D4.1 – AI-BASED DATA OPERATIONS V1

This is the first of the series of deliverables related to the activities of WP4 ("AI-based Data Management for Green Data Operations"). Following the MobiSpaces Reference Architecture defined under the scope of T2.1 ("Design of Reference Architecture") and its current release reported in D2.1 ("Conceptual Model & Reference Architecture v1"), this document gives more details about one of the major architectural pillars, the AI-based Data Operations Toolbox. 

The overall activities conducted under the scope of WP4 ("AI-based Data Management for Green Data Operations") that are being reported in this document, mostly focus on this particular pillar, with the exception of the T4.4 ("Privacy-driven Data Aggregation"). The latter is considered as part of the Trustworthy Data Governance Services, however, the progress of this task is reported in these series of deliverables that summarize the activities of the whole WP4. In this document, we present the individual software components that are part of the AI-based Data Operations Toolbox, we give details of their interactions, the background technologies that these components are currently being built upon, along with more detailed description of their internal building blocks. 

WP4 focuses on both the data management aspects of MobiSpaces and the data operations of the platform in terms of automating the definition of AI workflows in a declarative manner and their corresponding runtime deployment and orchestration of their entire data lifecycle. The first category of components consists of the Data Management Toolset of the integrated solution that offers a variety of different but complementary data management systems to be exploited by the data users and application developers. For the second category of components, we provide the tools and algorithms for automating the definition and execution of complex AI workflows, consuming data from the aforementioned Data Management Toolset in a transparent manner. The target objective is to execute these workflows in an energy efficient manner, using our novel resource allocator to reduce the carbon emission. 

The duration of WP4 spans from M04 to M34. This deliverable reports the work that has been conducted until M10, which accomplishes the milestone MS04 ("Software prototypes - Iteration I"). At this phase of the project, we have identified the internal building blocks of the AI-based Data Operations Toolbox, the details of their interactions and we have delivered the first release of the corresponding prototypes. In this report our primary focus is on the individual evaluation of the components, while D2.7 ("AI-based Data Operations Toolbox v1") will later focus on the integrated solution based on our prototypes described here, to be evaluated by the project's use cases. Given the different maturity levels of the different components in WP4 at this moment, in this document we either provide some initial evaluation results or a concrete plan for evaluation that will be followed during the next period. Two additional versions are planned to be submitted in M22 and M34, where the second and third release of the prototypes will be available, giving more details of the implementation and final evaluation, implementing all target objectives of the WP4.