Latency is the True Bottleneck of Web Applications

TL;DR Throw your worries away and ride your bike :) When discussing the performance of a web application, the focus is often on how many requests per second (RPS) the server can handle. However, what is often underestimated is network latency, which can have a greater impact on application speed than raw server performance.

Local Benchmark vs. Real Deployment

When testing an application on localhost, the numbers look different compared to a real network environment. For example, a Wapp-based CGI application (ufko.org), might handle 60 RPS locally, while PHP with FPM reaches 900 RPS. However, once deployed to a real network, these numbers drop significantly ... Wapp framework at 20 RPS and PHP at 60 RPS. This drop highlights that latency and network conditions are the true limiting factors.

What Causes Latency?

• Physical Distance – If a server is in the US and a user is in Europe, latency will naturally be higher.
• Number of Network Nodes – Each router, firewall, or proxy adds delay.
• TCP Handshake – Every new connection requires a three-way handshake.
• TLS Handshaking – HTTPS improves security but introduces additional delays.
• DNS Resolving – If a client needs to resolve the server's IP first, this takes several milliseconds.
• Network Congestion – If the network is overloaded, packet loss and retransmissions increase latency.

How to Optimize Latency and Performance?

• Keep-Alive & HTTP/2 – Reusing TCP connections instead of creating new ones repeatedly.
• Reducing the Number of Requests – A single page requiring 4 requests (HTML, CSS, favicon, logo) can be optimized with caching.
• Using a CDN – Offloading static files to a CDN improves speed.
• DNS Optimization – Using faster resolvers and reducing TTL where possible.

When Does Performance Really Become a Problem?

If an application can handle 20 RPS, that translates to 1.7 million requests per day. Assuming an average of 4 requests per page (HTML, CSS, favicon, logo), this supports 425,000 pageviews per day. That’s the point where scaling becomes necessary. If your site has only 1,000–5,000 pageviews daily, server performance is likely not your main bottleneck.

That’s why focusing on latency, caching, and request optimization is often more critical than simply chasing higher RPS.