Kamil Tokmakov (USTUTT), Ralf Schneider (USTUTT), Dennis Hoppe (USTUTT), Kalman Meth (IBM), Gad Maor (IBM), Elisabetta Di Nitto (POLIMI), Piero Fraternali (POLIMI), Rocio Nahime Torres (POLIMI), Saloni Kyal (POLIMI), Giovanni Quattrocchi (POLIMI)
Paul Mundt (ADPT), Yosu Gorroñogoitia (ATOS), Lucas Pelegrin (ATOS), Jorge Fernández (ATOS), Dragan Radolović (XLAB), Mihael Trajbarič (XLAB), Zoe Vasileiou (CERTH), Alfio Lazzaro (HPE), Indika Kumara (UVT/JADS)

In this deliverable, we report on the implementation progress in the second year of the project, its culmination in the second prototype of the SODALITE platform, and the evaluation of the release through a combination of technical KPI assessment and validation by the project’s three demonstrating use cases.

The key contributions and achievements with respect to the SODALITE platform are:

  • The introduction of the Edge as a unique infrastructure, as utilized by the Vehicle IoT UC. This has included the extension of the Cloud testbed to connect with a new Edge testbed in order to support Cloud-to-Edge deployments and experimentation. The Edge testbed includes a variety of hardware configurations and heterogeneous accelerators, managed as a self-contained Kubernetes cluster.
  • Quality controls for all software components have been implemented through SonarCloud, with each component required to pass a quality gate.
  • Development updates for individual SODALITE layers are presented below:
    • Semantic Modelling Layer: advanced features (e.g. design optimisation and Ansible models) were developed and released, the components were deployed and fully integrated. The SODALITE IDE has been improved for modules/extensions in the AADM, and extended to support the Ansible DSL, while the Semantic Reasoner and Semantic Knowledge Base have been extended with user management, support for TOSCA policies, and enriched context-assistance.
    • Infrastructure as Code Management Layer: During this period, some of the components developed during the first year were refactored and significantly improved. A new platform discovery service and application optimizer (MODAK) were released and integrated. IAM and secrets management have also been added.
    • Runtime Layer: Monitoring has been extended to include a visualization dashboard and alert manager, as well as Edge-based exporters. An alerting rules service has been developed to provide dynamic discovery of alerting rules. Metric exporters are dynamically discovered through Consul and Kubernetes.

The key contributions and achievements with respect to the evaluation of the use cases are:

  • Snow UC: All use case components were released as scheduled.
  • Clinical Trials UC: All use case components were released as scheduled. The original processing pipeline was extended by additional components.
  • Vehicle IoT UC: All use case components were released as scheduled. While the Y1 focus has been on developing use case components in Cloud, the Y2 focus has shifted towards Edge-based deployment through a Kubernetes cluster. Edge-based metric exports have been developed and tied into the SODALITE run-time monitor, allowing for direct reconfiguration of use case components at the Edge by the Cloud-based refactorer.