Asynchronous Broadcast Media Processing

Executive Summary

• An asynchronous processing architecture using the Matrox ORIGIN framework is described for live media workflows, positioned against limitations of synchronous, clock-driven processing in distributed cloud deployments.
• The architecture is implemented on AWS using a set of managed and infrastructure services, and it is presented as capable of scaling live production workloads by adding compute nodes to a Kubernetes cluster.
• A Matrox-tested distributed live mixer application is reported to scale from 10 concurrent 1080p50 live inputs to over 110 concurrent uncompressed video inputs through cluster expansion.

Key Industry Developments

• Shift from synchronous to asynchronous processing for distributed live media
• Synchronous, clock-driven processing is described as having limitations in distributed cloud deployments, including vertical scaling only and tight coupling between components.
• Matrox ORIGIN is described as implementing an asynchronous processing model that decouples control from compute, aiming to reduce coupling in live media pipelines.
• Three-part framework model for live media workflows
• Matrox ORIGIN is described as using three components: Media Services, Control Tracks, and Media Fabric.
• The framework is described as following a stateless architecture approach, using control tracks to declare stream characteristics and desired state.
• Deterministic timing model for frame-accurate coordination
• The system is described as using a deterministic timing model based on grain intervals and sequence numbers mapped to an absolute time reference.
• This timing approach is presented as a mechanism to coordinate media processing without relying on tightly coupled, synchronous execution.
• AWS-based implementation pattern for broadcast-quality scale
• The implementation described uses Amazon Elastic Kubernetes Service (Amazon EKS) and Amazon EC2 for compute and orchestration.
• It also includes Elastic Fabric Adapter (EFA), Amazon Time Sync Service, Elastic Load Balancing, Amazon CloudWatch, and Amazon Managed Grafana as part of the system’s networking, timing, traffic distribution, and observability stack.

Real-World Use Cases

• Cloud-based live production workflows
• The architecture is described for live media workflows in the media and entertainment industry, using the Matrox ORIGIN framework’s asynchronous model to structure processing across distributed resources.
• Distributed live mixer scaling with uncompressed inputs
• A Matrox-tested distributed live mixer application is reported to scale from 10 concurrent 1080p50 live inputs to over 110 concurrent uncompressed video inputs by adding compute nodes to an Amazon EKS cluster.
• Cross-Availability Zone resilience for frame publishing and consumption
• Resilience mechanisms are described, including redundant producers across Availability Zones and loosely coupled component lifecycles, to support continued operation when components fail or are updated.
• High-bandwidth uncompressed workflows
• Uncompressed 4K workflows are described as requiring up to 100 Gbps bandwidth, aligning with the inclusion of high-performance networking components in the referenced AWS implementation.

Why It Matters

• Decoupling control and compute changes scaling and operations
• The asynchronous model is described as decoupling control from compute, which directly addresses the stated limitations of synchronous, clock-driven processing in distributed cloud deployments (vertical scaling constraints and tight coupling).
• Deterministic timing supports coordinated processing without tight coupling
• A deterministic timing model based on grain intervals and sequence numbers mapped to an absolute time reference provides a concrete mechanism for coordinating distributed media processing while maintaining a defined temporal structure.
• Stateless design and declared desired state support resilience workflows
• The stateless architecture approach, using control tracks to declare stream characteristics and desired state, aligns with operational patterns such as rolling updates and fault recovery without system-wide disruption, as described in the use cases.
• Reference implementation ties architecture to deployable AWS services
• The described system maps the framework to specific AWS services (EKS, EC2, EFA, Amazon Time Sync Service, Elastic Load Balancing, CloudWatch, and Managed Grafana), providing an example of how timing, networking, scaling, and observability are assembled for broadcast-oriented workloads.

Sources

• https://aws.amazon.com/blogs/media/asynchronous-media-processing-on-aws-achieving-broadcast-quality-at-scale/