This type of federation is useful when you need a few select time series from another Prometheus instance locally for recording or alerting rules. Going back to the example scenario where you have a Prometheus instance tasked with scraping Node Exporters and another for applications, this pattern would allow you to federate specific OS-level metrics that could then be used in applications alerts, as shown in the following diagram:

The configuration of cross-service federation looks mostly the same as the previous pattern, but in this case the scraped Prometheus is in the same logical tier and the selectors that were used should be matching specific metrics instead of aggregates.

In the next section, we're going to introduce a new component that has been gaining traction in the Prometheus community and tackles the challenge of having a global view in a new and interesting way.