OLake

OLake is an open-source, Iceberg-first extract-and-load engine built by Datazip, Inc., led by CEO Rohan Khameshra and CTO Shubham Satish Baldava. It addresses the cost and complexity of moving operational databases into an open lakehouse without heavy Spark, Kafka or Debezium setups.

OLake Go replicates PostgreSQL, MySQL, MongoDB, Oracle, SQL Server, IBM Db2, Kafka and S3 into Apache Iceberg or Parquet on S3, with full loads and CDC. It works with catalogs such as AWS Glue, Hive, REST, Nessie, Unity, Polaris, S3 Tables and BigLake. OLake Fusion maintains Iceberg tables with tiered compaction of small and delete files. The software is self-hosted, so users pay only for infrastructure; logos on the site include Booking.com, Qualcomm, Razorpay, Physics Wallah and Lenskart.

• Database CDC to Iceberg: Full load and change data capture from major relational and NoSQL databases.
• Stream and file sources: Ingests from Kafka and S3 in addition to databases.
• Broad catalog support: Writes to Glue, Hive, REST, Nessie, Unity, Polaris, S3 Tables and BigLake.
• Iceberg v3 deletes: Writes deletion vectors for upserts and CDC with format interoperability.
• OLake Fusion maintenance: Tiered compaction and health metrics for Iceberg tables.
• Self-hosted open source: Runs on the customer's infrastructure, with enterprise support on request.

A Vietnamese e-commerce or fintech company can replicate MySQL and PostgreSQL into an Iceberg lakehouse on S3 and query it with Trino. A manufacturer running IBM Db2 or SQL Server can offload history to Iceberg for analytics. A data team can cut Spark costs by using OLake Fusion for table compaction.

Teams avoid per-row connector fees because OLake is open source and runs on their own infrastructure. The vendor's benchmark reports a 4.01 billion-row PostgreSQL-to-Iceberg full load in 1.91 hours for under $6. OLake Fusion compacted 1.8 billion rows in 27 minutes, 2.06x faster and about 52% cheaper per job than Spark compaction. Open Iceberg tables can be queried by Spark, Trino, Snowflake, Databricks, Athena or DuckDB.

Apache Iceberg is becoming the open table format of choice for lakehouses, letting companies avoid warehouse lock-in and separate storage from compute. Cost-sensitive data teams in South and Southeast Asia are adopting open-source ingestion to feed lakehouses on AWS and Google Cloud.

OLake is developed by Datazip, Inc.; its terms (updated 5 February 2026) list a Delaware agent address and are governed by Indian law. The former datazip.io domain now redirects to olake.io. Source code is on GitHub under datazip-inc. Published case studies include Xeno, Cordial and Bitespeed.