Optimizing Web Server Performance: Reducing Concurrency Pressure, Memory and CPU Usage

Modern web applications experience a rapid rise in concurrent connections due to richer page elements and more complex interactions, which puts significant pressure on server resources; simply adding hardware is costly, so technical optimizations are required.

Front‑end optimizations—such as leveraging HTTP cache headers, local storage, conditional requests (Last‑Modified/Etag), and merging multiple requests—can dramatically lower the number of HTTP requests and thus reduce server load.

On the server side, memory usage is critical; Apache’s prefork, worker, and event MPM models illustrate trade‑offs between process isolation, thread safety, and memory efficiency, while Nginx’s event‑driven architecture offers a lighter‑weight alternative.

Using kernel features like sendfile and memory‑mapped I/O cuts unnecessary data copies between kernel and user space, saving both memory and CPU cycles.

CPU consumption is further mitigated by replacing inefficient select/poll with epoll for I/O multiplexing, minimizing context switches, and reducing lock contention in multithreaded environments.

The article concludes that the optimal solution depends on specific business requirements, but combining front‑end caching, efficient server configurations (e.g., Nginx + PHP‑FPM or Apache event MPM), and kernel‑level optimizations yields the most resource‑efficient web system.