Oracle Big Data Service vs Unilake

Oracle Big Data Service is a managed cloud service for creating Hadoop-based data lakes that process large amounts of data. Its distribution includes Apache Hadoop, Ambari, HBase, Hive, Spark, and other services. Customers can build clusters with OCI compute shapes, from small development environments to large production clusters on dedicated hardware, and choose NVMe or block storage. Clusters can be resized; secure, high-availability clusters are an option. Oracle Cloud SQL queries data across Hadoop, Kafka, NoSQL, and object stores using Oracle SQL. The service also describes SparkR for working with HDFS, Spark DataFrames, and JDBC sources and building models in R. Access is available through the OCI Console, CLI, REST APIs, and SDKs. Block Volume data is encrypted at rest by default, and OCI IAM controls authentication and authorization. Oracle describes both short-lived clusters for specific tasks and long-running data lakes. The listed service fee is 0.02 USD per OCPU per hour; compute and storage charges apply separately. Oracle lists separate hourly rates for compute shapes and Cloud SQL. Clusters need release updates to remain current and supported.

Unilake is an open-source platform for integrating, managing, governing, and querying data across a unified lakehouse. It supports more than 350 pre-built connectors, plus custom integrations built with Unilake, Airbyte, or Meltano SDKs, including a low-code option. Loads can be full or incremental, with streaming ingestion, API flattening, privacy transformations, and SCD type 2 tables. Its query interface can combine Iceberg, Delta, Hudi, Kudu, and Paimon data; analytics use StarRocks and support SQL, object storage, and REST API endpoints. Management features include cataloging, search, lineage, data-quality checks, and automated classification. Governance includes attribute-based and context-aware controls, policies as code, and just-in-time access. Unilake supports SAML 2.0 or OpenID Connect, SCIM provisioning, exportable audit logs, and customer-managed keys. Deployment options range from public, private, and hybrid cloud to on-premises, with Kubernetes and storage among the minimum requirements. The project is in active development and working toward an Alpha release; its AI/ML product is on the roadmap. It is licensed under AGPL 3.0 and EUPL 1.2.