IoT Applications

Introduction

IoT workloads — connected cars, smart meters, factory sensors, wearables — generate billions of small, time-ordered events per day. MongoDB 5.0+ introduced time-series collections specifically for this pattern: automatic bucketing, columnar-like storage, 10×+ compression and purpose-built indexes.

Bosch, Toyota, Iron Mountain and dozens of utilities run IoT platforms on MongoDB Atlas time-series for exactly this reason — one technology spans ingest, query and analytics.

Understanding the topic

IoT-shaped MongoDB design:

• Time-series collection with timeField, metaField, granularity.
• metaField = device id / tenant — drives bucketing locality.
• Automatic bucketing — measurements grouped into compact buckets per device per time window.
• Window stages ($setWindowFields) for moving averages and anomaly detection.
• Online Archive moves cold data to S3 after N days — keeps hot working set tiny.
• Atlas Stream Processing for real-time alerts off Kafka/MQTT streams.

Informative example

Create a time-series collection and ingest sensor events:

db.createCollection("readings", {
  timeseries: {
    timeField: "ts",
    metaField: "device",        // { id, fleet, model }
    granularity: "seconds"
  },
  expireAfterSeconds: 60 * 60 * 24 * 90   // keep 90 days hot
});

db.readings.insertMany([
  { ts: new Date(), device: { id: "veh-42", fleet: "EU-1" }, speed: 88, temp: 22.4 },
  { ts: new Date(), device: { id: "veh-42", fleet: "EU-1" }, speed: 90, temp: 22.5 }
]);

// 1-minute moving average per device using window stages
db.readings.aggregate([
  { $match: { "device.fleet": "EU-1", ts: { $gte: from } } },
  { $setWindowFields: {
      partitionBy: "$device.id", sortBy: { ts: 1 },
      output: { tempAvg: { $avg: "$temp",
                           window: { range: [-60, 0], unit: "second" } } } } }
]);

Real-world use

Bosch streams 1M+ events/sec from connected appliances into Atlas time-series; Toyota powers fleet telemetry; Vaillant runs predictive maintenance on heating systems — all on MongoDB time-series + Atlas Stream Processing.

Best practices

• Always pick the right granularity at creation — it cannot be changed later.
• Use the metaField for the dimension you filter on most (device id, tenant).
• Combine with Online Archive to keep hot data lean while preserving history on S3.
• Apply $setWindowFields rather than self-joins for moving averages and rates.