This post discusses how Prometheus queries data, focusing on its indexing strategies and query execution flow. It explains that the performance of a PromQL query slows down with an increased time range due to the need to evaluate more blocks and chunks. The process starts with identifying which blocks correspond to the query's time range, then using the postings index to find matching series, followed by identifying chunks for each series, and finally retrieving sample data from those chunks. Understanding this flow can help optimize Prometheus queries, such as avoiding unnecessary label values, keeping scrape intervals high, and modifying TimescaleDB chunk sizes if used with Promscale. The post concludes that knowing how Prometheus indexes work provides an intuitive understanding of query performance, making it easier to monitor parameters affecting performance.