Welcome to Demystifying Kernel-Level Tuning for High Traffic VPS. When you install a fresh Linux distribution on a VPS, the default kernel parameters are optimized for a typical desktop or light server workload. If you're running a high-traffic web server, those defaults will bottleneck your hardware.
1. The TCP Stack Bottleneck
By default, Linux limits how many TCP connections it keeps in the 'TIME_WAIT' state and how long it keeps them. During a traffic spike, your server might run out of ephemeral ports because thousands of connections are stuck in a waiting state. Tuning net.ipv4.tcp_fin_timeout and enabling tcp_tw_reuse allows the kernel to recycle these ports immediately.
2. Maximizing File Descriptors
In Linux, "everything is a file," including network connections. A busy web server handles tens of thousands of concurrent connections. If the OS file descriptor limit (fs.file-max) is too low, Nginx or Apache will throw "Too many open files" errors, and users will see 502 Bad Gateway pages. Increasing this limit is a mandatory first step in server tuning.
3. Switching to BBR Congestion Control
TCP congestion control dictates how your server sends data over the internet. The traditional algorithms (Cubic/Reno) react to packet loss by drastically slowing down transmission. Google developed BBR (Bottleneck Bandwidth and Round-trip propagation time), which reacts to actual bandwidth limits rather than packet loss. Enabling net.ipv4.tcp_congestion_control=bbr can drastically improve website loading speeds for users on poor connections.
4. The Somaxconn Queue
When connections come in faster than Nginx can process them, the kernel puts them in a listen queue. The size of this queue is controlled by net.core.somaxconn. The default is often 128. If 200 users connect in the exact same millisecond, 72 of them are immediately dropped. Bumping this to 4096 or 8192 provides a buffer during traffic bursts.
5. Swappiness
Swappiness controls how aggressively the kernel moves RAM contents to the disk swap space. On a database server, swapping to disk destroys performance. Lowering vm.swappiness from the default 60 to 10 (or even 1) forces the OS to use physical RAM for applications as much as possible.
Conclusion
Before you pay to upgrade your VPS CPU or RAM, ensure your operating system isn't putting artificial speed limits on your existing hardware. Tuning sysctl.conf is the secret weapon of veteran systems administrators.