Intel-Switch: Optimized Adaptive Protocol Selection Mechanism for Video Streaming Application

With the introduction of QUIC (HTTP/3), a recent transport layer protocol QUIC (HTTP/3) has gained popularity for its better performance in challenging network conditions than TCP. Meanwhile, TCP (HTTP/2) often performs better than QUIC (HTTP/3) in high-bandwidth networks. Several prior works have proposed adaptive transport protocol selection mechanisms, but they rely on outdated machine learning models and only focus on web services, and none of them consider video streaming applications. Moreover, switching protocols during video streaming can degrade the Quality of Experience (QoE) due to the loss of congestion state. This work focuses on proposing an adaptive framework (Intel-Switch) for seamless protocol switching for services like video streaming. It includes a protocol selection mechanism for evolving network conditions, an optimized ABR algorithm on the face of protocol switching to minimize the impact on video QoE, and an eBPF-based mechanism to transfer the congestion state at the server side. Together, these innovations aim to enhance video streaming performance in dynamic environments.