After a long time, one device that excited me the most is the custom Edge Server that I assembled. It’s a DIY Intel NUC kit powered by a Dual Core i5 processor, 32GB DDR4 RAM, 500 GB SSD + 500GB M.2 SSD. This powerhouse is only twice the size of a Raspberry Pi but acts as a local endpoint for IoT devices. It is configured with Ethernet, WiFi, BLE, ZigBee, Z-Wave, and Thread protocols.
Since it emulates the public cloud functionality it runs the following stack –
- SurgeMQ as a MQTT Broker
- Apache Kafka as the ingestion endpoint
- PipelineDB for stream analytics
- Apache NiFi as the complex rules engine
- Node-Red for simple routing
- Apache OpenWhisk for Serverless Computing
- Minio as the Object Storage endpoint
- TensorFlow for Predictive Analytics and ML
- Grafana as the dashboard
All the above are deployed as containers running in a Kubernetes cluster on top of Ubuntu 64 bit. The stack can selectively switch to Azure, AWS, Bluemix, or GCP. The appliance can also run Azure IoT Gateway.
I will be shortly adding an external Nvidia GPU for CUDA.
I will publish the details of the setup when I am done. The entire setup and the use case will be demoed at an Edge Computing workshop that I am running in May.