Sysadmin Performance

System performance analysis

Analyze and optimize performance on a Debian/Ubuntu Linux machine. Measure before forming an opinion; tune the bottleneck, not a guess. Confirm before any change that alters kernel parameters, kills processes, or disables services.

Performance analysis steps

1. Current-state assessment

CPU:

top -bn1 | head -20
mpstat 1 5

Look at usage patterns, per-process CPU, and load average.

Memory:

free -h
vmstat 1 5

Look at RAM usage, swap usage, and memory pressure.

Disk:

df -h
iostat -x 1 5

Look at free space, I/O performance, and per-disk utilization.

Network:

ifstat 1 5
netstat -s

Look at throughput, packet statistics, and errors or drops.

Processes:

ps aux --sort=-%mem | head -10
ps aux --sort=-%cpu | head -10

Look for resource hogs, zombie processes, and process states.

2. Identify the bottleneck

Common bottlenecks: CPU saturation, memory exhaustion, disk I/O wait, network congestion, process limits. Name which one the measurements point to before tuning.

3. Optimization strategies

For CPU: identify CPU-intensive processes, check for runaway processes, consider nice/renice, review cron jobs, optimize application code.

For memory: check for leaks, review swap usage, adjust swappiness, kill memory hogs, add RAM if genuinely needed.

For disk: check usage, clean up old files, tune the I/O scheduler, consider an SSD, check for a failing disk with smartctl.

For network: check bandwidth, review firewall rules, tune network settings, check for attacks.

4. System tuning

Kernel parameters in /etc/sysctl.conf: vm.swappiness, fs.file-max, net.core parameters.

Service optimization: disable unused services, tune service configs, review startup items.

Resource limits in /etc/security/limits.conf: open-file limits, process limits, memory limits.

5. Monitoring setup

Set up ongoing monitoring: htop for a real-time view, logging for historical data, alerts for issues, and recorded performance baselines.

6. Before-and-after comparison

Benchmark before changes, apply optimizations, benchmark after, and document the result.

Guide me through systematic performance analysis and give specific recommendations for this system.

---

See full content at https://github.com/HermeticOrmus/linux-sysadmin-skills.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: HermeticOrmus
• Source: HermeticOrmus/linux-sysadmin-skills
• License: MIT

Install and usage instructions live in the source repository linked above.