已创建维度与事实表，如何在SQL Server中创建无需处理的虚拟OLAP Cube？

Absolutely, you can build a "virtual" Cube in SQL Server Analysis Services (SSAS) that only defines structural metadata (dimensions, hierarchies, measures) without tying to physical data or requiring processing. Here are two practical approaches tailored to your scenario:

Approach 1: Build an Empty Cube with Manual Metadata Definition #

This method creates a Cube entirely from scratch without linking to any existing data sources, so it never needs processing.

• Start in your SSAS project and launch the Cube Wizard. Select the option to Create an empty cube and proceed.
• Add virtual dimensions:
  • Create new dimensions (or repurpose existing ones) in the Dimension Designer. Instead of binding attributes to physical table columns, manually define attribute names, data types, and hierarchies in the "Dimension Structure" tab. Remove any links to data source views (DSVs) if modifying existing dimensions.
• Add virtual measures and measure groups:
  • In the Cube Designer, add empty measure groups. For each measure, set a name, data type (e.g., Decimal, Integer), and aggregation function (e.g., Sum, Count)—no need to map to a fact table column.
• Deploy the Cube to your SSAS instance. Since there's no data to process, you can skip the processing step entirely. The Cube will exist in the instance with a complete structural definition, but queries will return empty results.

Approach 2: Repurpose Existing Dimensions/Facts for a No-Data Cube #

If you already have existing dimension and fact table definitions in your SSAS project, you can adapt them to avoid physical data processing:

• For your existing dimensions:
  • In the Dimension Designer, go to the Data Source View tab and remove the association with physical tables. Switch to the "Dimension Structure" tab and ensure attributes are defined manually (you can keep their names/hierarchies, just unlink from columns).
• For the Cube's measure groups:
  • In the Cube Designer, modify each measure group to not bind to a physical fact table. Replace column bindings for measures with manual definitions (name, data type, aggregation).
• Set the storage mode for all measure groups to MOLAP (optional, but ensures no runtime dependency on physical data). When deploying, choose Process Structure Only instead of full processing—this will create the Cube's metadata without loading any data.

Key Notes #

• This virtual Cube is perfect for testing metadata structures, building report templates, or acting as a placeholder for future development.
• Queries against the Cube will return empty result sets, as there's no underlying data to retrieve.
• Be careful not to run full processing on the Cube later, as that will attempt to load data if you re-link to physical tables accidentally.

内容的提问来源于stack exchange，提问作者phalondon