Integrating a Stream Transformation Engine in the Distribution Pipeline for Next-Gen Streaming Efficiency

This paper proposes a distributed stream transformation architecture that integrates processing capabilities directly into IP-based media transport networks. As media distribution has transitioned from satellite to IP, solutions have been split between dedicated hardware and cloud platforms, each with scalability limitations. Our approach introduces a hardware-agnostic transformation framework supporting graphics processing unit (GPU), vision processing unit (VPU) (field-programmable gate array/application-specific integrated circuit (FPGA/ASIC)), and central processing unit (CPU) accelerators through unified Ku-bernetes orchestration, employing automatic repeat request (ARQ)-based protocols—Reliable Internet Stream Transport (RIST) and Secure Reliable Transport (SRT)—for reliable delivery over unmanaged networks. The architecture features a failure-tolerant control plane separated from the transport layer and a unified transformation engine running across high-performance core nodes and resource-constrained edge devices. Composable transformation pipelines eliminate duplicate processing by collocating multiple output requirements on optimal nodes. Cost analysis demonstrates that distributed commercial off-the-shelf (COTS) edge processing reduces per-stream costs from approximately ${\$}$281/month to ${\$}$97/month compared to managed cloud services over 36 months, supporting a hybrid deployment model.

Print ISSN

1545-0279

Electronic ISSN

2160-2492

Published

2026-04

Content type

Original Research

Keywords

distributed transcoding, edge computing, rist, srt, arq, kubernetes, gpu acceleration

DOI

10.5594/JMI.2026/YPFV5033