Caddy adoption and the Nginx versus Apache choice

Performance and Resource Requirements

Caddy 2.8 achieves 142,000 requests per second on 1KB static files, which beats Nginx 1.26 by 22% on 16-core ARM hardware. Nginx holds a 32.8% market share of known websites, while Caddy accounts for roughly 1% of the market. Nginx is the de facto standard for proxying to application servers like Node.js, Python, Ruby, or Go. Nginx was created in 2004 to solve the C10K problem, handling 10,000 concurrent connections. Nginx uses 2 to 3 MB of RAM at idle, whereas Caddy uses 25 to 35 MB because the Go codebase includes the runtime and TLS stack. You should account for this tenfold difference in memory consumption when you scale Go-powered containers. Nginx wins on large-file streaming with a 17% advantage for files over 1 MB. Nginx also uses 38% less memory for 1 GB plus streaming workloads. When a Go upstream returns 2 KB JSON responses, Nginx 1.30.0 sustained 88,000 requests per second, while Caddy 2.11.2 reached 81,000 requests per second. Nginx maintains its lead in large-file throughput because it uses a sendfile zero-copy path and tighter buffer management. Apache has been running the web since 1995 and uses a process-per-connection model that consumes more memory under load.

Configuration and Security Defaults

Caddy uses the Caddyfile syntax, which requires only two or three lines for a basic site. The Caddyfile syntax is designed to be as simple as possible. Nginx uses the nginx.conf file with a custom syntax. Caddy automates TLS certificate management through the ACME protocol, which removes the need for manual Let’s Encrypt configuration, and the system also uses On-Demand TLS to obtain certificates during the first TLS handshake. Nginx requires manual certificate management with tools like certbot to achieve similar results. Caddy’s memory-safe Go implementation resulted in 4 CVEs since 2020, but Nginx has reported 47 CVEs in that same timeframe. Caddy adds only 0.8 ms of p99 latency overhead when using automatic HTTPS, while Nginx adds 1.1 ms because of OCSP stapling stalls during cache misses. Caddy’s local CA uses Smallstep libraries to sign certificates. Caddy provides HTTP/2, HTTP/3, and TLS 1.3 out of the box. Caddy’s ease of use makes it a common choice for developers who want to avoid the complexity of manual certificate renewals. Nginx relies on a single-threaded-per-worker model to handle thousands of connections asynchronously, while Caddy uses lightweight goroutines to manage concurrency across operating-system threads via the Go runtime.

Deployment Logic and Use Cases

Teams choose Nginx when they need raw large-file streaming throughput or granular control through a domain-specific directive language. Apache remains a standard for shared hosting environments because it uses .htaccess files for per-directory configuration. Caddy works best for small-to-medium sites where the developer experience and automatic HTTPS matter more than squeezing the last bit of performance out of a CDN edge node. One test showed that Caddy used 5% CPU for 1,000 requests per second, while Nginx used 100% CPU for the same load. I would skip Nginx if your team priorities fall on simplicity and low operational risk for certificate renewal. Nginx remains a popular choice for high-traffic websites needing maximum performance and stability. Nginx relies on its C codebase to provide a tight binary, while Apache includes hundreds of modules like mod_php, mod_rewrite, and mod_security. Nginx 1.31.5 includes a new control API and the ngx_http_json_module, but Caddy 2.11.4 remains the stable target for users who prioritize the JSON config API. Does the performance gap justify the higher memory footprint for your specific Go microservices?