Wed Sep 08, 2021 10:01 am
If so, how do you measure the throughput? The packet that comes from the internet towards the public IP address reachable via the GRE tunnel occupies the download bandwidth of the WAN uplink, and then it occupies the upload bandwidth of the same WAN uplink as it is being sent encapsulated into GRE to the remote router. So if you measure the throughput first in one direction and then in the other one, this causes no conflict, but if you send the test traffic in both directions simultaneously, you already reduce the technically possible throughput down to just 500 Mbit/s per direction.
Another thing is MTU and PMTU discovery - if a 1500-byte packet is to be sent via GRE, the transport GRE packet carrying it gets fragmented, so the packets-per-second ratio doubles.
What is the load of the individual CPU cores of the 4011 while you test the throughput, i.e. what does /tool profile cpu=all show? And what device serves the other end of the GRE tunnel?