The Streaming Lab investigates the performance, optimization, and resilience of large-scale video delivery infrastructures. Our work spans the full spectrum of data-driven content distribution—from transport-layer congestion control to end-to-end Quality of Experience (QoE)—combining empirical measurement, system engineering, and hardware-based experimentation.

We explore how modern CDNs behave under real-world load and how they can be optimized for live TV, OTT streaming, and on-demand services. Our research leverages production-level monitoring data, advanced analytics, and controlled emulation environments to understand where bandwidth is lost, where latency originates, and how system behavior impacts end users.

Key Research Themes

Transport-Layer Optimization for Video Delivery

Our group studies TCP congestion control algorithms (e.g., CUBIC, BBRv1/v3) and their influence on CDN performance and user experience. Using both server-side TCP statistics and large-scale client-side telemetry, we analyze throughput, latency, congestion window dynamics, and buffer events.
Recent experiments in an operational European IPTV network demonstrated that BBR delivers substantial gains in throughput, RTT reduction, and congestion window size—resulting in measurable improvements in QoE for both live TV and VOD scenarios.

Large-Scale Monitoring and Analytics

High-resolution telemetry from CDN servers and end-user devices enables us to model bottlenecks, detect anomalies, and validate optimization strategies. We employ Python-based analytics, Elasticsearch, InfluxDB, and visualization pipelines to reveal patterns across tens of thousands of streaming sessions.

Energy- and Bandwidth-Aware CDN Optimization

We develop system-level techniques to improve energy efficiency and reduce bandwidth waste. These efforts include performance modeling, congestion analysis, and real-world A/B testing with ISP partners.

Hardware-Based “Streaming Rig” for Reproducible Experiments

A distinctive asset of our research area is The Rig—a Raspberry Pi–based CDN emulation environment.
The Rig allows us to:

• emulate real CDN cache nodes and client devices,
• reproduce congestion behavior and streaming conditions,
• benchmark algorithms under controlled but realistic workloads,
• provide a hands-on platform for student projects and reproducible research.

This hardware-driven approach enables controlled experimentation that is impossible to achieve in production networks and complements our work with real ISP data.

Applied Edge Systems for Streaming

We investigate how lightweight edge nodes, micro-servers, and Raspberry Pis can support:

• adaptive bitrate logic,
• stream prefetching,
• caching strategies,
• and energy-efficient content delivery.

This work connects streaming research with future distributed computing and IoT architectures.

Technologies Used

Elasticsearch, InfluxDB, Docker, Docker Swarm, Linux, Python, Raspberry Pi, advanced monitoring tools, and custom client-side analytics agents.

Opportunities for Students

We welcome students interested in:

• streaming analytics and QoE modeling,
• congestion control and network protocol evaluation,
• CDN optimization and large-scale measurement,
• hardware tinkering and Raspberry Pi–based experimentation.

Contact us!

Please send an email to Imre Lendák for further inquiries.