MYRTUS introduces advancements across its core pillars: the Reference Computing Continuum Infrastructure , MIRTO Cognitive Engine, and DPE , establishing new paradigms for compute-continuum systems.
Reference Infrastructure : MYRTUS redefines distributed computing with a composable, layered edge-fog-cloud continuum, integrating heterogeneous, autonomous, and federated nodes through seamless virtualization mechanisms. Apart from the advancements at the individual node level, including hardware acceleration via HMPSoCs-based platforms and RISC-V, unlike siloed or vendor-locked solutions, MYRTUS ensures strong interoperability, robust three-level security, and AI-optimized energy efficiency, leveraging Liqo, an innovative Kubernetes-based solution, to allow cross-layer virtualization. MYRTUS aligns with EUCEI initiatives, offering a flexible and adaptable foundation for dynamic WLs/resources management.
MIRTO Cognitive Engine: MIRTO significantly advances compute-continuum orchestration through a customizable decentralized strategy driven by distributed cognitive agents. Leveraging autonomic principles, this engine provides intelligence via continuous feedback, learning, and self-adaptation, enabling informed decisions on resource allocation/customization, task scheduling, and system reconfiguration in complex, dynamic environments. Customized management leverages a modular design strategy that allows opt for different AI-powered WL optimization technologies, including swarm-based and deep-reinforcement learning based strategies, expanding optimization drivers beyond existing frameworks, explicitly integrating privacy, security guarantees, and energy efficiency alongside performance.
MYRTUS DPE: MYRTUS DPE revolutionizes software development for computing continuum with end-to-end tooling and interoperable interfaces, overcoming current fragmented toolchains. Built upon OASIS TOSCA and the MLIR framework, it ensures robust, modular, and cross-platform design.
Within this unified framework, tools like Modelio TOSCA Designer, uniquely integrated with ADT Designer for security-by-design threat modeling, enable early-stage application design and high-level exploration and prospectively will also automated code generation for decentralized swarm-based agents and and countermeasures synthesis, largely, simplifying a traditionally manual process. Moreover, the DPE leveraging an MLIR-based compiler supports a diversity of high-level abstraction inputs targeting a plethora of different hardware platforms, including customizable FPGA-based ones, overcoming traditional vendor lock-in.
