CYBELE's final architecture

The core scientific and technological purpose of CYBELE is to serve as the enabler framework for Big Data, HPC and AI based algorithms that will be developed in order to construct different analytic processes embedded in end-to-end applications targeted at the agri-food domain by means of Directed Acyclic Graphs (DAGs). These graphs can be used to create structured pipelined workflows that can be instantiated on top of a programmable HPC Resource Abstraction Layer and a Distributed Execution Management Layer. As such, the analytic processes can be pre-categorized in three (3) categories which are: a)Micro-Batch and Batch Analytic Processes; b) Streaming Analytic Processes; and c) Distributed or Parallelised Machine Learning Processes for Deep Analysis.

The architecture is designed in a modular way facilitating maintenance, modifiability and extensibility, and can thus be easily extended and customised in order to include further end users’, farmers’, food growers’/producers’, data scientists’, agri-consultants and stakeholders’ needs not considered until now, including new inputs to reach different needs of interested parties, as well as new ones. It follows a layered approach which aims at ensuring interoperability among all involved components, putting emphasis on the way that pipelining of information (from data query to simulation formulation, to analysis and to visualization) is supported, safeguarding smooth interoperation of the supported services.

INTRASOFT led the work on the finalization of the CYBELE Architecture as well as the delivery of the APIS specifications which document how the components interact.

The core results documented are the validation of the APIs with the demo cases, the concretization of storage services to support auxiliary tasks and the development and the integration of CYBELE APIs under the umbrella of CYBELE Dashboard.