How Apache Polaris Vends Credentials: Securing Data Access Without Sharing Keys

The Only Mutable Piece: How Apache Iceberg Tracks Petabytes of Data Without a Database By Prithvi S · April 2026 Here's a question that haunts distributed systems engineers: how do you track millions of files across petabytes of data without a central database to coordinate them? The traditional answer was to give up. Use Hive partitions and pray. Keep a metadata service that becomes your bottleneck. Coordinate writes through some external lock server. Accept that your table is only as consistent as your metadata store. Netflix couldn't live with that answer, so they invented Apache Iceberg. At its core, Iceberg solves this problem with a deceptively simple insight: make almost everything immutable, and reduce the mutable part down to exactly one thing: a pointer . I've spent time digging through the Iceberg codebase while working on data platform projects, and the metadata architecture is where the magic happens. It's elegant, correct, and runs on object stores. Let me walk you throug