3 ms·
It uses Thread-Per-Core architecture where Netty/Reactor use thread-pools to process work and have dedicated threads for selecting from network.
by xedin 11y ago
It uses Thread-Per-Core architecture where Netty/Reactor use thread-pools to process work and have dedicated threads for selecting from network.
- crudbug 11y agoSo CPU affinity provides further optimization ? I am thinking less context switches. [1] http://conferences.sigcomm.org/imc/2010/papers/p218.pdf http://conferences.sigcomm.org/imc/2010/papers/p218.pdf
- xedin 11y agoYes, exactly since it allows for fewer context switches and cache locality not only for the network packets but for disk I/O as well.