Loki's New Architecture: 10x Faster Logs at Petabyte Scale | Grafana Labs

This GrafanaCon talk covers Loki's complete architectural overhaul for petabyte-scale structured logging. Engineers explain the Kafka-backed write path that decouples reads from writes, DataObjects (a columnar format storing each field in its own column), and the new object-storage-native index from the Logline acquisition that reduces massive scans to 8GB. The video is for platform engineers, SREs, and developers managing large-scale observability infrastructure who need to understand migration paths and performance optimization strategies.