Every source into Iceberg REST Catalog, in open file formats.

Parquet, CSV, JSON, and JSONL. Each flow run writes a file in the format you pick, with format-level controls - CSV delimiter, header, and date format, or the timestamp unit for Parquet, JSON, and JSONL.

Each flow run lands as its own dated file, so Iceberg REST Catalog is organized as a date-partitioned dataset. Query engines then read only the partitions they need instead of scanning everything, which keeps queries fast and costs predictable.

With the catalog URL (the REST endpoint). Authentication is optional - an OAuth token endpoint or username and password, left empty for open catalogs - and the warehouse location (e.g. s3://my-bucket/warehouse) is either set by you or server-side by the catalog.

Either the catalog vends short-lived storage credentials automatically, or you give Dataddo an S3 access key, secret and region directly. Data is written in namespace.table form (the namespace cannot contain whitespace) and tables are auto-created on the first write.

insert (append rows) and truncate_insert (empty the table, then write current data; the default). Upsert and composite-key modes are not available here, and the exact list can depend on the catalog and the current table state. Dataddo works with any REST-catalog implementation - Polaris, Unity Catalog, Nessie, Lakekeeper, S3 Tables, BigLake, OneLake and Glue.

Because Dataddo writes standard Apache Iceberg tables, you get ACID snapshot isolation for consistent reads during writes, safe schema evolution (add, rename, reorder or retype columns with no table rewrite), and time travel - query an older snapshot by ID or timestamp, or roll a table back to a known-good state.

The same table is readable by Spark, Flink, Trino, Presto, Hive, Snowflake, Amazon Athena, Google BigQuery, Dremio and more, with no copies and no vendor lock-in. Hidden partitioning prunes files automatically, so queries stay fast without manual partition filters.

Yes. Insert keeps writing new files, and create-new-or-replace overwrites the file at the same name. For lakehouse table formats like Apache Iceberg, Dataddo applies warehouse-style write modes to the table itself.

Dataddo is SOC 2 Type II and ISO 27001 certified. Data is encrypted in transit and at rest, PII can be masked or hashed, and EU or US data residency is available. You connect Iceberg REST Catalog with your own keys or an assumed role.

Incremental loads write only new or changed data, and columnar Parquet compresses well - so you store and scan less. You control the schedule and the partition layout, which keeps storage and query costs predictable.

No. Data lands in open file formats in a bucket you own, and pipelines are storage-agnostic - you can add or switch object stores, or point the same sources at a warehouse, without rebuilding anything.