Hi,
I have a 10G connection in a datacenter and a 1G (symmetrical) connection in a remote location, latency is around 30ms, 0 jitter. When I run the bandwidth test directly between both sides (using the public IP addresses) I get 900+ Mbps. However when I create an EoIP (between one CCR1009 and one CCR1036) I only get 150Mbps, I tried creating multiple EoIP links and bonding them together and same issue I only get 150 Mbps combined. I checked both CCR’s CPU and no cpu core is saturated.
I would say it is error in the testing method. I couldn’t do it properly (correctly, Device-under-test should not be the one which generates/consumes the traffic) so I just took the first two devices I know of where I already have EoIP and ran the btest between them.
In UDP mode, it easily went to 1Gbit (physical connection is 1Gbit). In TCP mode it got stuck on 190-200Mbit. I would rather guess that TCP mode of BTEST does not work correctly than assume there is any issue with EoIP. for precise results, I would rather recommend proper tools (e.g. iperf) and never generate/consume the traffic on device-under-test.
Does not matter if the CPU gets only few % utilization - there may be some effect which you just don’t see.
edit: Alternative possibility may be if you use IPSec - maybe the encryption is causing some slow downs?
Aside from what has been said already, note that packets traveling through EoIP tunnel may require fragmentation if you don’t have MTU set high enough on all routers/switches between tunnel end-points. Also check that you have same ROS version on both sides, since there were changes made to btest somewhat recently.