Microsoft SSAS: Should I Use Tabular or Multidimensional?

1
Helping Companies Learn From the Past, Manage the
Present and Shape the Future
www.senturus.com
Microsoft SSAS: Should I use Tabular
or Multidimensional?

2
This slide deck is part of a recorded webinar.
To view the FREE recording of this entire
presentation and download the slide deck, go to
http://info.senturus.com/2013-05-16-Microsoft-SSAS.html
The Senturus library has over 90 other recorded webinars,
whitepapers, and demonstrations on assorted business intelligence
topics which may interest you.
Go to the recorded resources

3
Agenda
• Introductions
• Company Overview
• Overview and History of SSAS
Multidimensional and Tabular
• Technical comparison of the two
versions
• Summary Recommendations
• Further Resources
• Q & A

4
Presenters Today
Greg Herrera
Co-Founder
CEO
David Shroyer
• Over 12 years experience
implementing BI solutions with
the Microsoft stack
• Worked for ProClarity, which was
later purchased by Microsoft
• Started “cube building career”
with Arbor Essbase.
• MBA

5
Who is Senturus ?
• Consulting firm specializing in Corporate Performance
Management
– Business Intelligence
– Tools of the Office of Finance
• Enterprise planning & budgeting
• Consolidate, close, report and file (CCRF)
– San Francisco Business Times Hall of Fame -- Four
consecutive years in Fast 100 list of fastest-growing
private companies in the Bay Area
• Experience
– 13-year focus on performance management
– More than 1,200 projects for 650+ clients
• People
– Business depth combined with technical expertise.
Former CFOs, CIOs, Controllers, Directors...
– DBAs with MBAs
www.Senturus.com 888.601.6010 info@senturus.com

7
SSAS Multidimensional
• MSOLAP debut with the release of SQL
Server 7.0
• The technology was purchased from an
Israeli company Panorama
• It quickly became one of the most used
OLAP engines, primarily because it was
included in the SQL Server suite and
there was no additional cost

8
SSAS Multidimensional
• SSAS was completely overhauled with
the release of SQL Server 2005
• This new version allowed for “sub
cubes” with the Scope statement. This
radically increased the functionality of
the cubes
• SSAS 2008R2 and 2012 have primarily
be concerned with query performance
and scalability

9
PowerPivot
• With the release of Excel 2010 came an
add-in called PowerPivot
• PowerPivot uses a local instance of
Analysis Services with the new xVelocity
engine
• The xVelocity in memory engine can
greatly increase query performance

10
SSAS vs PowerPivot
• PowerPivot is a client based tool. All
data is stored in the Excel workbooks.
• PowerPivot does not have any security,
other than securing a workbook on the
file system
• Microsoft position PowerPivot as “Self-
Service BI” whereas SSAS
Multidimensional is “Corporate BI”

11
SSAS Tabular
• With the release of SQL Server 2012
came the “server version” of PowerPivot
which is a “tabular” instance of SSAS on
the server
• Models can be created directly on the
server or PowerPivot models can be
imported

12

13
Languages
• Multidimensional cubes use MDX for
calculations, security, and queries
• Tabular uses DAX for calculations and
queries. It is also MDX compliant,
which means that MDX based tools, like
Excel, can query Tabular cubes

14
Multidimensional Modeling
• Based on a traditional star schema
methodology. This includes dimension
tables and fact tables. (Kimball
methodology)
• Multidimensional models can include
advanced calculations and sophisticated
business logic.

15
Tabular Modeling
• Organizes data into related tables
• Because tables do not need to be
designated as “dimensions” or “facts”
the development time is less with
tabular because all related tables serve
both roles

16
Data Model Considerations
• One criteria to use when making the
choice between versions is to look at
your data warehouse data model. The
relationships and characteristics of the
data can help guide you to the proper
product

17
Data Considerations
• One to Many Relationships
– Both models handle these out of the box.
Although with tabular you can only join on
ONE field
• Many to Many Relationships
– Multidimensional handles this out of the
box. Tabular does not allow for M2M
relationships but you can “work around it”
by using DAX calculations

18
Hierarchies
• Standard Hierarchies
– Multidimensional needs to have hierarchy
attribute relationships to be defined. The
cube then uses those hierarchies to pre-
calculated “totals” at those levels
– Tabular does not need any defined
relationships. Because it is “in memory”
there is no need for pre-calculated
aggregations

19
Hierarchies
• Ragged Hierarchies
– Multidimensional provides support for
ragged hierarchies by hiding any missing
values in the hierarchy
– Tabular does not support ragged
hierarchies

20
Hierarchies
• Parent/Child Hierarchies
– Multidimensional allows for parent/child
hierarchies out of the box. These are useful for
supervisor/employee dimensions as well a chart of
accounts dimension
– Tabular does not support the definition of parent-
child hierarchies; however, the DAX language
provides a set of functions that allows users to
explore parent-child hierarchies and to use these
hierarchies in formulas.

21
Financial Hierarchies
• Unary Operators
– Multidimensional offers out of the box support for
unary operators. This is very important for chart
of account hierarchies
– Tabular does not support unary operators
• Semi-additive Measures
– Multidimensional has built in aggregation functions
– Tabular handles these measures but you need to
use separate DAX functions for each level such as
ClosingBalanceMonth

22

23
Additional Features
• Perspectives – available in both models
• Translations – only available with
multidimensional
• Actions – only available in multidimensional
• Drillthrough – available in both models
• Write-back – only available in
multidimensional
• Role Playing Dimensions – not available in
Tabular

24
Additional Features
• Formatting Measures – Tabular does
not allow for custom formatting
• Display Folders – Tabular does not
allow for display folders
• Naming the All Level – Tabular does not
allow for the naming of the All Level

25
Calculations
• Both models have a wide variety of
calculation possibilities
• Multidimensional allows for extra
calculation properties such as font color
and background color based on logic
• Multidimensional allows for calculations
that are not located on the measures
dimension

26
Shell Dimensions
• In multidimensional you can create
“shell dimensions” which can store
“global” calculations
• This is often used for Date Calculations
such as year over year variances and
YTD
• These can also be used for unit of
measure conversions and other global
calculations

27
Date Calculations
Adding a date calculations dimension to a cube can
make creating dashboard type reports very easy

28
Date Calculations
• With tabular you have two choices
– You build your date calculations into the
measures dimension – Qty, Qty LY, Qty LY
Var, Qty LY Var %, Sales, Sales LY, etc
– Or you create a faux shell dimension and include
massive nested IIF statements to trap for every
possible intersection. (which is just not
reasonable) This also causes problem in that you
can only format a measure so shell dimensions are
just not suited for Tabular

29
SSAS Advanced Course
• In case you are interested in learning how to use
shell dimensions and create a date calculations
dimension, in multidimensional, we have an
upcoming class which teaches these concepts

30
MDX vs DAX
• MDX (Multi-Dimensional Expressions) is used
in Multidimensional and it is a language of
hierarchies and dimensions
• DAX (Data Analysis Expressions) is used in
Tabular and it is more columnar based
• One of the reasons why Microsoft invested
heavily into the xVelocity technology was
because they felt that MDX was too difficult
for the average DBA

31
MDX vs DAX
• MDX can be hard to conceptualize but it
has the advantage of being able to
navigate hierarchies
• Percent of Parent calculation in MDX
[Measures].[Sales Amount] /
([Product].[Product Categories].CurrentMember.Parent,
[Measures].[Sales Amount])

32
MDX vs DAX
• Here is the Percent of Parent calculation
in DAX
IF(
ISFILTERED(Product[Product])
,[Sales]/CALCULATE([Sales],ALL(Product[Product]))
,IF(
ISFILTERED(Product[Subcategory])
,[Sales]/CALCULATE([Sales],ALL(Product[Subcategory]))
,1
)
)

33

34
Named Sets
• In multidimensional named sets allow
you to retrieve common sets of
members that can be used in reporting,
such as “current day”, “current month”,
or “last 12 months”
• Named sets are not available in Tabular
but you can duplicate some of their
functionality using attributes and sql
date functions

35
Named Sets
• Named sets are critical for automating
date selection in dashboards
• Named sets will also be covered in our
SSAS Advanced Topics course

36
Data Storage
• Multidimensional stores data in data
partitions and then it creates
aggregations within the partitions
– Compression is normally 3X to 4X
• Tabular loads data into memory so
there is no need for aggregations
– Compression is normally 10X

37
Scalability
• Multidimensional can scale to billions of
records. This is accomplished through
the use of data partitioning
• Tabular is limited by the amount of
memory on the server
– The lower the cardinality of the data the
more records can be fit into memory

38
Processing Performance
• Multidimensional has dependencies
which can make cube processing
somewhat tricky
– All dimensions must be processed before
the fact tables
– Updating of dimensions drops all
aggregations and therefore they need to
be recalculated
– Full process of dimensions drops the data

39
Processing Performance
• Tabular is just a “collection of tables” so
there are no processing dependencies
• Therefore it is much more flexible when
it comes to processing
• You can partition the data within the
tabular model so that you don‟t have to
process the full data set each day but
the partitions cannot be processed in
parallel

40
Query Performance
• Multidimensional query performance is
dependent upon the partition design
and the aggregation design
• Tabular query performance does not
need to be tuned
• Generally tabular will provide for better
query performance especially if the
users are pulling large amounts of detail
data

41
Security
• Multidimensional bases security on
members within a dimension
• Tabular bases security on rows in a
table
• Both version support dynamic security
• Multidimensional also allows for visual
totals and allowing users to use
drillthrough

42
Microsoft Client Tools
• Microsoft Excel can connect to both models.
– Although the query performance benefits
of Tabular are lost when you query with
MDX
• PerformancePoint can only connect to
Tabular with the use of a custom data source
• PowerView can only connect to Tabular
• Most third party products can connect to both
models because Tabular is MDX compliant

43
The Senturus library has over 90 other recorded
webinars, whitepapers, and demonstrations on assorted business
intelligence topics which may interest you.

44
Demonstration
• How do these models look to the user
in Excel?

45
Summary Recommendations for Tabular
• You should consider Tabular if:
– You have a very short development time
span
– Your data model is fairly simple
– Your users are pulling large amounts of
detail data and query performance is
important
– Cube has a lot of Distinct Counts
– You want to use PowerView in SharePoint

46
Things to Consider
• Once you choose a path you cannot
migrate to the other version without
starting over
• You will not be able to “merge” data
between Tabular and Multidimensional
cubes
• Tabular can be risky if requirements
change mid-way through the project

47
Further Resources: New Training Course
Advanced Topics in Microsoft SSAS
June 11 (7am-3pm PDT)
• Learn how to
– create global date calculations,
– create named sets, which can help automate dashboards
and reports
– use „many to many‟ relationships
• $695
• For more information and to register:
http://www.senturus.com/Microsoft_SSAS.php

48

49
Helping Companies Learn From the Past, Manage the
Present and Shape the Futurewww.senturus.com
888-601-6010
info@senturus.com
Copyright 2013 by Senturus, Inc. This entire presentation is
copyrighted and may not be reused or distributed without the written consent of
Senturus, Inc.

Microsoft SSAS: Should I Use Tabular or Multidimensional?

More Related Content

Viewers also liked (13)

Similar to Microsoft SSAS: Should I Use Tabular or Multidimensional? (20)

More from Senturus (20)

Recently uploaded (20)

Microsoft SSAS: Should I Use Tabular or Multidimensional?