Scalable distributed repository framework for digital media production workflows


The work on the scalable distributed repository framework in TOSCA-MP developed an integrated repository for multimedia production systems. This integrated repository provides different performance characteristics, data structures and interfaces for different steps in media production workflows.

In the area of distributed processing and distributed access to repositories the distribution of requests to a multimedia repository becomes more and more important. Relational databases provide cluster solutions for that, but they are not basically designed for such distributed tasks. Therefore, TOSCA-MP investigated different approaches of different types of NoSQL databases to realize a true distributed repository. Characteristics like high performance, scalability, distributed execution and changeable data structures make NoSQL databases very attractive for the use in a distributed repository. In the focus of the investigations were object and graph oriented database structures. In addition, TOSCA-MP investigated the applicability of easy extensibility of the repository framework towards new data sources. E.g. domain-specific knowledge bases for particular event types (e.g. a football match) could improve the results of task-adaptive search and browsing.

A big problem in state of the art repositories is that although a lot of the included information can be linked together, no specific data are provided how relevant the stored information is in connection with the task at hand. TOSCA-MP provides data models and specialised repositories to solve this issue.

Furthermore, state-of-the-art data models in multimedia repositories are inflexible in allowing concise domain-specific annotation, such as descriptions of football games, news shows or musicals. Storing information of such diverse descriptions in a single repository requires a broad data model, which does not provide the best retrieval effectiveness of audio visual content retrieval compared to domain-specific ontologies.

Today, no frameworks are available which allow us to connect many different types of repositories in order to develop a single repository solution from the perspective of the applications. The scalable distributed repository framework developed in TOSCA-MP serves as the basis for the connection of various specialized repositories. It provides suitable interfaces for all processing steps and is able to serve all requirements of a distributed network based metadata workflow in the media industry.

© 2017 TOSCA-MP - Task-Oriented Search and Content Annotation for Media Production
The research leading to the presented results has received funding from the European Union's
Seventh Framework Programme (FP7/2007-2013) under grant agreement n° 287532. - Imprint