Linux Kernel Optimization: Critical Parameters
💡 Quick Tip
Technical Note: Always back up /etc/sysctl.conf before applying changes to the kernel parameters.
Introduction to Kernel Optimization
The Linux kernel is the heart of any distribution, acting as the fundamental mediator between hardware and software. While most modern distributions come with balanced general configurations, a sysadmin can "tune" the core behavior for specific workloads, such as high-traffic web servers, massive databases, or low-latency workstations. The primary mechanism for these runtime adjustments is the sysctl interface.
Memory Management and Swappiness
One of the most debated parameters is vm.swappiness. This value (0 to 100) defines the kernel's tendency to move processes from RAM to the swap partition.
- A low value (e.g., 10) is ideal for servers with plenty of RAM to avoid disk access latency.
- A high value is useful for systems with low memory to keep the filesystem in cache.
The Network Subsystem (TCP/IP Stack)
For servers handling thousands of simultaneous connections, the default network stack is often insufficient. Parameters like net.core.somaxconn define the listen queue limit for incoming connections. If this value is too low, the server will reject connections under heavy load. Another vital adjustment is net.ipv4.tcp_fin_timeout, which reduces the time a connection remains in the TIME_WAIT state, freeing resources faster.
I/O Schedulers
The kernel allows choosing how disk read and write requests are managed. For SSD and NVMe drives, algorithms like None or Kyber usually offer the best performance by reducing kernel processing overhead, whereas for older mechanical disks, BFQ helps prioritize interactive processes.
📊 Practical Example
Real-World Scenario: Preparing a Server for High HTTP Traffic
Suppose you manage an Nginx server experiencing drops during traffic spikes. Logs reveal "SYN flooding" errors. We will optimize the kernel via /etc/sysctl.conf.
Step 1: Increase connection limits. Edit the file and add:
net.core.somaxconn = 65535 (Increases the connection queue).
net.ipv4.tcp_max_syn_backlog = 8192 (Improves resistance to request floods).
Step 2: Optimize port reuse.
Add net.ipv4.tcp_tw_reuse = 1. This allows the kernel to safely reuse sockets in the TIME_WAIT state.
Step 3: Apply and verify. Run sudo sysctl -p to load the configuration. To verify the open file limit, check fs.file-max and adjust it to 200,000 if necessary.
Step 4: Monitoring. Use tools like netstat -s to confirm SYN packets are no longer being dropped and performance has significantly increased.
