Edge Computing Devices and Nodes

Edge computing runs on real hardware — physical machines that sit in factories, hospitals, vehicles, and street cabinets. Knowing the different device types helps you match the right hardware to the right job.

The Hardware Spectrum

Less Power ◄─────────────────────────────────────► More Power

[Microcontroller] [SBC] [Edge Gateway] [Edge Server] [Micro Data Center]
     Arduino     Raspberry Pi  Cisco IE  Dell PowerEdge  Pre-built rack
     $5–$30      $50–$100     $200–$2K    $2K–$20K        $50K+

Each tier handles different levels of data processing. Simpler devices collect and pass data. More powerful devices analyze, decide, and manage.

Tier 1: Microcontrollers

A microcontroller is a tiny computer on a single chip. It reads sensor inputs and sends simple signals. It has almost no storage and very limited processing power.

Real Example: Arduino in a Greenhouse

[Soil Moisture Sensor] ──► [Arduino Microcontroller]
                                     │
                         reads value every 10 minutes
                                     │
                         if moisture < 30% → turn ON water pump
                         if moisture ≥ 30% → keep pump OFF

The Arduino makes a yes/no decision locally with no internet. It costs $10 and runs on 5 volts. This is the simplest form of edge computing.

Tier 2: Single-Board Computers (SBC)

A single-board computer fits everything — CPU, memory, storage, and networking — on one small circuit board. It runs a full operating system and can execute real applications.

Popular Examples:

• Raspberry Pi 5: Used in retail kiosks, environmental monitors, and small AI inference tasks
• NVIDIA Jetson Nano: Designed for computer vision tasks — it can analyze camera feeds and recognize objects locally
• Google Coral Dev Board: Specialized for running AI models at low power

An SBC running a face recognition model at a building entrance processes 30 video frames per second locally. No cloud connection is needed to identify authorized staff.

Tier 3: Edge Gateways

An edge gateway is an industrial-grade device built to survive harsh environments — extreme temperatures, dust, vibration, and humidity. It connects multiple sensor types and runs filtering and routing software.

Gateway Use Case: Oil Platform

200 sensors (pressure, temperature, gas, flow) speaking 6 different protocols
        │
        ▼
[Industrial Edge Gateway]
  - Converts all 6 protocols to one standard (MQTT)
  - Applies safety thresholds
  - Stores 7 days of data locally
  - Sends alerts via satellite internet when triggered
        │
        ▼
  [Offshore Cloud Dashboard]

The gateway keeps the platform operating safely even when satellite connectivity is intermittent.

Tier 4: Edge Servers

An edge server is a full server — just deployed locally instead of in a remote data center. It runs multiple applications simultaneously, handles heavier AI workloads, and manages all gateways and devices in its area.

Form Factors:

• Rack-mount servers: Sit in standard server racks inside facilities
• Ruggedized servers: Built to survive vehicle vibration, used in trains, trucks, or military vehicles
• Mini PC servers: Compact units that fit inside street cabinets for smart city deployments

Tier 5: Micro Data Centers

A micro data center is a self-contained, climate-controlled cabinet that fits in a shipping container or a room. It houses multiple edge servers, networking equipment, UPS (battery backup), and cooling — all in one sealed unit.

Telecom companies deploy micro data centers at the base of 5G towers to bring powerful computing within 10ms of millions of mobile users. Retailers install them in large shopping centers to process store-wide analytics locally.

Specialized Edge Processing Chips

Beyond general hardware, specialized chips accelerate specific edge tasks:

• GPU (Graphics Processing Unit): Accelerates AI model inference and video processing
• TPU (Tensor Processing Unit): Google's chip designed specifically for machine learning tasks
• NPU (Neural Processing Unit): Found in modern smartphones; runs AI models on-device at low power
• FPGA (Field-Programmable Gate Array): Reprogrammable chip that performs specific tasks extremely fast; used in high-frequency trading and radar processing

Choosing the Right Hardware