GCP IoT Architecture Patterns

Reference Architecture

Standard IoT Pipeline

code

┌─────────────────────────────────────────────────────────────────────────┐
│                           Google Cloud Platform                          │
│                                                                           │
│   ┌─────────┐      ┌─────────────┐      ┌─────────────┐                │
│   │ ESP32   │ ───► │ Cloud Run   │ ───► │  Pub/Sub    │                │
│   │ Sensors │ HTTPS│ API Gateway │      │  Topic      │                │
│   └─────────┘      └─────────────┘      └──────┬──────┘                │
│                                                 │                        │
│                    ┌────────────────────────────┼────────────────┐      │
│                    │                            │                │      │
│                    ▼                            ▼                ▼      │
│            ┌─────────────┐             ┌─────────────┐   ┌───────────┐ │
│            │  Firestore  │             │   Push to   │   │  BigQuery │ │
│            │  Database   │             │  WebSocket  │   │ Analytics │ │
│            └─────────────┘             └──────┬──────┘   └───────────┘ │
│                    │                          │                        │
│                    └──────────────────────────┼────────────────────────┤
│                                               │                        │
└───────────────────────────────────────────────┼────────────────────────┘
                                                │
                                                ▼
                                        ┌─────────────┐
                                        │   React     │
                                        │  Frontend   │
                                        │ (Firebase)  │
                                        └─────────────┘

Pattern 1: Direct Push (Recommended for < 1000 devices)

Flow: ESP32 → Cloud Run → Pub/Sub (Push) → WebSocket Service → Frontend

Pros:

•Low latency (< 500ms end-to-end)
•Simple architecture
•Cost-effective for small deployments

Cons:

•Limited buffering
•Single region

Implementation:

javascript

// Cloud Run: Receive telemetry and publish
app.post('/api/telemetry', async (req, res) => {
  const { deviceId, readings } = req.body;

  await pubsub.topic('telemetry').publish(
    Buffer.from(JSON.stringify({ deviceId, readings, timestamp: Date.now() }))
  );

  res.status(200).json({ status: 'received' });
});

Pattern 2: Fan-Out with Processing

Flow: ESP32 → Cloud Run → Pub/Sub → Cloud Functions → Multiple Destinations

Use When:

•Need to store in database AND send to frontend
•Processing/transformation required
•Multiple subscribers need the data

javascript

// Cloud Function: Process and fan-out
exports.processTelemetry = async (message) => {
  const data = JSON.parse(Buffer.from(message.data, 'base64'));

  // Store in Firestore
  await db.collection('telemetry').add(data);

  // Update device status
  await db.collection('devices').doc(data.deviceId).set({
    lastSeen: FieldValue.serverTimestamp(),
    status: 'online'
  }, { merge: true });

  // Publish to frontend topic
  await pubsub.topic('frontend-updates').publish(
    Buffer.from(JSON.stringify(data))
  );
};

Pattern 3: Batch Processing (High Volume)

Flow: ESP32 → Cloud Run → Pub/Sub → Dataflow → BigQuery + Firestore

Use When:

•1000+ devices
•Need analytics/historical data
•High message volume (> 10 msg/sec)

Device Status Patterns

Heartbeat Pattern

cpp

// ESP32: Send heartbeat every 30 seconds
void loop() {
  if (millis() - lastHeartbeat > 30000) {
    sendHeartbeat();
    lastHeartbeat = millis();
  }
}

javascript

// Cloud Function: Mark offline if no heartbeat
exports.checkDeviceStatus = functions.pubsub
  .schedule('every 1 minutes')
  .onRun(async () => {
    const cutoff = Date.now() - 60000; // 1 minute
    const devices = await db.collection('devices')
      .where('lastSeen', '<', new Date(cutoff))
      .where('status', '==', 'online')
      .get();

    const batch = db.batch();
    devices.forEach(doc => {
      batch.update(doc.ref, { status: 'offline' });
    });
    await batch.commit();
  });

Last Will Pattern (MQTT-style)

Store expected disconnect behavior:

javascript

// On device connect
await db.collection('devices').doc(deviceId).set({
  status: 'online',
  lastWill: 'offline'
}, { merge: true });

Pub/Sub Configuration

Topic Setup

bash

# Create telemetry topic
gcloud pubsub topics create telemetry --project=$PROJECT_ID

# Create push subscription for websocket
gcloud pubsub subscriptions create telemetry-ws-push \
  --topic=telemetry \
  --push-endpoint=https://ws-service.run.app/pubsub \
  --push-auth-service-account=push-invoker@$PROJECT_ID.iam.gserviceaccount.com \
  --ack-deadline=10 \
  --project=$PROJECT_ID

# Create pull subscription for processing
gcloud pubsub subscriptions create telemetry-process \
  --topic=telemetry \
  --ack-deadline=60 \
  --project=$PROJECT_ID

Dead Letter Queue

bash

# Create dead letter topic
gcloud pubsub topics create telemetry-dead-letter

# Update subscription with dead letter
gcloud pubsub subscriptions update telemetry-ws-push \
  --dead-letter-topic=telemetry-dead-letter \
  --max-delivery-attempts=5

Cloud Run Configuration

Recommended Settings for IoT API

yaml

# service.yaml
spec:
  template:
    metadata:
      annotations:
        autoscaling.knative.dev/minScale: "1"  # Avoid cold starts
        autoscaling.knative.dev/maxScale: "10"
    spec:
      containerConcurrency: 80
      timeoutSeconds: 60
      containers:
        - resources:
            limits:
              memory: 512Mi
              cpu: "1"

WebSocket Service Config

bash

# Enable session affinity for socket.io
gcloud run services update ws-service \
  --session-affinity \
  --min-instances=1

Security Best Practices

Device Authentication

cpp

// ESP32: Include API key in header
http.addHeader("X-API-Key", API_KEY);
http.addHeader("X-Device-ID", DEVICE_ID);

javascript

// Cloud Run: Validate device
app.use('/api/telemetry', (req, res, next) => {
  const apiKey = req.headers['x-api-key'];
  const deviceId = req.headers['x-device-id'];

  if (!validateDevice(apiKey, deviceId)) {
    return res.status(401).json({ error: 'Unauthorized device' });
  }
  next();
});

Use Secret Manager

bash

# Store secrets
echo -n "my-api-key" | gcloud secrets create api-key --data-file=-

# Access in Cloud Run
gcloud run services update my-service \
  --set-secrets=API_KEY=api-key:latest

Cost Optimization

Component	Free Tier	Optimization
Cloud Run	2M requests/mo	Batch telemetry, use min-instances=0 for dev
Pub/Sub	10GB/mo	Compress messages, batch publishes
Firestore	50K reads/day	Cache reads, batch writes
Functions	2M invocations/mo	Use Cloud Run for high-volume

Monitoring Setup

bash

# Create uptime check
gcloud monitoring uptime-check-configs create api-health \
  --display-name="IoT API Health" \
  --http-check=request-method=GET,path=/health,use-ssl=true \
  --monitored-resource=gce_instance

# Create alert for errors
gcloud alpha monitoring policies create \
  --notification-channels=[CHANNEL] \
  --condition-display-name="High Error Rate" \
  --condition-filter='resource.type="cloud_run_revision" AND metric.type="run.googleapis.com/request_count" AND metric.labels.response_code_class="5xx"'

Telemetry Schema

Recommended JSON Structure

json

{
  "deviceId": "esp32-001",
  "timestamp": 1705337600000,
  "type": "telemetry",
  "readings": {
    "temperature": 25.5,
    "humidity": 60.2,
    "pressure": 1013.25
  },
  "meta": {
    "firmware": "1.2.0",
    "rssi": -65,
    "battery": 85,
    "uptime": 3600000
  }
}

Firestore Document Structure

code

/devices/{deviceId}
  - status: "online" | "offline"
  - lastSeen: Timestamp
  - firmware: "1.2.0"
  - config: { ... }

/devices/{deviceId}/telemetry/{timestamp}
  - readings: { temperature, humidity, ... }
  - meta: { rssi, battery, ... }

Common Anti-Patterns

❌ Don't: Poll from frontend every second ✅ Do: Use realtime listeners (Firestore onSnapshot, WebSocket)

❌ Don't: Store every reading in Firestore ✅ Do: Aggregate/downsample for long-term storage

❌ Don't: Use single Cloud Run instance for everything ✅ Do: Separate API gateway from WebSocket service

❌ Don't: Hardcode credentials in ESP32 ✅ Do: Use secure provisioning and API keys