e6 Query Engine
e6data.com
- Linux
- Web
- API
- Self-hosted
What e6 Query Engine is
e6 Query Engine runs SQL and AI workloads directly on existing lakehouse storage, avoiding a separate warehouse copy. It supports joins, window functions, and aggregations on large fact tables without down-sampling or pre-aggregation. Vector search uses the same tables and supports cosine similarity for semantic lookups. The product page claims up to 10x faster queries at p95 and more than 1,000 QPS at p95 under two seconds. Compute scales in one-vCPU increments, with configurable autoscaling floors and ceilings. Per-cluster thresholds can log, alert on, or cancel queries in real time. Listed storage includes S3, ADLS Gen2, and GCS, with Iceberg, Delta, and Hudi formats. JDBC, ODBC, Python, and BI tools are supported. Deployment can be serverless or inside a customer Kubernetes cluster in a VPC, including on-premises, hybrid, air-gapped, and sovereign environments. Pay-as-you-go compute is $0.18 per contact, billed at $0.17 per vCPU-hour, with no minimum commitment.
Who it suits
e6 Query Engine suits data teams querying lakehouse storage across SQL, AI, and vector workloads. Its deployment and security options also address controlled or isolated environments.
Good to know, and what to watch
- Queries lakehouse data without warehouse copying
- Supports SQL and vector search together
- Autoscaling uses one-vCPU increments
- Real-time query guardrails
- Offers air-gapped and sovereign deployment
- No free plan
- Consumption pricing requires metering
- Setup-specific support goes through assigned staff
Plans
Every plan and priceOutcome-Based Pricing
Pay for Compute
The verdict
e6 Query Engine combines federated SQL and vector search over existing lakehouse tables, with bounded scaling and query controls. Its varied deployment models fit organizations with specific cloud, on-premises, or isolation requirements.
What the maker publishes
- Compatibility
- The product page lists Databricks, Snowflake, Trino, SageMaker, and Microsoft Fabric as platforms it runs alongside. source
- Data handling
- The security page says the data plane runs in the customer's cloud account, while the control plane receives metadata and metrics rather than customer data. source
- Deployment
- It can run serverless or inside the customer's Kubernetes cluster in a VPC; the page also lists on-premises, hybrid, air-gapped, and sovereign environments. source
- Founded
- 2021 source
- Headquarters
- San Francisco, California, USA source
- Interfaces
- Listed client interfaces include JDBC, ODBC, Python, and BI tools. source
- Pricing limits
- The pricing page gives a consumption rate of $0.175 per vCPU-hour and says bring-your-own-cloud pricing is by contact. source
- Purpose
- e6 Query Engine runs SQL and AI workloads directly against existing lakehouse storage without copying data into a separate warehouse. source
- Query guardrails
- Per-cluster thresholds can log, alert on, or cancel a query in real time. source
- Scaling
- Compute capacity scales in 1-vCPU increments, and autoscaling can be bounded with configured floors and ceilings. source
- Security certifications
- The Security & Trust page lists SOC 2 Type II, ISO 27001, and GDPR. source
- SQL features
- It supports joins, window functions, and aggregations over large fact tables without down-sampling or pre-aggregation. source
- SQL performance
- The product page claims up to 10x faster queries at p95 and 1,000+ QPS at p95 under 2 seconds. source
- Storage and formats
- The product page lists S3, ADLS Gen2, and GCS, and the Iceberg, Delta, and Hudi table formats. source
- Support
- The documentation directs customers to their e6data CSM or support engineer for setup-specific questions. source
- Vector search
- Vector search runs on the same tables as SQL and supports cosine similarity for semantic lookups. source
Checked against the maker's pages on 30 Sept 2026.