Organization Hierarchy

Singdata Lakehouse uses a four-tier organization structure to manage resources and data: Instance → Catalog → Schema → Data Objects.

Instance ├── Workspace ← Internal Catalog, resource isolation unit │ ├── Schema ← Namespace, organizes tables/views/functions │ │ └── Tables, Views, Functions, etc. │ └── External Schema ← Special Schema that maps an external Hive data source └── External Catalog ← External data catalog, entry point for federated queries └── External Schema └── External Table

Tier Descriptions

TierDescriptionReference
InstanceThe top-level resource unit, corresponding to one Singdata Lakehouse instance. Instance-level configuration is shared across all Workspaces.Managing Instances
WorkspaceThe resource isolation unit within an instance, essentially an internal Catalog. Contains independent users, permissions, compute clusters, and data objects. Different Workspaces are invisible to each other by default.Workspace
External CatalogAn external data catalog at the same level as a Workspace, mapping external data systems such as Hive, Databricks, and Snowflake for federated queries.External Catalog
SchemaThe namespace within a Workspace, used to organize data objects by business domain or data layer (e.g., ods, dwd, ads).Schema
External SchemaA special Schema within a Workspace that maps an external Hive data source, enabling direct queries without data migration.External Schema

Recommendations

When to create multiple Workspaces:

  • Isolating development, test, and production environments
  • Different business teams need independent permissions and compute resources
  • Business units that require separate billing

When to create multiple Schemas:

  • Dividing by data layer within the same team (ODS / DWD / ADS)
  • Dividing by business domain within the same Workspace (orders, users, products)
  • Granting permissions to different data collections as a whole

When to use External Catalog vs. External Schema:

ScenarioRecommendation
Querying external systems such as Hive, Databricks, or SnowflakeExternal Catalog
Mounting an external Hive database into the current Workspace for direct schema.table referencesExternal Schema
Federated queries across platforms without data migrationExternal Catalog

In-place data lake acceleration: Data stays in the original object storage (OSS/COS/S3/HDFS). Connect it to Lakehouse via External Catalog or External Schema, then use Lakehouse directly as a replacement for Spark/Hive for ETL processing, or as a replacement for Presto/Trino for ad-hoc queries — no data migration required, and you immediately gain Lakehouse's performance and SQL capabilities.

Object Model
Workspace
Singdata © 2024 Singdata, Inc. All Rights Reserved.