps
Fusion Hub processor loads do not cross 1% on either appliance and the Max Transit does not cross 5%.
Testing a second WAN (cellular) for bonding on the Max Transit results in an awkward 5Mbps down/20Mbps up. This cellular connection supports 130Mbps down/20Mbps up by itself.
I’d like to say it worked at full speed on the previous Max Transit firmware (8.2.0 build 5312) but rolling it back does not fix the issue.
Also, there is 0 packet loss and latency holds steady at about 90ms the entire time.
HI Jason,
the Transit Duo has 60Mbit SpeedFusion throughput with an ecrypted tunnel. Maybe you can try without encrytion?
BR
Dennis
That was one of my first thoughts, but turning off encryption on both ends doesn’t make it any faster. Still doodles along at 30Mbps.
May be worth also using the WAN Analysis tools to check the path performance outside of the PepVPN tunnel.
Set the FH up as a server and then run the remote end as a client and you can test that the end to end capacity of each WAN to each FH is what you expect it to be, generally speaking this should be able to deliver the line rate performance of each WAN as there are no overheads of the PepVPN or SF bonding to account for.
When using the built in SF speed tester do you notice any reported excess latency / loss / out of order packets on the detailed graphs?
Fusionhub is currently set up as server with the Max as a client.
WAN analysis tools show latency is fairly stable and very low most times, although under load at PepVPN speeds it does climb up to around 170ms on Starlink and 150ms on Verizon (internal cellular). Starlink (145Mbps) and Verizon (53Mbps) are stable at 170ms at full throughput outside of PepVPN.
Using the built in speed tester is showing some issues on the Starlink WAN and Verizon cellular. The issue is that results are anything but consistent between tests- sometime PepVPN flatlines around 5Mbps and other times it gets 50Mbps. Also, FEC and WAN Smoothing are disabled on both ends. I don’t know why either are being tagged on the graphs.
