Data Plane
A Data Plane is an external data warehouse or lakehouse where Ascend persists data. It's the infrastructure platform that provides both storage (where your tables live) and compute (where your queries run) for a given Flow. A Flow can only run on one Data Plane, but a Project can have many Flows that run on many Data Planes, enabling a data mesh architecture.
What the Data Plane does
The Data Plane is responsible for:
- Storage: Persisting all data from your Flow's Components as tables or views
- Compute: Data processing operations via execution of SQL and Python code
- Data ownership: Your data lives in external infrastructure, not in the Ascend platform
Ascend sits on top of the Data Plane
The following applies to BigQuery, Databricks, and Snowflake Data Planes. DuckDB with DuckLake keeps compute within Ascend Flow runners.
Ascend acts as an orchestration and metadata layer on top of your Data Plane. When you build a data pipeline in Ascend, Ascend generates and optimizes the queries and transformation logic. Your Data Plane executes those queries using its compute resources. Then, your Data Plane stores the resulting tables in its storage layer, Finally, Ascend tracks the metadata, lineage, and orchestration state.
This architecture is called pushdown execution — Ascend pushes all data processing down to your Data Plane's native engine for maximum performance.
What the Data Plane does not do
The Data Plane is not responsible for:
- Orchestration: Ascend handles scheduling, dependencies, and workflow coordination
- Metadata management: Ascend tracks lineage, schemas, and data quality
- Observability: Ascend monitors Deployments and tracks pipeline performance and errors
- Access control: While your Data Plane has its own permissions, Ascend manages user access to the Ascend platform
Compute and storage
Compute
Compute is the processing power that executes queries and transformations. When Ascend runs a Transform Component:
- BigQuery: BigQuery's distributed compute engine processes the SQL query
- Databricks: Databricks clusters run the Spark jobs or SQL queries
- DuckLake: The query runs in the Ascend Flow runner on the DuckDB engine
- Snowflake: Snowflake's virtual warehouses execute the query
The compute resources are provided and billed by your Data Plane (except DuckDB with DuckLake, which is managed by Ascend).
Storage
Storage is where your data tables are permanently stored. For example, when you create and run a Transform titled sales_summary, Ascend automatically creates a sales_summary table in your Data Plane's storage:
- BigQuery: The table exists in a BigQuery dataset in your GCP project
- Databricks: The table exists in a Unity Catalog schema or Hive metastore
- DuckLake: The table exists in a DuckLake catalog
- Snowflake: The table exists in a Snowflake schema in your account
You can query these tables directly from your Data Plane using any tool that connects to it — they're real tables in your infrastructure.
Ascend-managed vs. external Data Planes
External Data Planes (you manage)
For BigQuery, Databricks, and Snowflake Data Planes, you own the infrastructure including accounts, projects, and workspaces. You control access through credentials, roles, and network policies. You manage costs as compute and storage billing goes through your cloud provider, and you handle availability based on your platform's SLA.
Ascend connects to your existing infrastructure via secure credentials and executes operations on your behalf.
DuckDB with DuckLake (Ascend-managed)
DuckDB with DuckLake is Ascend's fully managed Data Plane option. Ascend owns the infrastructure with no setup required and manages compute and storage with automatic scaling and optimization.
Unlike other Data Planes, DuckDB with DuckLake doesn't persist data permanently; this means lower storage costs and is ideal for use cases where you don't need long-term data retention. Billing is simplified as it's included in your Ascend plan, and there's zero configuration needed — just start building Flows.
Although every free trial Instance of Ascend comes bootstrapped with Ascend-managed DuckLake, you can also bring your own DuckDB with DuckLake Data Plane to the Ascend platform.
Supported Data Planes
Ascend currently supports the following Data Planes:
- BigQuery
- Databricks
- DuckDB with DuckLake (Ascend-managed)
- Snowflake
Summary
- Your Data Plane is external infrastructure where data is stored and processed
- Ascend sits on top of your Data Plane and pushes down all data processing to your Data Plane's native engine (except for DuckDB with DuckLake)
- Storage = where tables persist; Compute = where queries execute
- You can access your data directly through your Data Plane in the form of real, queryable tables