Metrics comparison summary

Here’s the executive summary on metrics collection.

For AWS production workloads, we recommend metric streams. They have a better cost profile at scale, lower latency for alerting, and are easier to maintain long-term. Scrape jobs are still great for quick evaluation and proof-of-concept setups.

For Azure, the serverless scrape jobs work great for production. Simple setup with no infrastructure to manage.

For GCP, scrape jobs are your primary option, with Alloy available for more complex needs.

If you’re running multi-cloud or want the security benefits of managed identity instead of API keys, Grafana Alloy gives you one collector that handles everything.

A common pattern: start with scrape jobs to prove value quickly, then migrate to the recommended production approach once you’re ready to scale.