TGW - Operating an Efficient and Resilient Cloud Gateway at Scale

Learn about the design, deployment, and operation of Tencent Gateway (TGW), a large-scale cloud gateway system engineered to handle the demanding requirements of online gaming and live streaming services. This 18-minute conference presentation from USENIX ATC '25 explores how TGW addresses the critical challenges of efficient packet forwarding, scalable state management, and high resilience in cloud data center environments. Discover the system's highly decoupled and modular architecture, featuring efficient forwarding planes, scalable state migration mechanisms, resilient failure recovery systems, and comprehensive failure detection and localization capabilities. Examine the performance achievements including 2.9× improved forwarding capacity compared to prior systems, 4-second state and traffic migration between clusters without packet loss, and the ability to handle tens of Tbps of traffic with packet drop rates ranging from 10-7 to 10-4. Gain insights from eight years of real-world industrial deployment experience, including strategies for maintaining 100% availability while serving the stringent latency, jitter, and packet loss requirements of Tencent Cloud's gaming and streaming workloads.