Mastering Server Performance Tuning: A Practical 5W+1H Guide

Overview

In this blog post the author explains performance tuning in plain language, using the 5W+1H method (What, Why, When, Where, Who, How) to structure the discussion.

Hardware configuration: CPU Xeon E5620 x2 (8 cores), 16 GB RAM, RAID 10 disks, OS: CentOS 6.4 x86_64.

What is performance tuning?

Performance tuning means deeply understanding and adjusting the relationships among hardware, operating system, and applications to maximize overall system performance and continuously meet business demands.

Hardware includes CPU, memory, disks, NICs, etc.; the OS includes processes, virtual memory, file system, network, and more; applications include common services such as Apache, MySQL, Nginx, Memcached, and others.

Why is performance tuning needed?

Two main reasons: to achieve better system performance and to satisfy growing business demands. The author illustrates this through hardware selection, OS distribution choices, and application configuration.

1. Hardware selection

Server hardware should be chosen based on the specific application type (load balancer, web server, application server, cache server, database server, backup server, monitoring server, etc.), with recommended configurations for each.

2. Operating system

The book Linux Performance Tuning by Fernando Apesteguia explains that default OS settings aim for broad compatibility, often leaving performance‑critical options disabled. Choosing the right distribution (e.g., CentOS 6.4/6.5 rather than a brand‑new release) can avoid unnecessary bugs while still providing needed features.

3. Application

Example Apache MPM configurations are shown:

<IfModule prefork.c>
StartServers       8
MinSpareServers    5
MaxSpareServers    20
ServerLimit        256
MaxClients         256
MaxRequestsPerChild 4000
</IfModule>

and

<IfModule worker.c>
StartServers          2
MaxClients            150
MinSpareThreads       25
MaxSpareThreads       75
ThreadsPerChild       25
MaxRequestsPerChild   0
</IfModule>

These default settings are conservative and suitable for small‑to‑medium sites; higher‑performance environments require further tuning.

When to perform performance tuning?

Two phases: before launch (basic/experience tuning) and after launch (continuous monitoring and optimization).

Where to focus?

Hardware (CPU, memory, disk, NIC)

Operating system (processes, file system, kernel, etc.)

Applications (Nginx, MySQL, etc.)

Who should be involved?

Performance tuning is a team effort involving operations engineers, developers, QA, product, and monitoring staff—not just ops.

How to conduct performance tuning?

Define performance metrics (throughput, concurrency, response time).

Perform testing to verify metrics.

Analyze results to locate bottlenecks.

Apply targeted optimizations.

Monitor to validate improvements.

Performance metrics

Throughput – requests per unit time.

Concurrency – simultaneous users.

Response time – time from request to response.

Testing tools

Use appropriate load‑testing tools to measure the three metrics and generate reports.

Analysis tools

Hardware: vmstat, sar, iostat, netstat. OS: process, file system, swap, kernel parameters. Applications: mysqlreport, mysqlsla.

Optimization steps

Set clear tuning goals (e.g., 10‑20% improvement).

Understand the full stack and code base.

Make incremental changes, one parameter at a time.

Validate each change with tools such as ApacheBench.

Monitoring

Server‑level: CPU usage, load, memory usage, disk I/O, network traffic, disk space, processes. Service‑level (MySQL): query throughput, connection utilization, cache hit rates, index usage, lock statistics, etc.

Conclusion

The article offers a practical roadmap for performance optimization, emphasizing that with the right methodology anyone can tackle performance challenges effectively.