I run a small Meshtastic mesh network out of Benicia, CA. Five nodes, a mix of fixed and mobile hardware, with the gateway publishing packets to an MQTT broker on my NAS. I wanted the telemetry - battery levels, signal quality, GPS positions, environmental sensors - flowing into InfluxDB and visualized in Grafana. This post covers how I built the pipeline and the specific problems I hit along the way.
I picked up a Ugreen DXP4800 Pro NAS earlier this year. It’s a 4-bay unit with an Intel Core i3-1315U, 10GbE + 2.5GbE LAN, and dual M.2 NVMe SSD slots. I upgraded the RAM to 2x8GB DDR5-5600 SO-DIMMs (replacing the stock single stick) for 16GB total. The i3-1315U is a chip more commonly found in laptops than NAS enclosures, which turns out to matter a lot when you want to run Plex Media Server.
The Ring smart doorbell does far more than your ordinary wired doorbell; it’s able to deliver audio & video recordings in real-time over a WiFi link connected to the Internet. When your doorbell is pressed or motion is detected, the doorbell automatically begins streaming video to the Ring servers. Ring will send a push notification to your phone and archive the video, where it can be downloaded or viewed through the Ring website or app.
Kinesis is a hosted service from Amazon Web Services that offers the ability to publish & consume durable streams of high-volume data. There’s no limit to the number of sources that feed a Kinesis stream, making Kinesis a convenient ingestion point for a data processing pipeline.
I’ve been using Apache Flink to perform streaming aggregations for real-time alerting across several attributes. It’s a phenomal stream-processing engine with an easy-to-use API, simple switching between local and cluster job execution, and great support for job checkpointing to simplify recovery when systems fail.