Uploaded June 2026 | Updated September 2026, 3 weeks ago
HyperEdge: An Edge CDN Infrastructure for Cost Efficient Video Streaming
Dehui Wei, National University of Singapore; Jiao Zhang, Beijing University of Posts and Telecommunications, and Purple Mountain Laboratories; Haozhe Li, Rui Han, Zhichen Xue, Yajie Peng, Xiaofei Pang, and Yan Ma, ByteDance; Jialin Li, National University of Singapore
Awarded Outstanding Paper!
As ByteDance’s business expands, the cost of video streaming using content delivery networks (CDN) has become prohibitively high. We have discovered a sea of under-utilized edge devices with the potential to reduce content distribution cost. The unreliable performance of an edge network, however, presents deep challenges to video streaming services. In this work, we introduce HyperEdge, an edge-assisted content delivery system for video streaming. HyperEdge seamlessly integrates the robustness of a conventional CDN with the cost-efficiency of an edge network. It offers dependable streaming quality to users while minimizing traffic expenses. HyperEdge employs a centralized tracker cluster to optimize content distribution to a pool of edge devices, based on real-time monitoring. To ensure satisfactory video playback quality, we develop a novel multi-path protocol for client-edge video transmission. Having been in stable operation for six years, HyperEdge manages over a hundred thousand edge devices, serving about a hundred million users daily, and saving hundreds of millions of dollars in content delivery cost annually.
View the full NSDI '26 program at usenix.org/conference/nsdi26/technical-sessions
HyperEdge: An Edge CDN Infrastructure for Cost Efficient Video Streaming
Dehui Wei, National University of Singapore; Jiao Zhang, Beijing University of Posts and Telecommunications, and Purple Mountain Laboratories; Haozhe Li, Rui Han, Zhichen Xue, Yajie Peng, Xiaofei Pang, and Yan Ma, ByteDance; Jialin Li, National University of Singapore
Awarded Outstanding Paper!
As ByteDance’s business expands, the cost of video streaming using content delivery networks (CDN) has become prohibitively high. We have discovered a sea of under-utilized edge devices with the potential to reduce content distribution cost. The unreliable performance of an edge network, however, presents deep challenges to video streaming services. In this work, we introduce HyperEdge, an edge-assisted content delivery system for video streaming. HyperEdge seamlessly integrates the robustness of a conventional CDN with the cost-efficiency of an edge network. It offers dependable streaming quality to users while minimizing traffic expenses. HyperEdge employs a centralized tracker cluster to optimize content distribution to a pool of edge devices, based on real-time monitoring. To ensure satisfactory video playback quality, we develop a novel multi-path protocol for client-edge video transmission. Having been in stable operation for six years, HyperEdge manages over a hundred thousand edge devices, serving about a hundred million users daily, and saving hundreds of millions of dollars in content delivery cost annually.
View the full NSDI '26 program at usenix.org/conference/nsdi26/technical-sessions
