[KubeCon + CloudNativeCon EU recap] Better histograms for Prometheus

Björn "Beorn" Rabenstein's talk at KubeCon + CloudNativeCon EU 2020 focused on the future of Prometheus histograms, proposing a solution to existing issues by introducing histograms with effectively infinite buckets arranged in a logarithmic schema. This approach, which handles sparseness efficiently, has been used in other systems but presents challenges within the Prometheus model. Rabenstein has conducted research using Grafana Labs Cortex clusters and developed proof-of-concept implementations for instrumentation and encoding in Prometheus. While confident in overcoming initial challenges, further work is needed to optimize querying and assess the cost of storing and querying these histograms. The adoption of high-resolution, sparse histograms will require significant changes across the Prometheus stack, but Rabenstein is optimistic about its potential benefits and seeks community input for further development.