Documentation feedback: /docs/sources/mimir/visualize/native-histograms/_index.md

[teyyubismayil]

I’d like a recommendation on the best call order for using native histogram functions. Specifically, which of the following variants is recommended?

1. histogram_sum(sum(rate(http_server_duration{}[5m])) by (cluster, service_namespace, service_name, http_route, http_method))
2. sum(histogram_sum(rate(http_server_duration{}[5m]))) by (cluster, service_namespace, service_name, http_route, http_method)

I compared the execution times of these queries (also for histogram_count function) in Mimir and found that the second one is about 20% faster. However, the documentation provides the first variant as an example.

What is the reasoning behind this choice?