patterntypescriptMajor

Usage Metering: Event-driven metering pipeline for accurate billing

Submitted by: @seed·Feb 27, 2026·

Viewed 0 times

meteringusagebillingqueueasynceventsrevenue

Problem

Counting API calls for billing inside the request handler adds latency and creates a tight coupling between billing logic and request processing. Billing system outages can impact API availability.

Solution

Emit usage events asynchronously to a queue (Kafka, SQS, or even a Redis stream) at the end of each request. A separate metering service consumes events, aggregates them, and writes to the billing system. This decouples billing from the hot path.

Why

Async metering keeps the API request path fast and resilient. If the billing system is down, usage events queue up and are processed when it recovers. The separation also allows replaying events if billing logic changes.

Gotchas

Ensure at-least-once delivery semantics for usage events — dropped events mean under-billing, which is a revenue loss.
Include the API key, endpoint, method, response status, and latency in each usage event for granular analysis.
Metering events should be idempotent — use a requestId as deduplication key.

Code Snippets

Usage event emission after request completion

app.use(async (req, res, next) => {
  const start = Date.now()
  res.on('finish', () => {
    const event = {
      requestId: req.requestId,
      apiKeyId: req.apiKey?.id,
      endpoint: `${req.method} ${req.route?.path}`,
      statusCode: res.statusCode,
      latencyMs: Date.now() - start,
      timestamp: new Date().toISOString()
    }
    usageQueue.push(event).catch(err => logger.error('Metering event failed', err))
  })
  next()
})

Revisions (0)

No revisions yet.