AetherLake

AetherLake is a self-hosted, Kubernetes-native data lakehouse that brings storage, cataloging, federated SQL, streaming, orchestration, BI, and identity into a web control panel. Its listed stack includes MinIO, Trino, Apache Polaris, Kafka, Flink, dbt, Airflow, Superset, Spark, Milvus, Prometheus, Grafana, PostgreSQL, and Keycloak. The panel shows pod health and resource use, supports service restarts, and provides workspaces for dbt, Kafka, Flink SQL, Trino queries, and Iceberg catalogs. Kafka and Flink are enabled by default, and project examples describe streaming Kafka data into Iceberg tables and querying Kafka topics through Trino. Keycloak provides OIDC single sign-on with data-admin, data-scientist, and data-engineer role names. Deployment prerequisites include Kubernetes, Helm 3.12 or later, kubectl, Docker, and an NGINX ingress controller. Materials cover AWS, Azure, and Google Cloud. The BUSL-1.1 license allows internal, evaluation, development, testing, and non-commercial use at no cost; commercial use requires a separate license.

AetherLake combines a broad data stack with a central web control panel, but it has substantial deployment prerequisites. Its free license covers internal and non-commercial use; commercial use requires separate licensing.

AetherLake suits teams that can operate Kubernetes and want a self-hosted lakehouse with streaming, SQL, catalog, and orchestration components. Its free terms cover internal, evaluation, development, testing, and non-commercial use.

AetherLake is a self-hosted data lakehouse that combines storage, SQL, streaming, orchestration, and BI services behind a web control panel. It suits teams already equipped to run Kubernetes and wanting one environment for a broad data stack. Its appeal is the range of tools it brings together; that breadth comes with meaningful deployment and production-setup work.

Overview

The stack spans object storage, cataloging, federated SQL, stream processing, workflow orchestration, business intelligence, and identity. Components include MinIO, Trino, Apache Polaris, Kafka, Flink, dbt, Airflow, Superset, Spark, Milvus, Prometheus, Grafana, PostgreSQL, and Keycloak. For a team building across these areas, that gives a common platform to organize related workflows. It also means taking responsibility for a large collection of services rather than relying on a single managed product.

A Next.js control panel brings together platform status, pod health, resource use, SQL, catalog, Kafka, Flink, dbt, and observability views. It can restart services and provides workspaces for dbt, Kafka, Flink SQL, Trino queries, and Iceberg catalogs. The English and Turkish interface gives teams in either language a shared management surface.

Key features

Lakehouse workflows and streaming

Kafka and Flink are enabled by default. Project examples describe moving Kafka data into Iceberg tables and querying Kafka topics through Trino. dbt examples use Bronze, Silver, and Gold stages with Kafka, Flink, Parquet, Iceberg, Superset, and Trino. These pieces make AetherLake a plausible foundation for teams connecting ingestion, transformation, and analytics; teams that need only a query engine or one processing service may be taking on more stack than they need.

Identity, security, and operations

Keycloak provides OIDC single sign-on across services, with data-admin, data-scientist, and data-engineer roles. The documented external Kafka listener uses TLS and SCRAM-SHA-512 authentication. Credentials are randomly generated for each installation, but the project advises replacing placeholders, setting production secrets—including Control Panel secrets and the MCP server's AIRFLOW_AUTH—and configuring real ingress TLS certificates before exposing a cluster. That preparation is essential for production use.

An anonymous heartbeat tracks installation counts, detected cloud environments, and component health. The project says it excludes personal information, schemas, and dataset contents, and can be disabled. Teams evaluating telemetry should account for that heartbeat and its opt-out when deciding how to configure deployments.

AI assistant access

The MCP server gives compatible assistants such as Claude and Cursor tools to inspect platform status and logs, restart services, run Trino queries, access Polaris catalogs, and inspect Airflow DAGs. This can make routine platform interaction available through an assistant, while the documented access also makes production authentication settings particularly important.

Pricing

AetherLake is free for internal, evaluation, development, testing, and non-commercial use under its BUSL-1.1 license. That makes it possible to assess and self-host the platform without a subscription price, but the free terms do not cover commercial use. Internal deployment and evaluation are the clearest fits for this plan.

Commercial use requires a separate license with custom pricing. Licensing is handled case by case and may include revenue-sharing terms, so commercial teams should resolve those terms before adopting AetherLake in a business offering. The license permits internal use, modification, and self-hosting; releases convert to Apache-2.0 after four years.

Platforms

AetherLake is self-hosted and accessed through the web control panel. The quick start requires Kubernetes, Helm 3.12 or later, kubectl, Docker, and an NGINX ingress controller. Those prerequisites make it a better match for teams with Kubernetes operations capacity than for individuals or organizations seeking a quick, managed workspace. The project documents Terraform and Helm deployment materials for AWS, Azure, and Google Cloud Platform.

Who it's for

AetherLake is best for data engineering teams that want to self-host a platform spanning lakehouse storage and catalogs, SQL, streaming, orchestration, BI, and identity—and already have the infrastructure skills to operate Kubernetes. It is also worth considering for internal or non-commercial evaluation of an integrated stack. Teams that need a managed service, lack Kubernetes capacity, or intend commercial use without first securing a license should look elsewhere.

Pros and cons

Pros

• Broad stack in one environment: storage, catalog, SQL, streaming, orchestration, BI, identity, and observability components support several connected data workflows.
• Useful operational workspace: pod health, resource use, service restarts, and focused workspaces bring day-to-day platform views into one control panel.
• Streaming is part of the default setup: Kafka and Flink are enabled by default, with examples connecting Kafka, Iceberg, and Trino.
• Free internal evaluation: internal, development, testing, and non-commercial use are covered without a plan fee.

Cons

• Substantial deployment prerequisites: Kubernetes, Helm, kubectl, Docker, and NGINX ingress are required, raising the operational bar.
• Production exposure needs deliberate configuration: placeholder values, secrets, and ingress certificates must be replaced or set before exposing a cluster.
• Commercial rights require a separate agreement: commercial licensing has custom pricing and may involve revenue sharing.
• Telemetry is enabled unless disabled: the anonymous heartbeat can be turned off, but teams must make that choice as part of deployment.

Alternatives

Data Lakehouse Platforms is a useful starting point for comparing options across the category.

• Amazon SageMaker Autopilot is a paid API and web option with a free trial and pay-as-you-go pricing without minimum fees or upfront commitments; consider it when that pricing model suits better than self-hosting AetherLake.
• Bauplan offers a free sandbox with a full CLI, SDK, and API, community support, and public datasets or uploads in a shared environment; choose it when a shared sandbox fits better than operating a Kubernetes stack.
• Databricks Notebooks has a free web edition with one serverless workspace and limited compute size and usage; choose it when a serverless notebook workspace is a better fit than self-hosting a broad platform.
• LavaLake offers a free Core plan with up to 16 vCPU on one machine, plus a paid Team plan; consider it when a single-machine starting point suits the workload.
• Oracle Autonomous AI Lakehouse has an always-free plan for an unlimited time subject to capacity limits; consider it when that free offering is a better fit.
• Starburst Data Platform offers a free plan with up to three clusters for ad hoc queries in standard execution mode; choose it when that focused query setup meets the need.
• Starburst Galaxy also offers a free plan with up to three clusters for ad hoc queries in standard execution mode; consider it for that free cluster-based option.
• Apache Hudi is a free, self-hosted Linux lakehouse platform; consider it when that platform is the better match.

Verdict

Choose AetherLake if your data team can run Kubernetes and wants a self-hosted environment that brings lakehouse storage, streaming, SQL, orchestration, BI, and identity together. Its strongest reason to choose it is the breadth of connected tools under one control panel, with free terms for internal and non-commercial use. Look elsewhere if you need managed operations, a lighter deployment, or commercial use without a separately negotiated license.