Evolution of UPYUN Cloud CDN Architecture from 2010 to 2016

Traditional CDNs saw little technical progress from 2000 to 2010, often relying on Squid and a limited number of nodes. With the rise of mobile internet, demand for faster acceleration grew, prompting UPYUN to develop its own cloud CDN system starting in 2010.

In 2010, UPYUN operated roughly 30‑40 edge nodes, using LVS for load balancing and deploying 5‑10 servers per node running ATS and Nginx. The system faced two main challenges: traffic statistics and acceleration rule distribution, leading to extensive Nginx C‑module development.

The architecture comprised three layers: about 40 edge nodes, four transit nodes for route optimization, and a data center in Hangzhou containing LVS, ATS, and cloud storage. Transit nodes reduced bandwidth pressure on origin servers by aggregating traffic.

2013 System Adjustments

All user configuration rules were centralized in the data center and pushed to edge nodes via LVS, eliminating the need for direct data‑center interaction. The first‑generation system was completely rewritten to accommodate large customers' custom requirements, reducing development cycles to about one week.

By 2013, edge nodes increased to 60, transit nodes to 8, and data centers to 2.

2014 Business Expansion

The focus shifted to enhancing transit nodes, adding an extra Nginx layer for better control, and supporting multi‑origin, multi‑IP, hot‑backup, load balancing, and multi‑line optimization. Edge nodes grew to 130, transit nodes to 16, providing roughly 1 TB of bandwidth redundancy.

2015 and Future Outlook

In 2015, UPYUN began international acceleration, establishing a Hong Kong‑to‑Zhejiang fiber link and planning backbone nodes in the US (Los Angeles) and France. By 2016, the network would include seven international acceleration nodes and three core backbone nodes.

Future plans include a proprietary DNS scheduling system to enable multi‑datacenter load balancing within provinces, enhanced DDoS/CC/WAF protection, and a stronger focus on video and live‑stream acceleration. Mobile SDK optimization promises 20‑30% performance gains without requiring UPYUN services.