Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

Making the Decision

Multidimensional vs. Tabular


Dustin Ryan
Data Platform Solution Architect, Microsoft
duryan@microsoft.com
Amount of Data
Server Resources
Disk Space

4
Tabular Multidimensional
• In-Memory Technology (x- • Pre-Aggregated Data From Disk
Velocity) • Can Store Very Large Amounts
• Can Store Large Amounts of of Data
Data • Uses Aggregations to Increase
• No Aggregations. Column- Query Performance
Based Storage. • Data Compression 3x
• Data Compression 10x

5
Scalability

Tabular Model Multidimensional Cube


Aggregate Data
Detail Data
Cache Results

7
Tabular Multidimensional
• Generally Speaking Tabular will • Pre-Aggregated Data From Disk
Perform Faster • Can Store Very Large Amounts
• Tabular Engine Does Not of Data
Require a Great Deal of • Uses Aggregations to Increase
Performance Tuning Query Performance
• Best at Returning Low
• Often Faster Than Tabular When
Granularity Data Pulling From Warm Cache

8
9
Dimension Design
Calculation Creation
Key Performance Indicators

10
Tabular Multidimensional
• Much less time to develop • More planning usually required
• Can be created from Power • No upgrade path
Pivot model • Requires dimensional model
• Does not usually require
dimensional model but still
recommended
• Much simpler interface (similar
to Power BI Desktop)

11
Time to Develop

Tabular Model Multidimensional Cube


13
Requires more training
Requires more planning
Translating existing skills

14
Tabular Multidimensional
• Uses DAX (Data Analysis Expressions) • Uses MDX (Multi-Dimensional
for Query Language Expressions) for Query Language
• If You Know Excel Formulas Then • Difficult to Learn but has Benefits
DAX Will Be Easy (Navigating Hierarchies)
• Simple Drag and Drop KPI Creation • More Complex KPI Creation
• Relationships are simple (no • Multiple relationship types
composite keys) • Many developers have been
• If you are familiar with modeling in designing cubes for a long time
Power BI Desktop, then your skills
will translate

15
Learning Curve

Tabular Model Multidimensional Cube


17
After you Choose…
• There is no migration path between the technologies
• Tabular Best on its Own Machine, Not a Good Candidate for
a Shared Service

18
Consider Tabular…
If You Have a Short Development Timeline
If You are Working with a Plethora of Memory
If Your Have Many Disparate Data Sources
If Users Need to Query Large Amounts of Detail Data

Consider Multidimensional…
If You are Using SQL Server 2008 R2 or Earlier
If Your Data Model cannot fit into memory
If You Have a Complex Data Model
If You Need Multidimensional only features (Actions, Data Mining, Writeback, Translations)

19
Multidimensional Tabular
Actions Yes No
Aggregations Yes No
Calculated Measures Yes Yes

Custom Assemblies Yes No

Custom Rollups Yes No


Distinct Count Yes Yes (via DAX)
Drillthrough Yes Yes

Hierarchies Yes Yes


Key Performance Yes Yes
Indicators
Linked objects Yes No
Many-to-many Yes Yes (via DAX)
relationships
Parent-child Hierarchies Yes Yes (via DAX)

Partitions Yes Yes


Perspectives Yes Yes
Semi-additive Measures Yes Yes (via DAX)

Translations Yes No
User-defined Hierarchies Yes Yes

Writeback Yes No
20

You might also like