>> This is the only study I know that did something like that. It was >> limited to a single router and is 2 years or so old. > this isn't a relevant study because the device in question was a software-forwarded device (i.e. no separation between control plane and forwarding plane). The devices we're talking about in this discussion are systems which have separate management planes which can be overloaded by excess traffic from the forwarding planes, unless specific mitigating configuration is used. > > Not sure why the authors of that report used a C1841 - that device dates from 2004. The VPP project maintains lots of performance measurements for modern Intel CPUs. https://docs.fd.io/csit/rls1810/report/vpp_performance_tests/packet_throughput_graphs/index.html We haven’t measured extension header cost as far as I know. Ole
