Work package 7

Objective: Resource Management

Task 7.1 – Modeling Compute Continuum Infrastructure Resources

Establish abstract infrastructure models for the Resource Ensemble Manager (REM) to represent diverse cloud–edge continuum resources for complex data analytics.
Using frameworks from WP4 (T4.1) a cloud-native APIs (extending Kubernetes API), the models define:

  1. Cloud and edge resources, including constraints, energy use, and carbon footprint.

  2. On-premise edge resources for managing COTS hardware.

  3. SLOs and SLIs for SLA management.
    These models align with provisioning and adaptation systems (T7.2) to support dynamic workload management.

Task 7.2 – Compute Continuum Provisioning and Adaptation 

Develop provisioning and adaptation techniques for infrastructure resources within the cloud–edge continuum to support the REM (T7.1).
Using GitOps for infrastructure synchronization, and open-source tools like Cluster API, Crossplane, and Karmada, this task enables dynamic deployment, configuration, and scaling.
A service mesh management system will ensure seamless interconnection in hybrid edge–multi-cloud environments, enabling flexible, scalable, and adaptive resource provisioning for complex data pipelines.

Task 7.3 – Compute Continuum Access Control, Monitoring, and Holistic Explainability

Create a unified access control and monitoring interface across edge–cloud providers, following Zero Trust principles for secure, privacy-aware operation (aligned with T3.2).
The task provides trusted Compute Continuum Resources a runtime transparency for data pipelines using insights from the Unified Knowledge Layer (T4.1, T4.2).
Focus areas include energy consumption, trustworthy execution, data reliability, and performance, supported by ML-based monitoring to detect anomalies and ensure regulatory compliance.

Task 7.4 – Auction-based Models and Algorithms for the XaaS Marketplace 

Develop an on-demand XaaS Marketplace for complex data analytics, integrating cloud and edge services through low-code/no-code tools for providers and agents.
Leveraging WP6 results and the Smart Data Management layer, the Marketplace uses semantic annotations from UKL (T4.1) to enhance service discovery and matchmaking.
It supports real-time agent-based auctions for resource allocation via the REM (WP7), ensuring resilience through distributed runtime deployments and agent restart strategies.
Additionally, it introduces swarm models (homogeneous, heterogeneous, meta-swarms) for collaborative and adaptive agent coalitions.

Návrat hore