using a Max Transit Duo here with a FusionHub. We have two tunnels, one with Wan-Smoothing+FEC enabled, the other one is just for SpeedBonding.
Whats the best way to find the maximum MTU size for a device that is sending packets out thru the tunnel so they don’t split up / fragmented into two pieces?
The Max Transit Duo is using a normal WAN connection (cablemodem) and 2x LTE connection as WAN interfaces.
When i try pingstests (deny fragmentation) from a linux machine via the Tunnel (2x LTE-Wan active), the test shows a ping-mtu size of 1472 to be ok, adding 28 bytes to that will result in the standard 1500 bytes mtu size. However, when the traffic is forced to be not routed via the tunnel but for example via LTE-1 directly, that drops to an mtu size of 1428. When a LTE-Connection is disabled, so the tunnel only established on one LTE-Connection instead of two, than the max mtu size using the SpeedFusion tunnel also drops down to that 1428.
Ok, but even with that 1428 mtu size for example, if you transmit pakets out there with that mtu size, the graphics on the SpeedFusion tunnel are showing fragmented packets. So, the overhead for the tunnel itself is not in this mtu size. I have done a test with one LTE-Wan connection and the tunnel and i had to set the mtu size on the client back to 1332 to not produce any fragmented packets. Thats also did not change when i actived again 2x LTE-Wan connections for the tunnel.
So, can someone explain to me what is happening when there are 2x WAN or more available? Are packets from LAN-Clients that are larger than this 1428 be split into two paths on the tunnel?
Whats the optimum max. MTU size in that scenario coming from a LAN-Client going out via the tunnel into the world?
In my test, i used an unencrypted tunnel.