Whether they’re hosted in the cloud or on-premise, AI microservices will be key for the broadcast business. But the degree to which an organization can benefit from these services depends on how well they’re integrated into operational workflows. This is where the open Media Cloud and Microservice Architecture (MCMA) comes in, writes EBU Principal Project Manager Jean-Pierre Evain.
18 months after its inception, the MCMA project now provides a common framework for the integration of cloud-based artificial intelligence microservices into media workflows, making a significant contribution the development of service-based architectures, and enabling one-click deployments of fully functional multi-cloud and AI-driven metadata-extraction workflows. The MCMA framework now also addresses security and authentication.
The Media Cloud and Microservice Architecture is based on the original FIMS (Framework for Interoperable Media Services), but adds a focus on:
- REST (Representational State Transfer)
- Artificial Intelligence (AI) cloud solutions
- the integration of AI metadata extraction tools in broadcast workflows, from ingest to data visualisation
The project is led by Bloomberg with key contributors from NHK, Triskel and Glookast, and the support of EBU Members such as VRT and YLE.
The MCMA approach consists of a lightweight high-level REST service interface, plus complementary code libraries for adaptation to the microservices of cloud AI platforms (such as AWS, Microsoft, IBM and Google) and their respective protocols and data structures. The implementation also extends existing cloud services to support media workflow functions such as the cataloguing of services, the monitoring of long-running processes and transactions, or job tracking.
Developer-friendly and open
The MCMA framework is developer-friendly, and the code is open-source and publicly available on Github (mcma-libraries and mcma-projects). The example project demonstrates how to set-up a complete cloud infrastructure across several providers with a single mouse click, using technologies like Gradle, Node.js and Terraform.
A typical integrated workflow was demonstrated on the EBU stand at IBC 2018: the example workflow began with a content ingest event, triggering a series of processes on cloud platforms of various vendors, extracting data with speech-to-text, translation and celebrity identification services, and making the results available via a front-end interface also developed as part of the project.
Hands-on MCMA showcase at PTS 2019
Participants at the EBU’s Production Technology Seminar 2019 will be able to get some first-hand experience with MCMA. Two tutorials have been organised to explain the fundamentals of MCMA and how to setup the simple framework needed to utilize one-click deployments of fully operational, multi-cloud and AI-driven metadata-extraction workflows.
For more information, visit the MCMA page and click on “join this group”.