Server-side Optimization with Nginx and pm-static

This article details optimizing a multi-image gallery blog application, reducing its initial load time from a sluggish 28 seconds to a swift 0.7 seconds. The optimization journey focuses on on-the-fly thumbnail generation.

After optimization, production-ready speeds were achieved:

Key Improvements:

• Dramatic Load Time Reduction: Server-side optimization slashed initial load time from 28 seconds to 0.7 seconds.
• PHP-FPM Process Management: Adjusting PHP-FPM's pm setting (process manager) significantly impacted performance. While pm=static consumed more resources, it eliminated process spawning overhead, resulting in a 20% performance boost.
• Nginx and FastCGI Caching: Leveraging Nginx caching for both static and dynamic content drastically improved performance. Median request time dropped to 170 milliseconds, and failed requests plummeted from 17% to 0.53%.
• Resource Efficiency: The primary performance bottleneck wasn't hardware, even with modest resources.

Troubleshooting:

If using Homestead Improved on Windows, shared folder issues may arise. Adding type: "nfs" to the folder in Homestead.yaml often resolves this:

Run vagrant up with administrative privileges if problems persist. Before these fixes, load times were 20-30 seconds per request:

Testing Methodology:

Locust load testing was used with 100 concurrent users. The server stack comprised PHP 7.1.10, Nginx 1.13.3, and MySQL 5.7.19 on Ubuntu 16.04. Ngrok tunneled HTTP connections for testing via a static URL.

PHP-FPM and pm Setting:

The pm setting in /etc/php/7.1/fpm/pool.d/www.conf controls PHP-FPM process management. dynamic, ondemand, and static modes were tested. static provided the best performance but at the cost of higher resource utilization.

Nginx and FastCGI Caching Configuration:

Nginx caching was implemented using proxy_cache for static assets and fastcgi_cache for dynamic content. This significantly reduced response times and failure rates.

Pingdom testing confirmed the substantial performance improvements:

Conclusion:

This optimization demonstrated the significant impact of server-side tuning using Nginx caching and strategic PHP-FPM process management. The results highlight the potential for substantial performance gains even with modest server resources. A HAR file of the final test is available (not included here). Further optimization strategies are welcome.

(The initial image remains at the top, and all subsequent images maintain their original order and format.)

The above is the detailed content of Server-side Optimization with Nginx and pm-static. For more information, please follow other related articles on the PHP Chinese website!