BigData / Apache Iceberg Interview questions
What is the difference between Apache Iceberg and Delta Lake?
Both are open table formats bringing warehouse-like reliability to data lakes, but they differ in metadata architecture and their degree of coupling to a specific engine, stemming from their different origins — Iceberg from Netflix, Delta Lake from Databricks.
| Apache Iceberg | Delta Lake |
| Tree-based metadata: catalog, metadata file, manifest lists, manifests. | Append-only JSON transaction log (_delta_log), periodically checkpointed to Parquet. |
| Catalog decoupled from storage; broad multi-engine support. | Historically more tightly coupled to Spark; broader engine support has grown over time. |
| Hidden partitioning and true partition evolution without rewrites. | Liquid Clustering allows changing clustering keys without full rewrites, a different paradigm. |
| REST Catalog has become a widely adopted, vendor-neutral standard. | Universal Format (UniForm) can generate Iceberg-compatible metadata alongside Delta's own. |
The two ecosystems have been converging in practice: Delta Lake's UniForm can generate Iceberg metadata for the same physical table, and Iceberg's REST catalog has been adopted broadly across vendors, which means the choice increasingly comes down to which platform's native tooling and defaults best fit an organization, rather than the two formats being permanently incompatible silos.
What metadata structure does Delta Lake use, compared to Iceberg's tree of manifests?
What Delta Lake feature lets you change clustering without a full rewrite, similar in spirit to Iceberg's partition evolution?
More Related questions...
