High amount of data send from node-agent to controlplane

We have setup a private cluster on GKE and connect to our company network via VPN (10Mbit/s up and downstream). After this we imported the cluster in our rancher 2.2.8. For the first 2 months it works fine.

But now we noticed some connection and performance problems of our services.
After a while of analyzing we saw taht the VPN tunnel dropped packets because the bandwidth was reached every 2 minutes. The node-agents sends 700Mi data over wss every second minute to the controlplane. First we increased the bandwidth up 50 Mbit/s but the transfered
data increases up to 900Mi and we also noticed connection and perfomance problems

Can you help to fix this problem? Which logs can help to find the issue? How can we analyze which data was send and how can we reduce this data? Are there parameter to send only a minimum of data to the controlplane?

@superseb, @vincent

Any thoughts here?

In addation here is a traffic snapshot.
To mention are those spikes exactly triggered every 120 seconds (Imported GKE <-> Rancher )

How often are your back ups set. I had an epiphany over the weekend. This may be backup traffic.