Glossary

Definitions of the core e6data terms used throughout the documentation and Console.

Definitions of the terms you'll see throughout the e6data docs and Console.

Platform

e6data - the product: a lakehouse compute engine that runs SQL and AI workloads directly on your data lake.

Account - the top-level entity for your organization; billing, support, and membership are scoped to it.

Tenant - your organization within an account. Users belong to a tenant, and resources (workspaces, catalogs, clusters) belong to it.

Control Plane - the shared e6data Console where you manage your organization, workspaces, account, billing, and global access control. It doesn't process queries or touch your data.

Compute Plane - the per-workspace environment (opened with Open) that holds catalogs, clusters, the SQL Editor, and Run History. The compute that runs queries.

Workspace - an isolated environment for a team or project, with its own catalogs, clusters, access control, and run history.

Workspace Type - a workspace property with two values: Serverless or In Your VPC. See Deployment models.

Serverless - the deployment model where e6data runs and manages the compute.

In Your VPC - the deployment model where the Compute Plane runs inside your own Kubernetes cluster.

CloudPrem - e6data's name for an In Your VPC deployment, where the Compute Plane runs in your own cloud account and Kubernetes cluster.

Compute

Cluster - the compute that runs queries inside a workspace, selected by name in every connection.

Cluster Runtime (CRT) - the cluster's version, chosen when you create or update a cluster.

Cluster Size - a cluster's capacity, expressed as a size such as XS, S, or M.

Executor - a node that runs query work; cluster size is expressed in executor count.

Autoscaling - automatic adjustment of a cluster's executors between a configured minimum and maximum based on load.

Scale-to-zero - autoscaling with a minimum of zero executors: the cluster drops to no compute when idle and resumes on the next query.

Auto-suspend / auto-resume - a cluster suspends after an idle period (zero cost) and resumes automatically on the next query.

Execution mode - the path a query runs on: Standard (the full SQL surface) or Native (a vectorized path optimized for scan-heavy queries, with automatic fallback to Standard for unsupported features).

Result cache - on-cluster caching that serves repeated identical queries without re-executing them, with a configurable TTL.

Data

Catalog - a connection to a metastore (Hive Metastore, AWS Glue, Databricks Unity, Microsoft Fabric, Apache Polaris, Iceberg REST, or AWS S3 Tables) that lets e6data discover databases, tables, and columns. Read-only.

Table format - the on-disk layout of a table that e6data reads: Apache Iceberg, Delta Lake, or Hive tables.

Time travel - querying a table as it existed at an earlier snapshot or version (supported for Iceberg and Delta Lake).

Schema evolution - changing a table's columns over time (add, drop, rename) without rewriting existing data.

Partitioning - physically dividing a table's data (for example, by date or region) so queries can skip irrelevant partitions, known as partition pruning.

Metadata Service (MDS) - the metadata catalog service, upgradeable from Settings → Version Upgrades.

Platform Components - the shared workspace infrastructure (console, routing, and related services), upgraded together as a platform release.

Data governance

Data governance - access controls applied within a catalog: catalog-level access control, column masking, and row filtering.

Catalog privilege - a data-access grant that controls which databases, tables, rows, or columns a user can read within a specific catalog. Deny-by-default.

Column masking - a policy that obscures sensitive column values from users who lack explicit privilege.

Row filtering - a policy that restricts which rows a user can read, based on a condition.

Identity and access

Personal Access Token (PAT) - an e6pat_ token that authenticates the API as a human user.

Service Account - a non-human identity scoped to one workspace, authenticating with an e6sa_ API key.

Role - a named set of permissions. The built-in roles are Admin, Manager, and Viewer; custom roles are supported.

Role binding - attaches a role to a user, group, or service account, optionally scoped to a specific resource.

Group - a named collection of users (and, on the Compute Plane, service accounts) that roles are assigned to collectively.

SSO (single sign-on) - signing in through an external identity provider instead of a local e6data password.

Identity provider (IdP) - the external system that authenticates SSO users (for example, Okta, Microsoft Entra ID, or Google).

JIT provisioning - just-in-time account creation on a user's first SSO sign-in, with a default role.

Domain auto-join - automatic organization membership for users whose email domain matches the org's configured domain.

Support access - scoped, audited access granted to an e6data engineer (a Support User) for support and debugging, which you can enable, audit, and revoke.

Querying

SQL Editor - the in-browser console for writing and running queries.

Run History - the searchable log of queries run in a workspace (timing, status, user, query text).

SQL transpiler - translates SQL written for another dialect (Snowflake, Trino, Databricks) into e6data SQL before execution.

Networking

Endpoint - the network address (host and port) that clients use to connect to a cluster.

IP set - a named list of IP ranges (CIDRs) that controls which clients are allowed to connect.

TLS certificate - secures endpoint connections. e6data manages it for Serverless; you manage it for In Your VPC.

Connectivity and APIs

JDBC - standard database drivers for connecting BI tools and applications to a cluster.

PostgreSQL protocol - the PostgreSQL wire protocol, supported for client connections alongside JDBC and REST.

Management API - the REST API for organization and workspace management (workspaces, users, service accounts, version upgrades).

Operations

Observability export - exporting workspace metrics, logs, and traces to an external backend using the OpenTelemetry (OTLP) standard.