D2.1 - Decentralised Orchestrator Early Release

This document reports on the results of the research and technical development activities conducted under Work Package 2 (WP2) of the Swarmchestrate project that aim to empower decentralized, smart, and sustainable application orchestration in a wide range of Cloud-to-Edge settings. This document introduces the key architectural designs and implementation choices on decentralised orchestration, context-aware semantics, adaptive monitoring, and AI-driven energy optimisation.

The deliverable describes the design of a TOSCA-based semantic model that enables rich, machine-interpretable application descriptions. These models capture not only structural and resource requirements, but also contextual dependencies, quality-of-service objectives, and monitoring specifications and constraints which are crucial for informed, autonomous decision-making across heterogeneous infrastructure. From this, the system enables applications to deploy, using a peer-to-peer orchestration layer built from resource agents that collaboratively discover and provision available resources. The deployment of the application, swarm formation, as well as the runtime management along with different types of potential reconfigurations are thoroughly detailed. 

Another portion of the work addresses the delivery of a decentralised monitoring framework, the Event Management System (EMS), with a context-aware metric model, which supports both raw and composite metrics, and integrates mechanisms for defining service-level objectives as enforceable constraints. The EMS provides the feedback loop necessary for intelligent orchestration, supporting reconfigurations and self-healing through a decentralised complex event processing system.

In parallel, WP2 introduces the first iteration of Swarmchestrate’s energy optimisation framework, which integrates ecological considerations into the orchestration loop. By embedding environmental cost as a first-class concern in resource selection and task allocation, the orchestrator could balance performance and sustainability in real time.

Taken together, these results constitute the first fully integrated release of Swarmchestrate’s decentralised orchestrator. The outputs of this work package include an extended TOSCA specification tailored to swarm-based application modelling, a functioning prototype of the orchestration and monitoring subsystems, and a validated approach to energy-aware deployment. These developments establish the architectural and semantic backbone upon which subsequent work packages will build more advanced features in monitoring, optimisation, intelligence and full lifecycle management of application.