The TaRDIS project has developed and demonstrated a powerful prototype for decentralized monitoring and configuration management in swarm-based systems. This milestone showcases an innovative system architecture, novel APIs, and runtime configuration strategies—all supporting the dynamic and intelligent orchestration of distributed edge applications.
Overview
The goal of this task is to design and implement solutions for:
- Acquiring telemetry from deployed SIEM components
- Aggregating and propagating telemetry data for ML model training using both centralized, and peer-to-peer approach
- Runtime reconfiguration of application components (e.g., changing patterns, modifying communication paths, removing or redeploying components) using both direct, and peer-to-peer approach
The system is structured into two core elements:
- A Control Plane, acting as the brain of the system—managing orchestration, logic, and user interaction
- A Swarm/Data Layer, representing the distributed nodes where applications and telemetry sources reside
The entire system is based on open-source tools and emphasizes the use of OpenMetrics standards for interoperability and extensibility, but also emphasizes cloud-edge collaboration.
Key Features and Architecture
Users interact with the system via a Command-Line Interface (CLI) tool that communicates with the control plane to send commands, request data, or trigger reconfigurations.
Key architectural elements include:
- Label-based object management using arbitrary key-value pairs, for querying purpose
- Loosely coupled interactions between system components, using labels
- Namespace management for logical isolation and hierarchical organization of swarms
- Resource quota and security profile management per namespace
What’s Next
Looking ahead, we will work on peer-to-peer metrics aggregation, continue to expand integration points with other work packages, enable more sophisticated reconfiguration capabilities, support real-world deployment scenarios of decentralized edge intelligence and facilitate collaboration with AI model training and inference pipelines.
A live demo of the system was presented, showcasing the interaction between CLI, APIs, the control plane, and the swarm layer—highlighting the prototype’s flexibility and real-time control capabilities.
For more details about this task and other innovations within the TaRDIS project, stay tuned to our news section and follow our progress on tardis-project.eu.
Watch demo on our YouTube channel or here below: