Optimizing Nginx Reverse Proxy Performance for High-Load Environments

When operating high-throughput network endpoints, standard Linux and Nginx configurations often become a bottleneck. Under severe load, servers hit boundary conditions such as ephemeral port exhaustion, file descriptor allocation limits, and sub-optimal TCP window clamping. This guide outlines practical parameters to enforce robust L4/L7 traffic proxying.

1. Linux Kernel Tuning (sysctl.conf)

Before optimizing Nginx itself, the host underlying network stack must be configured to handle large connection queues. Append the following settings to your /etc/sysctl.conf:

# Maximize local port ranges allocation
net.ipv4.ip_local_port_range = 1024 65535

# Enable TCP window scaling and fast recycle
net.ipv4.tcp_window_scaling = 1
net.ipv4.tcp_timestamps = 1
net.ipv4.tcp_tw_reuse = 1

# Maximize backlog queues
net.core.somaxconn = 65535
net.ipv4.tcp_max_syn_backlog = 65535
net.core.netdev_max_backlog = 65536

2. Optimizing nginx.conf Core Directives

Ensure that file descriptor allocation (worker_rlimit_nofile) scales linearly with the connection limit. A worker handling multiple connections must never drop frames due to filesystem pooling constraints.

worker_processes auto;
worker_rlimit_nofile 65535;

events {
    worker_connections 8192;
    use epoll;
    multi_accept on;
}

3. HTTP/2 and Buffer Management

Mitigate L7 response clipping by increasing client body buffers. This prevents Nginx from creating temporary files on disk, moving the overhead strictly to volatile memory pools.

http {
    include       mime.types;
    default_type  application/octet-stream;

    # Timeouts
    keepalive_timeout  65;
    keepalive_requests 10000;
    send_timeout       10;

    # Buffer Configurations
    client_body_buffer_size 128k;
    client_max_body_size 10m;
    client_header_buffer_size 1k;
    large_client_header_buffers 4 4k;

    # Fast TCP Data Transfer
    sendfile        on;
    tcp_nopush      on;
    tcp_nodelay     on;
}