Boost Nginx High-Concurrency Performance by Up to 10×: Practical Tuning Guide

Worker Processes

Nginx uses a master+worker multi‑process model; each worker is single‑threaded and can run on a separate CPU core. The default single worker underutilizes multi‑core CPUs. Setting worker_processes auto (or the number of cores) enables parallel request handling and reduces context‑switch overhead. Optionally bind CPU affinity with worker_cpu_affinity auto.

File Descriptor Optimization

In Linux, network connections are file descriptors. Nginx consumes a descriptor for each connection, log file, and backend socket. Insufficient descriptors cause “too many open files” errors. Raise the system limit fs.file-max and the per‑process limit ulimit -n. In Nginx config set worker_rlimit_nofile 100000; and worker_connections 65535; to support high‑concurrency long‑connections and massive static file serving.

TCP Connection Queue Optimization

When request arrival exceeds worker processing speed, new connections queue. A small queue leads to dropped connections or client timeouts, especially in flash‑sale (秒杀) systems. Increase the Linux kernel backlog with net.core.somaxconn = 65535 to allow more pending connections.

Production‑Ready System Settings

Rather than using temporary sysctl -w commands, persist the tuned parameters in system configuration files. Example settings:

fs.file-max = 1000000
net.core.somaxconn = 65535
net.ipv4.tcp_max_syn_backlog = 65535
net.ipv4.tcp_tw_reuse = 1
net.ipv4.tcp_fin_timeout = 15
net.core.rmem_max = 16777216
net.core.wmem_max = 16777216

These values together enable Nginx to handle ten times more concurrent requests in production environments.