AetherLake

AetherLake is a self-hosted Kubernetes data platform that bundles lakehouse storage, streaming, SQL, orchestration, BI, and identity into one control panel. It suits teams that already operate Kubernetes and want an integrated open-source-oriented stack rather than a single-purpose service. The trade-off is substantial deployment and production hardening work, plus licensing boundaries for commercial use.

Overview

AetherLake combines MinIO storage, Trino federated SQL, Apache Polaris catalogs, Kafka and Flink streaming, dbt transformations, Airflow orchestration, Superset BI, Spark, Milvus, and observability and identity services including Prometheus, Grafana, PostgreSQL, and Keycloak. That breadth can reduce the need to stitch together separate tools, but it also creates a large operational footprint for the team running it.

The Next.js control panel presents pod health and resource consumption alongside SQL, catalog, Kafka, Flink, dbt, and observability views. Operators can restart services and open workspaces for dbt, Kafka, Flink SQL, Trino queries, and Iceberg catalogs. English and Turkish interfaces make it more accessible to teams using either language.

Key features

Lakehouse pipelines and streaming

Kafka and Flink are enabled by default. The documented workflow streams Kafka data into Iceberg tables and lets Trino query Kafka topics, while dbt examples move data through Bronze, Silver, and Gold stages using Kafka, Flink, Parquet, Iceberg, Superset, and Trino. This is a coherent path from ingestion to transformation and reporting, provided your team is comfortable managing each service.

Identity and operations

Keycloak supplies OIDC single sign-on across services, with data-admin, data-scientist, and data-engineer roles. The panel's health, resource, and restart controls are useful for day-to-day administration, but they do not remove the need for Kubernetes expertise. External Kafka access is documented with TLS and SCRAM-SHA-512 authentication.

AI-assisted administration

The MCP server can expose platform status, logs and restarts, Trino queries, Polaris catalogs, and Airflow DAGs to assistants such as Claude Desktop and Cursor. That can shorten diagnostic and query workflows, while still requiring careful control over what an assistant is allowed to inspect or restart.

Deployment and security

Quick start deployment requires Kubernetes, Helm 3.12 or later, kubectl, Docker, and an NGINX ingress controller. Terraform and Helm materials cover AWS, Azure, and Google Cloud. Credentials are randomly generated per installation, but production exposure requires replacing placeholders, setting production secrets, overriding the MCP server's AIRFLOW_AUTH, and installing real ingress TLS certificates. Anonymous telemetry records installation counts, detected cloud environments, and component health; it excludes personal data, schemas, and dataset contents and can be disabled.

Pricing

The free plan is appropriate for internal teams, prototypes, development, and testing, with no subscription charge. It is not a free commercial production license: hosted or managed commercial offerings require a separate agreement. Releases convert to Apache-2.0 after four years under the stated BUSL-1.1 terms.

Platforms

AetherLake runs self-hosted on Kubernetes and is managed through a web interface. Deployment configurations target AWS, Azure, and Google Cloud, but the prerequisites still place cluster, ingress, and container responsibility on the operator.

Who it's for

Choose AetherLake if you have Kubernetes capability and need storage, open table formats, SQL analytics, streaming ingestion, governance catalogs, orchestration, BI, and identity in one self-hosted environment. It is a poor match for a small team that wants a managed lakehouse, minimal infrastructure work, or a straightforward commercial license.

Pros and cons

• Pro: Broad stack coverage puts storage, catalog, SQL, streaming, transformations, workflows, BI, and identity behind one interface.
• Pro: Kafka and Flink defaults, plus Iceberg and Trino workflows, support an end-to-end streaming lakehouse pattern.
• Pro: OIDC roles, TLS/SCRAM guidance, and operational health controls provide a useful administration foundation.
• Con: Kubernetes, Helm, Docker, and ingress prerequisites make deployment unsuitable for teams without platform operations capacity.
• Con: Production use requires replacing placeholders, configuring secrets and TLS, and securing MCP access rather than simply exposing the quick-start installation.
• Con: Commercial hosted or managed use falls outside the free terms and may require case-by-case licensing.

Alternatives

See the Data Lakehouse Platforms list for more options.

• Oracle Autonomous AI Lakehouse is a better fit when you want a freemium cloud service with an always-free plan and API, Linux, self-hosted, and web access.
• Starburst Data Platform suits teams wanting a managed-oriented option with a free tier for up to three clusters and paid expansion.
• Starburst Galaxy is the choice when web and API access matter more than running the full platform yourself, with a free tier capped at three clusters.
• Amazon SageMaker Autopilot fits pay-as-you-go model-building needs rather than operating a complete self-hosted lakehouse.
• Apache Hudi is preferable when you want a focused, free, self-hosted lakehouse project for Linux instead of an assembled service suite.
• Bauplan is a better starting point for a hosted sandbox with CLI, SDK, API, and community support.
• Databricks Notebooks fits users seeking a web workspace with a free edition and limited serverless compute.
• Google Cloud Lakehouse for Apache Iceberg is suited to teams wanting Google Cloud's managed Iceberg metadata services rather than self-hosting.

Verdict

Pick AetherLake if your team can operate Kubernetes and wants a broad, self-hosted data stack with streaming, SQL, catalogs, orchestration, BI, and identity already coordinated. Look elsewhere if you need turnkey managed operations or unrestricted commercial hosting without a separate license.