We encountered slowdowns in routing, and implemented a simple cache
to speed it up:
https://lore.kernel.org/netdev/[email protected]/
Our patch shows considerable improvement in our use-case (hundreds
of daddrs), and non-negligible improvement in other scenarios as well.
We received some pushback from the kernel community claiming this
improvement is modest in real-life scenarios.

We measured our changes using udp floods with different numbers of
daddrs. The benchmarking setup is comprised of 3 machines: a sender,
a forwarder and a receiver. We measured the PPS received by the receiver
as the forwarder was running either the mainline kernel or the patched
kernel, comparing the results.

Does anyone have a good idea for more accurate benchmarking methods?