It looks like thread-per-core stuff like seastar (C++) and glommio (Rust) are still the state of the art, along with userland network drivers. glommio seems to

It looks like thread-per-core stuff like seastar (C++) and glommio (Rust) are still the state of the art, along with userland network drivers. glommio seems to be like a seda-thread-per-core using multiple rings as staged event buffers while running on dedicated cores.