Master Nginx Performance: 5 Key Settings to Boost Your Backend

Nginx is a core component of large‑scale architectures; understanding its performance bottlenecks is crucial for backend engineers.

1. worker_processes

Defines the number of worker processes Nginx starts, usually matching the number of CPU cores. Optimization tip: set it to the physical or logical CPU core count (e.g., auto) to fully utilize multi‑core parallelism and avoid excessive context‑switch overhead. Adjust for blocking backend modules or high I/O workloads as needed.

2. worker_connections

Specifies the maximum number of simultaneous connections each worker can handle. Optimization tip: calculate a reasonable value based on concurrency needs and the file‑descriptor limit: worker_processes × worker_connections. Increase when using keep‑alive to support persistent connections. Ensure the OS ulimit -n limit is raised accordingly.

3. keepalive_timeout

Controls the timeout for idle persistent (keep‑alive) connections. Optimization tip: for latency‑sensitive, high‑reuse scenarios, extend the timeout (seconds to tens of seconds) to reduce TCP handshake overhead; for massive short‑lived requests, shorten it to free resources. Beware that long keep‑alive periods can consume connection slots and reduce concurrency.

4. sendfile, tcp_nopush, tcp_nodelay

These switches affect I/O and TCP performance. Optimization tip: enable sendfile and tcp_nopush when serving large static files to minimize kernel copies and packet count. Enable tcp_nodelay for low‑latency interactions (small responses, real‑time requirements). Test different combinations under load to find the best configuration.

5. client_max_body_size

Limits the maximum size of a client request body and influences buffering strategy. Optimization tip: set this value according to the maximum upload size your application permits, and adjust client_body_buffer_size to balance memory and disk usage (small buffers may cause large uploads to be written to temporary files). Monitor temporary directory I/O and space when handling frequent uploads.