Amagi metadata graph architecture achieves sub-500ms latency for 2,500 channels
Amagi Media Labs has implemented a hybrid metadata architecture using Amazon Neptune for graph-based relationship modeling and Amazon OpenSearch Service for high-frequency transactional reads. The system manages over 500 million triples across 2,500 channels to enable sub-500ms query performance for complex media supply chain operations.
Key Takeaways
- The Global Metadata Store manages over 500 million triples and one million media assets using an RDF-based knowledge graph.
- A custom synchronization layer resolves non-persistent blank nodes from Neptune to OpenSearch to maintain sub-500ms read latency.
- The architecture supports peak loads exceeding 1,000 requests per second for high-frequency scheduling and distribution APIs.
- Automated rights inheritance allows localized clips and trailers to inherit restrictions from parent series without manual reconciliation.
Why It Matters
The shift from relational joins to a graph-based model addresses the increasing complexity of media metadata, where a single asset may have hundreds of regional licensing and talent credit dependencies. By decoupling deep semantic queries in Neptune from high-speed transactional reads in OpenSearch, Amagi provides a blueprint for managing massive content libraries without the performance degradation typical of SQL-based systems. This infrastructure allows for rapid ingestion of AI-generated metadata and real-time rights inheritance, which is critical as streaming platforms move toward more granular, automated distribution models. Watch for whether Amagi integrates ML-powered metadata enrichment to further automate the tagging of live sports and news highlights.
Additional Context
Amagi Media Labs has been expanding its cloud-native media operations platform across multiple AWS services beyond the Neptune and OpenSearch combination described in the Global Metadata Store. In a recent blog post, Amagi detailed how it uses AI tools for scene-level metadata extraction, including speech-to-text and credit parsing from end titles, to enrich content descriptions across its FAST channel portfolio. This metadata enrichment pipeline feeds directly into the kind of graph-based relationship modeling that Neptune handles, suggesting the GMS architecture is designed to absorb AI-generated metadata at scale without manual tagging bottlenecks.
The polyglot persistence approach Amagi adopted, pairing a graph database for deep semantic traversal with a search engine for high-throughput reads, reflects a pattern AWS has promoted across its media and entertainment reference architectures. AWS has published guidance on using Amazon Neptune to store subject-relationship-object tuples, a workflow that mirrors the RDF triple ingestion Amagi performs at production scale. The key difference is that Amagi's implementation operates at 500 million triples across 2,500 live channels rather than as a reference demo, making it one of the largest documented Neptune deployments in the media vertical.
For streaming platform operators evaluating similar metadata architectures, the synchronization layer between Neptune and OpenSearch that Amagi built on Amazon EKS represents a critical engineering decision. The custom application intercepts Neptune Stream events, resolves blank nodes to their nearest named parent URI, flattens the data into structured JSON, and indexes it into OpenSearch for sub-500ms transactional reads. This pattern of graph-to-search synchronization is not unique to Amagi. AWS has demonstrated building RDF knowledge graphs on Neptune, though those examples target smaller-scale discovery use cases rather than the real-time scheduling and distribution APIs that Amagi's 800 content brands depend on.
Companies in this story
In short
Amagi Media Labs has deployed a new metadata graph architecture using Amazon Neptune and OpenSearch to manage 500 million triples across 2,500 channels. By replacing traditional relational databases, the system achieves sub-500ms read latency, enabling efficient management of complex media supply chain relationships and automated rights inheritance for global content distribution.
FAQ
What technologies power Amagi's new metadata architecture?
Amagi uses a hybrid approach combining Amazon Neptune for deep semantic graph storage and Amazon OpenSearch for high-speed transactional reads, synchronized via a custom layer on Amazon EKS.
How many channels does the Amagi metadata system support?
The architecture is designed to support 2,500 live channels and manages over 500 million triples and one million media assets.
Why did Amagi move away from relational databases?
Relational databases struggled with the complexity of modern media metadata, where assets have numerous regional licensing and talent dependencies. The graph-based model allows for better semantic queries and real-time rights inheritance.
What is the benefit of the synchronization layer?
The custom synchronization layer resolves non-persistent blank nodes from Neptune and flattens data into JSON for OpenSearch, enabling sub-500ms read latency for scheduling and distribution APIs.
Read full article at aws.amazon.com
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source