Skip to content

09 Cloud, Edge & Infrastructure (ATLAS)

1. Purpose

ATLAS is the compute substrate of ORION. It defines exactly where and how code executes, ranging from massive multi-region cloud clusters down to ruggedized battery-powered servers bolted into the trunks of responder vehicles.

2. Scope

Included: Cloud environments, edge hardware specs, containerization strategies, distributed compute topologies, and deployment models. Excluded: Software logic, database schemas, and NATS topic routing (defined in NEXUS and OMNIS).

3. Major Components

  • The Global Core (Cloud): The central aggregated state (AWS/GCP/Azure).
  • The Tactical Edge (Vehicles/Camps): High-compute physical nodes deployed in the field.
  • The Deep Edge (Sensors/Drones): Low-compute, low-power end devices.
  • Container Orchestrator: The runtime managing workloads across the spectrum.

4. Architecture

ATLAS employs a Heavy Edge / Decentralized architecture. Unlike traditional IoT where edge devices are "dumb" sensors reporting to a "smart" cloud, ORION edge nodes are fully autonomous micro-clouds capable of sustaining local operations indefinitely.

5. Responsibilities

Ensure that the ORION software stack (FastAPI, OPA, NATS, Postgres) has a deterministic, immutable execution environment regardless of whether it's running in Virginia or the middle of the Pacific Ocean.

6. Relationships with other ORION parts

ATLAS provides the physical and virtual compute for NEXUS (Architecture). It is physically transported by AEGIS (Comms).

7. Future Roadmap

Full Kubernetes (KubeEdge/K3s) orchestration stretching from the cloud down to the tactical edge nodes, allowing seamless live-migration of AI workloads to moving vehicles.

8. Trade-offs

A "Heavy Edge" requires buying expensive, ruggedized servers for every responder vehicle. We trade capital expenditure (CapEx) for absolute operational survivability.

9. Risks

Configuration Drift. If an edge node is offline for 3 months, its container images and security patches will be severely out of date when it reconnects.

10. Research Questions

How to automatically and securely pull 20GB of container image updates to a fleet of 500 edge nodes over a highly constrained satellite connection without suffocating SOS traffic?

11. Security Considerations

Edge nodes are physically exposed. ATLAS mandates TPM-backed full disk encryption and secure boot to ensure that a stolen edge node cannot be reverse-engineered or used as a trojan horse to breach the Supercluster.

12. Current Status

Phase 0/1: Docker Compose is standard for local and edge deployments. Kubernetes is explicitly deferred.