Linux System Tuning: Essential Tips for Peak Performance

Optimizing your Linux system unlocks hidden potential, whether you’re running a web server, database, or workstation. Here’s a detailed guide to key tuning areas:

1. Kernel Parameter Tuning via /etc/sysctl.conf

• Increase open file limits:

   fs.file-max = 1000000   # Max system-wide open files

• Optimize memory usage:

   vm.swappiness = 10       # Reduce swapping tendency (0-100)
   vm.vfs_cache_pressure = 50  # Favor inode/dentry cache retention

• Network performance:

   net.core.somaxconn = 4096   # Increase connection backlog
   net.ipv4.tcp_fastopen = 3    # Enable TCP Fast Open (TFO)

  Apply changes: sudo sysctl -p

2. I/O Scheduler Selection

Choose a scheduler matching your workload:

• SSDs/NVMe: kyber or none (noop) for low latency.
• HDDs: mq-deadline for balanced throughput.
• Database/VMs: bfq (Budget Fair Queueing) for fairness.

  # Set scheduler for NVMe drive
  echo "kyber" | sudo tee /sys/block/nvme0n1/queue/scheduler

3. Filesystem Optimization

• Mount options in /etc/fstab:

   # For XFS/SSD: noatime, discard (TRIM)
   UUID=... /mountpoint xfs noatime,discard 0 0
   # For ext4: data=writeback for faster writes (trade reliability)

• Reserve disk space: Reduce reserved blocks (default: 5%) for large disks:

   sudo tune2fs -m 1 /dev/sda1  # Set to 1%

4. Memory Management: Transparent Huge Pages (THP)

Disable THP for database workloads (e.g., MongoDB, PostgreSQL) to avoid latency spikes:

   echo "never" | sudo tee /sys/kernel/mm/transparent_hugepage/enabled

Permanent option: Add transparent_hugepage=never to kernel boot parameters.

5. Network Tuning

• Increase buffer sizes:

   net.core.rmem_max = 16777216
   net.core.wmem_max = 16777216

• TCP congestion control: Use bbr (Google’s BBR) for high-speed networks:

   net.ipv4.tcp_congestion_control = bbr

6. Security vs. Performance Tradeoffs

• Spectre/Meltdown mitigations: Disable for performance-critical systems (if risk-acceptable):
  Add mitigations=off to kernel boot parameters.
• Limit sysctl hardening: Avoid overly restrictive settings like kernel.kptr_restrict=2 on internal systems.

7. User Limits: /etc/security/limits.conf

Prevent resource exhaustion for services (e.g., Nginx):

   www-data  hard  nofile  50000  # Web server user
   *         soft  nproc   65535  # Global process limit

8. CPU Governor

Set CPUs to performance mode for servers:

   sudo apt install cpufrequtils
   echo 'GOVERNOR="performance"' | sudo tee /etc/default/cpufrequtils
   sudo systemctl restart cpufrequtils

Monitoring & Validation Tools

• Check disk I/O: iotop, iostat -x 1
• Network stats: ss -tulpn, nload
• Memory: vmstat 1, htop
• Tuning impact: Use perf or sysbench for benchmarks.

> ⚠️ Critical Advice:
> – Test changes in staging environments first.
> – Monitor system stability/performance after each tweak.
> – Document customizations for recovery/auditing.

Final Tip: Automation tools like Ansible or tuned (for RHEL/CentOS) simplify applying tunings across systems. Start small, measure relentlessly, and tailor settings to YOUR workload’s DNA! 🚀

Need deeper dives? Explore kernel documentation at kernel.org.