How to Tune Nginx for 100k+ Concurrent Connections: Step‑by‑Step Guide

Optimization Overview

A typical Nginx Linux server can handle 500,000–600,000 requests per second; with proper tuning it can reach a stable 904,000 req/s, dramatically increasing concurrent access.

Optimization Steps

Identify the Nginx bottleneck.

Apply configuration tweaks.

Retest with a load‑testing tool.

Identify Bottleneck

Use Apache ab to generate load, e.g. ab -n 200000 -c 5000 http://localhost:8080/index.html, and check the status page http://127.0.0.1/status for connections, waiting and other metrics.

Configuration Tweaks

Worker processes and CPU affinity

worker_processes 8;
worker_cpu_affinity 00000001 00000010 00000100 00001000 00010000 00100000 01000000 10000000;

Event model (epoll for CentOS)

events {
    worker_connections 10240;
    use epoll;
}

Worker connections worker_connections 10240; Maximum open files per process

worker_rlimit_nofile 65535;  # usually matches ulimit -n

Keepalive timeout keepalive_timeout 60; Gzip compression

gzip on;
gzip_min_length 1k;
gzip_buffers 4 32k;
gzip_http_version 1.1;
gzip_comp_level 6;
gzip_types text/css text/xml application/javascript;
gzip_vary on;

Proxy timeout settings

proxy_connect_timeout 90;
proxy_send_timeout 90;
proxy_read_timeout 4k;
proxy_buffers 4 32k;
proxy_busy_buffers_size 64k;

Efficient file transfer

sendfile on;
 tcp_nopush on;

Kernel Parameters (sysctl)

net.core.somaxconn = 262144

net.core.somaxconn = 4096

net.ipv4.tcp_tw_recycle = 1
net.ipv4.tcp_tw_reuse = 1

net.ipv4.tcp_syncookies = 0
net.ipv4.tcp_max_orphans = 262144

Set the maximum number of open files for the system: ulimit -n 30000 After applying these settings, run the ab test again to verify the performance gain and adjust parameters as needed.