Pull doesn't scale - or does it?

This article argues against the common claim that Prometheus's pull-based metric collection doesn't scale. It explains that Prometheus differs significantly from systems like Nagios, focusing on collecting aggregated time series data rather than executing checks or handling raw events – minimizing the load from pulling. Ultimately, the author contends that scaling bottlenecks in Prometheus typically lie in data storage and processing, not the pull mechanism itself, and that pull-based systems can even offer advantages in configuration, flexibility, and discovery.