Same Job, Two Engines: InnoDB and Iceberg Internals
Summary
An in-depth comparison of OLTP InnoDB vs OLAP Iceberg-based lakehouse (Polaris/MinIO/Spark). It explains MVCC, partitioning, and the multi-layer metadata architecture (data files, manifests, manifest lists, and metadata.json), and highlights how columnar pruning enables reading only needed columns for faster analytics. The post also shares practical, hard-earned lessons from migrating analytics workloads from OLTP to OLAP.