Industrial edge computing platforms 2027: AWS IoT Greengrass v2 (Lambda@Edge, ML inference, AWS ecosystem), Azure IoT Edge (Docker modules, Azure ML, Microsoft stack), Siemens Industrial Edge (TIA Portal integration, PROFINET, OT-native), Litmus Edge (protocol-agnostic, 250+ drivers, Microsoft partner), Cisco IOx (network edge, IT-managed). Architecture: edge gateway → container runtime → cloud sync. OPC UA + MQTT standard protocols.
For manufacturing IIoT architects and IT/OT teams in 2027, edge computing is no longer optional — it’s the architectural foundation for industrial data processing. Edge platforms enable: real-time data processing at the source (sub-second response without cloud round-trip), offline operation (plant continues when cloud connection drops), data filtering and aggregation (reduce cloud bandwidth/cost by 90%+), ML inference at the machine (defect detection, anomaly detection, predictive maintenance), and protocol translation (convert OPC UA, PROFINET, Modbus, EtherNet/IP to MQTT/REST for cloud). This guide compares the major edge platforms, details architecture patterns, and provides selection criteria for manufacturing use cases.
Edge platform comparison matrix
| Platform | Vendor | Runtime | OT protocols | ML inference | Cloud integration | Pricing model |
|---|---|---|---|---|---|---|
| AWS IoT Greengrass v2 | Amazon | Java/Python Lambda, Docker containers | OPC UA (via SiteWise Edge), Modbus, protocol adapters | SageMaker Neo, TensorFlow Lite, ONNX Runtime | AWS IoT Core, S3, SiteWise, SageMaker, Kinesis | Free runtime + AWS service usage |
| Azure IoT Edge | Microsoft | Docker containers (Linux/Windows), IoT Edge modules | OPC UA (OPC Publisher module), Modbus, custom modules | Azure ML, ONNX Runtime, Custom Vision, OpenVINO | Azure IoT Hub, Blob Storage, Stream Analytics, Digital Twins | Free runtime + Azure service usage |
| Siemens Industrial Edge | Siemens | Docker containers, Siemens Edge Apps | PROFINET, OPC UA, S7, Modbus — native TIA Portal integration | TensorFlow, ONNX, Siemens AI SDK | MindSphere (Siemens), Azure, AWS via connectors | Hardware + app subscription (€200-500/yr per app) |
| Litmus Edge | Litmus | Docker/K3s containers | 250+ OT protocol drivers (OPC UA, PROFINET, EtherNet/IP, Modbus, BACnet, MT Connect, FANUC, Siemens S7, etc.) | TensorFlow, PyTorch, ONNX | Azure (Microsoft partner), AWS, GCP, any MQTT/REST | Subscription per edge device |
| Cisco IOx | Cisco | Docker containers on Cisco routers/switches | Limited OT (Modbus, OPC UA via partners) | Limited (partner solutions) | Cisco DNA Center, ThousandEyes, Meraki, any cloud | Included with Cisco hardware |
| ADLINK Edge | ADLINK | ROS 2 (robotics), Docker, DDS | OPC UA, DDS, PROFINET adapters | NVIDIA Jetson integration, OpenVINO | Azure, AWS, custom | Hardware + software license |
| Advantech WISE-EdgeLink | Advantech | Proprietary + Docker | 200+ drivers (Modbus, OPC UA, BACnet, MQTT, proprietary PLCs) | Limited (partner) | Azure, AWS, WISE-PaaS (Advantech cloud) | Hardware + software license |
| Stratus ztC Edge | Stratus (Rockwell partner) | VM-based (zero-touch redundancy) | OPC UA, Kepware integration | Partner solutions | FactoryTalk (Rockwell), any cloud | Hardware + subscription |
Architecture patterns: edge-cloud hybrid for manufacturing
Pattern 1: Edge gateway + cloud analytics
Most common pattern for OEE + IIoT:
- Edge: gateway collects data from PLCs (OPC UA), sensors (MQTT), machines → filters, aggregates, computes real-time KPIs (OEE per minute) → sends aggregated data to cloud every 1-60 seconds
- Cloud: stores time-series, runs analytics (trends, Pareto, benchmarking), serves dashboards, trains ML models, multi-site aggregation
- Offline resilience: edge buffers data locally during cloud disconnection, auto-syncs when reconnected
- Data reduction: edge sends only aggregated KPIs + events (alarms, state changes) = 90-99% bandwidth reduction vs raw sensor data
Pattern 2: Edge ML inference + cloud training
For predictive maintenance, defect detection, anomaly detection:
- Cloud: trains ML models on historical data (weeks/months), pushes trained model to edge
- Edge: runs inference in real-time on live sensor/camera data, generates predictions/alerts locally (<100ms response)
- Feedback loop: edge sends inference results + edge cases back to cloud for model retraining
- Platforms: AWS SageMaker Edge Manager, Azure ML Edge, Siemens AI SDK, NVIDIA Triton on Jetson
Pattern 3: Edge-to-edge (peer) for distributed control
Emerging for multi-machine coordination, digital twin synchronization:
- Edge gateways communicate peer-to-peer via MQTT or DDS
- No cloud dependency for real-time coordination
- Cloud used only for fleet management, configuration, long-term analytics
- Enabled by OPC UA Pub/Sub multicast and Sparkplug B topic hierarchy
Container orchestration at the industrial edge
| Technology | Use case | Notes |
|---|---|---|
| Docker | Single-node edge gateways | Standard container runtime, supported by all edge platforms |
| K3s (Rancher/SUSE) | Multi-container edge with orchestration | Lightweight Kubernetes for edge. Litmus Edge, Red Hat Device Edge use K3s |
| MicroK8s (Canonical) | Similar to K3s, Ubuntu ecosystem | Used in Ubuntu-based industrial edge deployments |
| Azure Arc + GitOps | Fleet management of edge K8s clusters | Azure IoT Edge + Arc for managing 100s of edge devices centrally |
| AWS ECS Anywhere / EKS Anywhere | AWS-managed K8s at edge | Extends AWS container management to on-premise edge hardware |
Best practice 2027: Docker for simple single-app edge gateways (most OEE deployments), K3s for multi-workload edge nodes requiring orchestration (ML + protocol conversion + OEE + local dashboard).
Download the white paper
Enter your email address to receive our White Paper
OT protocol support: the critical differentiator
The most important edge platform selection criterion for manufacturing is OT protocol driver coverage:
| Protocol | Used by | AWS Greengrass | Azure IoT Edge | Siemens Edge | Litmus Edge |
|---|---|---|---|---|---|
| OPC UA | Modern PLCs (Siemens S7-1500, Beckhoff, B&R) | ✅ SiteWise Edge | ✅ OPC Publisher | ✅ Native | ✅ Native |
| Siemens S7 (MPI/PROFINET) | Siemens S7-300/400/1200/1500 | ⚠️ Partner | ⚠️ Partner | ✅ Native | ✅ Driver |
| EtherNet/IP | Rockwell CompactLogix/ControlLogix | ⚠️ Partner | ⚠️ Partner | ⚠️ Via connector | ✅ Driver |
| PROFINET | Siemens, B&R, Beckhoff (field level) | ❌ | ❌ | ✅ Native | ✅ Driver |
| Modbus TCP/RTU | Legacy PLCs, energy meters, sensors | ✅ Component | ✅ Module | ✅ Native | ✅ Driver |
| MTConnect | CNC machines (Mazak, Haas, DMG MORI) | ⚠️ Custom | ⚠️ Custom | ❌ | ✅ Driver |
| MQTT | IIoT sensors, Sparkplug B | ✅ Native | ✅ Native | ✅ Native | ✅ Native |
| BACnet | Building automation, HVAC | ❌ | ⚠️ Partner | ❌ | ✅ Driver |
| FANUC FOCAS | Fanuc CNC controllers | ❌ | ❌ | ❌ | ✅ Driver |
Key insight: Litmus Edge wins on OT protocol breadth (250+ drivers out-of-box). Siemens Industrial Edge wins for Siemens-centric factories (native TIA Portal + PROFINET). AWS/Azure require partner solutions for many OT protocols — stronger for cloud integration than OT connectivity.
ML at the industrial edge
- Hardware options: NVIDIA Jetson Orin (AI inference 100-275 TOPS), Intel Movidius (OpenVINO), Google Coral TPU (4 TOPS, low power), Hailo-8 (26 TOPS, PCIe/M.2 module), Qualcomm QCS8550 (AI engine)
- Common use cases: visual defect detection (camera + CNN at edge, <100ms inference), vibration anomaly detection (autoencoder at edge, real-time), OEE prediction (LSTM at edge, forecast next shift), energy optimization (load prediction at edge), quality SPC alert (real-time process parameter monitoring)
- Framework support: TensorFlow Lite, ONNX Runtime, PyTorch Mobile, OpenVINO (Intel), TensorRT (NVIDIA), ArmNN
- Edge-cloud ML lifecycle: train in cloud (GPU clusters), optimize for edge (quantization, pruning), deploy to edge (OTA update), monitor at edge (drift detection), retrain in cloud (feedback loop)
Selection criteria: which edge platform for your factory
| Criterion | Best choice |
|---|---|
| Siemens-centric factory (TIA Portal, S7, PROFINET) | Siemens Industrial Edge (native integration) |
| Multi-vendor PLC landscape (Siemens + Rockwell + Fanuc + others) | Litmus Edge (250+ protocol drivers) |
| AWS cloud strategy, IoT Core, SageMaker | AWS IoT Greengrass v2 |
| Azure/Microsoft strategy, Power BI, Dynamics 365 | Azure IoT Edge |
| IT-managed network edge (Cisco infrastructure) | Cisco IOx |
| Robotics + machine vision (ROS 2) | ADLINK Edge + NVIDIA Jetson |
| OEE specialist platform integration | Any — TeepTrak Box operates independently as dedicated OEE edge sensor (no dependency on general-purpose edge platform) |
Integration with OEE platforms
Edge platforms and OEE platforms are complementary, not competing:
- General-purpose edge platform (Greengrass, Azure Edge, Litmus, Siemens Edge): provides broad data collection from diverse OT protocols, ML inference, general-purpose compute, multi-application hosting
- OEE specialist platform (TeepTrak Pulse): provides focused OEE A × P × Q measurement, Six Big Losses analysis, standardized methodology (ISO 22400-2), multi-site benchmarking, operator-facing dashboards
- Integration: general-purpose edge platform feeds normalized machine data to OEE specialist via OPC UA or MQTT. Or: OEE specialist runs as container/module on general-purpose edge platform. TeepTrak Pulse: operates standalone (TeepTrak Box edge sensor) OR integrates with existing edge platform via OPC UA/MQTT.
FAQ: Edge computing platforms for industrial IoT
AWS Greengrass or Azure IoT Edge?
Choose based on cloud strategy: AWS Greengrass if using AWS IoT Core, SageMaker, S3, SiteWise. Azure IoT Edge if using Azure IoT Hub, Power BI, Dynamics 365, Digital Twins. Both free runtime + cloud service charges. Both limited OT protocol support vs specialized platforms (Litmus, Siemens). For multi-cloud or cloud-agnostic: Litmus Edge supports both AWS and Azure.
Why Siemens Industrial Edge for Siemens factories?
Siemens Industrial Edge provides: native TIA Portal integration (direct PLC data access without separate OPC UA configuration), native PROFINET support, Siemens Edge App marketplace (pre-built apps for OEE, energy, quality), managed from Siemens Industrial Edge Management (IEM) central platform, hardware certified for industrial environments (SIMATIC IPC series). If factory is 80%+ Siemens PLCs, Siemens Edge is simplest path.
What is Litmus Edge’s advantage?
Litmus Edge: 250+ OT protocol drivers out-of-box (OPC UA, Siemens S7, Rockwell EtherNet/IP, PROFINET, Modbus, MTConnect, FANUC FOCAS, BACnet, etc.). Key for heterogeneous factories with multiple PLC brands. Microsoft partner (Azure integration). K3s-based container orchestration. Data normalization layer: converts diverse OT protocols to unified MQTT/REST output for any cloud/analytics platform.
Do I need edge computing for OEE?
For basic OEE measurement: TeepTrak Box is a dedicated edge sensor that handles OEE data collection independently — no general-purpose edge platform needed. 1 hour install per machine, Ethernet or 4G/5G connectivity, cloud-managed. For advanced use cases (ML at edge, multi-application hosting, complex protocol conversion from 10+ PLC brands): general-purpose edge platform (Litmus, Siemens Edge, Greengrass, Azure Edge) provides broader capability.
What hardware for industrial edge?
Industrial edge hardware: Siemens SIMATIC IPC (IPC127E, IPC227G, IPC427E), Advantech UNO/EPC series, ADLINK DLAP/ROSCube, Dell Edge Gateway 3000/5000, HPE Edgeline EL8000, Lenovo ThinkEdge SE70, OnLogic Helix/Factor series. For ML inference: NVIDIA Jetson Orin (275 TOPS), ADLINK DLAP-411 (Jetson-based). Requirements: DIN rail/panel mount, -20 to +60°C, 24V DC, fanless, industrial certifications (CE, UL, ATEX/IECEx for hazardous).
Container orchestration at edge: Docker or Kubernetes?
Docker for simple edge (single OEE app, single protocol converter) — most manufacturing edge deployments. K3s (lightweight Kubernetes) for complex edge (multiple containers: ML inference + protocol conversion + local dashboard + OEE + data buffering). K3s adds: auto-restart, rolling updates, resource management, multi-node clustering. 2027 trend: K3s adoption increasing as edge workloads grow more complex.
How does edge computing reduce cloud costs?
Edge data reduction: raw sensor data (1000 data points/sec × 100 machines = 100K points/sec = ~50 GB/day) → edge aggregates to 1-minute KPIs (100 machines × 10 KPIs × 1/min = 1000 points/min = ~50 MB/day) = 99.9% reduction. Cloud ingestion costs (AWS IoT Core, Azure IoT Hub) scale with message volume — edge aggregation reduces cloud costs 10-100×. Plus: edge provides sub-second response without cloud round-trip latency.
Cybersecurity for edge computing IEC 62443?
Edge cybersecurity per IEC 62443: (1) secure boot (verified firmware), (2) encrypted storage (data at rest), (3) TLS 1.3 for all communications, (4) certificate-based mutual authentication (mTLS), (5) RBAC per container/module, (6) audit trail (SR 2.8), (7) automatic security patching (OTA updates), (8) network segmentation (edge in DMZ between OT and cloud). NIS2 Directive requires edge platforms in manufacturing to meet minimum cybersecurity baseline.
What about Red Hat Device Edge?
Red Hat Device Edge (RHEL for Edge + MicroShift lightweight K8s): enterprise Linux at edge, managed via Red Hat Advanced Cluster Management. Strong for organizations with Red Hat/OpenShift strategy. RHEL for Edge: immutable OS image, OTA updates, rollback capability. MicroShift: single-node OpenShift for edge. Adoption growing in telecom and industrial. Alternative to K3s with enterprise Red Hat support.
How does TeepTrak Box compare to general edge platforms?
TeepTrak Box is purpose-built OEE edge sensor: clamp-on installation (no PLC modification), 1 hour per machine, dedicated to OEE A × P × Q measurement, Ethernet + 4G/5G connectivity. Not a general-purpose edge platform — doesn’t host custom containers or ML workloads. Advantage: zero IT dependency, instant deployment, OEE-specific. For plants needing both OEE and general edge compute: TeepTrak Box for OEE + Litmus/Siemens/AWS/Azure edge for other workloads — complementary architecture.
Future of industrial edge computing?
2027-2030 trends: (1) edge AI becoming standard (NVIDIA Jetson Orin, Hailo-8 as commodity), (2) K3s/lightweight K8s replacing Docker-only at edge, (3) fleet management at scale (1000+ edge devices managed centrally via Azure Arc, AWS Systems Manager, Siemens IEM), (4) edge-native applications (designed edge-first, cloud-optional), (5) OPC UA FX + TSN enabling deterministic edge-to-edge communication, (6) sovereign edge (data processed in-country per NIS2, PIPL, GDPR). Edge is the new plant floor computing standard.
Conclusion
Industrial edge computing in 2027 is the architectural foundation for manufacturing IIoT. Platform selection depends on: Siemens-centric factory → Siemens Industrial Edge (native TIA Portal/PROFINET), multi-vendor PLC landscape → Litmus Edge (250+ protocol drivers), AWS strategy → AWS IoT Greengrass v2, Azure/Microsoft strategy → Azure IoT Edge, IT-managed network → Cisco IOx. Architecture: edge-cloud hybrid with Docker containers (simple) or K3s (complex multi-workload). Key differentiation: OT protocol driver coverage (Litmus and Siemens lead), ML inference (NVIDIA Jetson + all major platforms), cloud integration (AWS/Azure lead). For OEE: TeepTrak Box edge sensor operates independently (purpose-built OEE, zero IT dependency) OR integrates with general-purpose edge platforms via OPC UA/MQTT — complementary architecture. Edge reduces cloud costs 10-100× via data aggregation, enables sub-second response, and provides offline resilience.
Next step: download the TeepTrak edge computing architecture guide or request a free edge platform assessment for your factory IIoT stack.
0 Comments