leandromoreira/linux-network-performance-parameters

This repository is a reference guide, not a software library. It explains where specific Linux kernel configuration settings, known as sysctl variables, sit within the network processing pipeline of the Linux operating system. The goal is to help you understand what happens to a network packet from the moment it arrives at the network card to the moment your application reads it, and where each tunable setting affects that journey. The README starts with an important caution: newer Linux kernels are already well-tuned for most workloads, and blindly copying configuration values from the internet (sometimes called cargo-culting) can actually reduce performance rather than improve it. The guide is meant to help you make informed changes when you have a specific problem, not a list of magic numbers to apply everywhere. The content is organized around two directions of traffic. Incoming packets (ingress) travel from the network card through several queues and buffers in the kernel before reaching the application that is waiting for them. Outgoing packets (egress) take the reverse path, starting from the application and moving through the kernel until the network card transmits them. At each step, there is a sysctl parameter that controls queue sizes, timing thresholds, or memory limits. For each parameter, the guide explains what it does in plain terms, why you might want to adjust it (for example, to reduce latency or handle sudden bursts of traffic without dropping packets), and the exact commands to check its current value and change it. The guide also includes a section on tools for monitoring network behavior in real time, such as reading kernel statistics files and using the perf tracing utility. It is a companion to several external posts about Linux networking internals, particularly articles from Cloudflare and packagecloud.