OLake Go can now write Iceberg v3 deletion vectors for upserts and CDC. From immutable Parquet and merge-on-read to equality and positional deletes—and why bitmap deletion vectors close the gap with query engines.
Same Iceberg table, same catalog, same data yet Spark reads it fine while Databricks fails. See what query-engine gaps break cross-engine interoperability.
Write Apache Iceberg tables into Google BigLake with OLake Go: GCP prerequisites, REST catalog configuration field by field, and common failure modes.
We benchmark Spark rewrite_data_files against OLake Fusion compaction on Apache Iceberg by running a full TPCH lineitem load from Postgres to GCP, applying 200k-record CDC batches every 2 minutes, and tracking TPC-H Query 6 performance, runtime, resource usage, and infrastructure cost.
We ran TPC-H queries on Iceberg tables with many small files, then compacted them and ran the same queries again. Here's how much faster compaction made them.
Make OLake's Merge-on-Read Iceberg tables work in Databricks with an automated MOR to Copy-on-Write script that gives accurate analytics queries.