network-performance-tuning - Performance Tuning for Networking
This page contains a summary of some performance tuning techniques that this system is using.
This system is automatically using SMP affinity for every physical network controller, if supported.
A processor core is assigned to handle all interrupts of a certain network controller which will result in minimising cache misses, reducing network latency and quite possibly increasing throughput.
The algorithm is trying to balance all network controllers across all processors.
See /proc/interrupts for the distribution of interrupts.