Track: WebRTC and Real-Time Applications Programmable Real-Time Networks and Applications Emerging Technologies and RTC |
| Scaling WebRTC Globally: GPU-Powered Docker & Kubernetes Strategies |
| Real-time, high-quality video in WebRTC applications is increasingly in demand across industries—from telehealth consultations to live education, broadcasting, gaming, and emergency communications. But scaling WebRTC across regions while having low-latency is one of the toughest problems developers and architects face. Traditional WebRTC setups often struggle with resource bottlenecks, unpredictable traffic spikes, GPU dependency management, and monitoring across distributed environments. This session introduces a cloud-native, container-based architecture for scaling WebRTC applications using Docker, Kubernetes, and GPU-powered media services. Attendees will learn how to containerize GPU-enabled media workloads such as real-time background blur, live language translation, and AI-driven video enhancements. We’ll explore how to deploy these services on multi-region Kubernetes clusters, taking advantage of autoscaling, GPU scheduling, and traffic routing to maintain high-quality performance under load. We will also cover observability and resilience strategies, showing how to integrate Prometheus, Grafana, and OpenTelemetry into the pipeline to monitor session quality, latency, and system health in real time. Security and compliance considerations—such as isolating workloads, handling sensitive video streams, and using policy-as-code for runtime governance—will also be addressed. A live demo will showcase a Docker-crafted WebRTC service running on Kubernetes, scaling dynamically under traffic load while maintaining low latency and quality of experience. By the end of this session, attendees will walk away with code-ready patterns, sample Dockerfiles, Kubernetes manifests, and architectural insights they can immediately apply to build production-grade, AI-enhanced WebRTC systems that run reliably in the cloud and at the edge. |
|
| Presentation Video |