Comparing OLAP and MTDI cubes

The following table lists a comparison between Online Analytical Processing (OLAP) and Multi-Table Data Import (MTDI) cubes.

	OLAP (In-Memory, Intelligent Cube)	MTDI (In-Memory)	MTDI (Live Connect)
Data Retrieval	An in-memory version of report data that can be manipulated by the MicroStrategy Analytical Engine. It retrieves report results from a schema source and stores it in memory. Visualizations in the dashboard that use data from the cube display subsets of the stored report, which increases the speed in which data appears.	Retrieves dataset results from a data source and stores them in-memory. Visualization in the dashboard that use data from the dataset display subsets of the stored results, which increases the speed in which data appears.	Retrieves visualization results directly from a data source. Performance when you drag dataset objects onto visualizations using Live Connect mode can be slower than in-memory mode because the system retrieves dataset results before they appear. However, in some cases, Live Connect can perform better as a dashboard will spend less time in SQL execution and data transfer phases performing targeted data fetch.
Data Volume	Each partition can have up to 2 billion rows.	Each partition can have up to 2 billion rows.	Individual visualizations are limited by the amount of rows a single report can fetch from the warehouse.
Schema	A single table dataset, pre-joined	Multiple tables, not pre-joined. Closer to raw data staged in-memory; Multiple in-memory fact tables, including varying grains, many-to-many relationship tables, and entity-relation model semantics. Starting in MicroStrategy 2021 Update 3, users can control attribute relationship models. For more information, see KB485145.	Multiple tables, not pre-joined Schema for MTDI dataset is loaded from Metadata at dashboard execution time, including varying grains, many-to-many relationship tables, and entity-relation model semantics. Starting in MicroStrategy 2021 Update 3, users can control attribute relationship models. For more information, see KB485145.
Definition	User specifies the definition of the cube on a Report Template with Attributes, Metrics, and Filter	User specifies the tables to load into memory; Attributes/Metrics are mapped to these tables	User specifies the tables for SQL engine to consider when generating SQL query; Attributes/Metrics are mapped to these tables.
Query Generation	Intelligence Server generates the SQL to execute against the data source (can use Multi-Source) By default, final pass of SQL joins lookup tables to fact table and/or metric temp tables to retrieve attribute descriptions	Queries submitted to the RDBMS are simple SELECT statements against the tables—no joins among tables	Queries submitted to the RDBMS are multi-pass SQL statements per visualization.
Partition	Available	Available	Not applicable
Parallel	Parallel Data Fetch Option set at Intelligent Cube level	Maximum Parallel Queries Per Report set at Project level	Maximum Parallel Queries Per Report set at Project level