The project's technical team gathered in Athens at the end of January 2020 for a two-day meeting. The goal was to formulate a common technical vision and approach, take stock of the available components, understand what else might be needed to realise the overall project ambition, and foresee any challenges that could hamper progress.
DUET technical meeting underway in Athens
Much of the discussion centered around DUET's ultimate solution i.e. a digital twin platform. Partners agreed it should be modular and consist of multiple layers.
Infrastructure Layer: Corresponds to a cloud-based infrastructure with support for the High Performance Computing (HPC) workloads. Main considerations include
Development and production of HPC and cloud environment (one or more servers per environment), with some components running on partner’s server
CI/CD with Gitlab for open sourced components (Deployment Templates required). Other tools for proprietary code in partner’s premises
Docker with Docker compose or Kubernetes
Singularity for HPC components
Data Layer: Refers to the repositories/databases of the platform where each database is deployed in a distributed mode that spans multiple nodes in the cloud cluster for enhanced scalability, availability and performance. Two main considerations are
Databases: IoT data, Open Data, content from social media and blogs, models, apps and pilot data
Possible database servers: MongoDB, MySQL, PostgreSQL, ElasticSearch, Neo4j
Business Layer: Corresponds to the processing components that implement business requirements of the platform. All components will expose a REST API for consumption by the visualization/UI components. Key considerations include
Modelling (Data Modeler, Data Process Engine, Simulation Engine)
A.I. (NLP Engine, Sentiment Analysis)
Gamification (Gamification Engine)
Semantic (Semantic Enricher)
Access Control Layer: Responsible for aggregating data from various external sources and exposing a unified API to all Open Linked Data in the repository. The IoT API component facilitates the ingestion of new data pipelines from sensors and other sources. Key services include
Presentation Layer: Provides the interfaces between the systems and the user. This requires
Dashboards per role encapsulating the various UI components
React JS for implementation
D3 and Resium for visualisations
Kibana and Grafana for monitoring
One of the main takeaways from the meeting is that DUET should act as a data broker relying on Open Data compliant standards e.g. OGC sensorthings. For better performance, caching adapters may be added while random data access from models should be avoided. Since pretty much everything these days is event based, plugins/adapters can facilitate the replay of time series by publishing content to the event broker, as shown in the T-CELL concept below.
High-level overview of (DUE)T-CELL Concept
Towards the end of the meeting, partners identified and agreed on the components that need to be implemented and/or adapted in the first year of the project. They had a better sense of each other's responsibilities and how the different activities, tasks and outputs align with one another. Importantly, they introduced and agreed on the collaboration procedures to determine key priorities for the first phase, and thrashed out a detailed plan for the next six months.
To keep abreast of future technical and other developments, make sure to subscribe to our newsletter and follow DUET on Twitter (@DuetH2020).