How fluidos-project/fluidos-edge works
Project type and purpose
• This is a full-stack IoT Edge system (the “Fluidos Edge” project) that spans from low-level firmware on a sensor board, through edge-side microservices, to Kubernetes-based cloud/edge control planes.
• Purpose: manage and orchestrate edge nodes and devices, collect sensor data (via BLE/MEMS), run mapping/transformation services, and expose/control everything via CloudCore/EdgeCore on Kubernetes.
File‐structure analysis
• Top level:
• manifests/ – sample Kubernetes CRDs and Deployment YAMLs for CloudCore/EdgeCore, mapper rollouts, router/policy
• sw/ – submodules for edge system software components:
– FluidosEdge-System (CloudCore, EdgeCore, NodeCore etc.)
– FluidosEdge-Mappers (sensor-to-API translation)
– FluidosEdge-REAR: two Go microservices (connection_manager and edge-resource-manager)
• testbed/ – scripts and YAMLs to spin up a Kind cluster for local testing of CloudCore/EdgeCore and node deployments
README insights
• Refers to a multi-layered Edge architecture:
– Physical devices → Edge microservices → CloudCore/EdgeCore (K8s)
– Illustrates Fluidos Edge layers: Device layer, NodeCore/EdgeCore layer, CloudCore layer
• Describes installation of Kubernetes, CloudCore, EdgeCore, and sample deployments
Main components
a. Device layer:
b. Edge node software:
• connection_manager (discovers/manages BLE devices)
• edge-resource-manager (manages compute/storage resources on node)
• mappers (FluidosEdge-Mappers) that subscribe to device data and translate to Fluidos data models
• NodeCore (from FluidosEdge-System) that glues together connection_manager, resource-manager, mappers
c. Kubernetes control plane:
• EdgeCore (runs on edge clusters, orchestrates NodeCore, sidecar pods)
• CloudCore (runs on cloud cluster, central control, stores state)
d. External systems:
• etcd/DB (CloudCore state store)
• CLI/UI (kubectl or Fluidos UI)
Relationships and data flows
• Firmware → (BLE) → connection_manager
• connection_manager → publishes events to mappers or NodeCore event bus
• mappers → transform sensor data → NodeCore API
• NodeCore → EdgeCore CRDs (via Kubernetes API)
• EdgeCore ↔ CloudCore sync (object sync CRDs)
• CloudCore → stores global state, provides UI/REST API
Architectural patterns/design principles
• Microservices on edge nodes (connection_manager, resource-manager, mappers)
• Kubernetes CRD-based control plane (CloudCore/EdgeCore)
• Event-driven data flow (BLE events → message bus → mapper → API)
• Separation of concerns: firmware vs. edge services vs. orchestrator vs. cloud
Technologies, frameworks, libraries
• Firmware: STM32Cube HAL, CMSIS, BlueNRG BLE stack, X-Cube-MEMS middleware
• Edge services: Go modules (go.mod), standard library, custom Fluidos utilities
• Orchestration: Kubernetes, CRDs (apps_v1alpha1_edgeapplication, nodegroup, device models, router rules, etc.)
• Testing: Kind cluster, Helm charts, sample YAMLs