Introduction to

Business Intelligence
Prem Shanker
Sr. Software Engineer Credit Suisse

Goals
Learn about the concept of Data Warehousing and what BIDS offer. Learn about how to design and implement a Data Warehouse Dimensional database. Learn about what is a cube. Learn about the SQL Server Analysis Services Architecture Learn what is new in Analysis Services 2008 Learn about what is a MDX Language.

What BIDS can do?

Source Systems/OLTP

Cubes

SQL Server Data Warehouse

Analysis Services

Clients

Query Tools Reporting Analysis Create OLAP Cubes

4

Design the Data Warehouse

Populate Data Warehouse

Query Data

Data Warehouse
Table and Cube Star Schema and Snowflake Schema Fact Table and Dimension Table

Table vs Cube
A simplified example:
A typical relational table
Data are organized by rows

Make it into a cube

Data are organized by intersections

Sales table
Product Donut Donut Milk Milk Region East West East West Sales $ 1 2

Region dim
East West Total

Donut Milk

1 3
4

2 4
6

3 7
10

3 Total 4

Product dim

The basic ingredients to make a cube

Two kinds of table in a data warehouse DB 1. fact table 2. dimension tables. Question: 1. Which one is a fact table and which one is a dimension table?

Star Schema
A Star Schema contains a fact table and one or more dimension tables. 1. A Fact Table: The central fact table store the numeric fact (measures) such as Sales dollars, Costs, Unit Sales etc. 2. Dimension Tables: They surround the central fact table, and they store descriptive information about the measures The shape looks like a Star

Star schema

Snowflake Schema

Review: Data Warehouse Schemas

The Data Warehouse is either a Star Schema or a Snowflake Schema:
Fact tables that contain foreign keys and numeric measures Dimension table contains the data describes the measures.

The schema is ready for Analysis Services to build a cube.

Client Server Architecture

Excel OLEDB TCP ADOMD Client Apps

Analysis Server
IIS

XMLA AMO

MOSS
BIDS SSMS ADOMD .NET SSRS

HTTP

A Logical Cube - Example

Product
Donut
Sandwich
The Sales$ by Soda by West in Yr of 2001

Region
North South East West 1999 2000 2001 2002
Time

Milk Soda

Beer

Tools to connect to Cubes

SQL Server Management Studio (SSMS) Business Intelligence Development Studio (BIDS) Query Analyzer (SSMS) To write MDX Excel 2007 Uses MDX

Physical Cube- BIDS

Analysis Services Database Unified Dimensional Model Data Source connection Data Source View Dimensions Cube Creation Wizard

Analysis Services Database

An Analysis Services database is the top level container for other dependent objects: A database includes Data Source Data Source View Cube Dimension Security Role

Creating an Analysis Services Database

You can use one of the following to create a new empty database on an instance of SQL Server 2005 Analysis Services.
SQL Server Management Studio Business Intelligence Development Studio.

Unified Dimensional Modeling

Common Name: UDM New feature Since AS 2005 Combine all Relational Sources in one single environment. A single data model, called Unified Dimensional Model (UDM) over one or more physical data sources

Unified Dimensional Model - Concept

The user needs to understand the particulars of each technology (e.g. the dialect of SQL used) to generate reports. Within one single Analysis Services, you can have more than one data sources to pull the data from.

Data Source Connection

The data sources of your AS database is your Data Warehouse databases (SQL). It defines the connection string and authentication information for a database on an OLE DB data provider. You can use the Data Source Wizard to specify one or more data sources (SQLDB) for Analysis Services databases.

The Functions of the Data Sources

Integrate your Analysis Services databases with the data warehouses They are used for the following: Processing the Cubes and dimensions Data Retrieval if ROLAP or HOLAP is used as the storage. Write Back

Different Storage types of Cube

Data Sources connection to SQL Server

For SQL Server, you can pick from the following providers: OLE DB provider for SQL Server SQL Native Client .NET Provider/SqlClient Data Provider
(Avoid using .NET data sources OLEDB is faster for processing in practices)

Data Source Views

New feature Since AS 2005 A single unified view of the metadata (UDM) from specified tables and views that the data source defines in the project. It hides the physical implementation of the underlying data sources from the reporting users. Basic Data Layout for Cubes Define Data Relationships Can Leverage Multiple Data Sources The key to effective cube design Named Query As Objects Not only Tables or Views

Demo

Dimension
All dimensions are based on tables or views in a data source view. All dimensions are shared since AS 2005 The structure of a dimension is largely driven by the structure of the underlying dimension table or tables. The simplest structure is called a star schema, which is where each dimension is based on a single dimension table that is directly linked to the fact table by a primary key - foreign key relationship.

Dimension Consists of
A dimension consists of:

Attributes that describe the entity User-Defined Hierarchies that organize dimension members in meaningful ways
such as
Store Name Store City Store State Store Country

Attributes
New feature since AS 2005 Containers of dimension members Typically have one-many relationships between attributes in the same dimension: City State, State Country, etc. All attributes implicitly related to the key

User Defined Hierarchies

User Defined Hierarchies are created from Attributes Tree-like structure City State Country All Provide navigation paths in a cube

Typical Example Calendar Hierarchy

The Year, Quarter, and Month attributes are used to construct a hierarchy, named Calendar, in the time dim. The relationship between the levels and members of the Calendar dimension (a regular dimension) is shown in the following diagram.

Measure Group
In a cube, a measure is the set of values, usually numeric, that are based on a column in the fact table in the cube. A measure group contains one or more or all the measures from a single fact table. It cant contain measures from different fact table.

Measure Group Advantages

Measure groups provide the following advantages:
They can be partitioned and processed separately They allows to include measures from diff fact tables. They are grouped by granularity: Same measure group same granularity. Security can be applied to specific measure groups

Cube
A cube is defined by its measures and dimensions.

Inside a Cube
Measures and Measure Groups Dimensions Relationships Calculations Actions Partitions Perspectives

Demo

Dimension Design
Different Dimension Relationships Regular Dimension Relationship Reference Dimension Relationship Fact Dimension Relationship Role Playing Dimension Parent-Child Hierarchy

Regular Dimension Relationships

A traditional star schema design The Primary Key in the dimension table joins directly to Foreign Key in the fact table.

Reference Dimension Relationships

Snowflake schema A Reference dimension using columns from multiple tables, or the dimension table links a dimension that is directly linked to the fact table.

Role Playing Dimension

It is used in a cube more than one time, each time for a different purpose. Each role-playing dimension is joined to a fact table on a different foreign key.
Example, you might add a Time dimension to a cube three times to track the times that

products are ordered, products are shipped, Orders are due..

Parent-Child Hierarchy
A parent-child hierarchy is a hierarchy in a standard dimension that contains a parent attribute. A parent attribute describes a self-join, within the same dimension table. Example: Employee Hierarchy An employee is an employee who reports to his/her manager. His manager is an employee as well Employee Key self joins to ParentEmployeeKey

Slowly Changing Dimension

Some attribute values may change over time. Two basic techniques: Type 1 change Type 2 change

Slowly Changing Dimension Type 1

A Type 1 change, is to simply overwrite the old value with the new one.

Slowly Changing Dimension Type 2

You create a new dimension row with the new value and a new surrogate key, and mark the old row or timestamp as no longer in effect The fact table will use the new surrogate key to link new fact measurements

Calculated Member
A Calculated Members is a member of a dimension or a measure group that is defined based on a MDX expression. The value for the member is calculated at runtime. The result values are not stored in the disk.

Calculated Member Properties

Named Set

A named set is a MDX expression that returns a set of dimension members. You can define named sets and save them as part of the cube definition. It allows you to reuse the same named set throughout the cube. Typical example: Create a list Top 10 customers based on Sales You can reuse same Top 10 customers in diff queries.

Best practices for Cube Design

Use integer or numeric for key columns. Avoid ROLAP storage mode, particular with custom rollup or unary operators. MOLAP is the fastest storage structure in SSAS. Use parent-child dimensions prudently, especially those containing custom rollup and unary operators. No aggregation support in PC dimension.

Best practices for Cube Design (Contd..)

Use role playing dimensions (e.g. OrderDate, BillDate, ShipDate) - avoids multiple physical copies. If the dimensions are base from the same physical table(s), use role playing dimensions.

What's New (Analysis Services - Multidimensional Database)

New Attribute Relationship designer. The dimension editor has a new Attribute Relationship designer that makes it easier to browse and modify attribute relationships. New AMO Warnings. These new warning messages alert users when they depart from design best practices or make logical errors in database design.

What's New (Analysis Services - Multidimensional Database)

Backup and Restore Improvements The backup and restore functionality in Analysis Services has a new storage structure and enhanced performance in all backup and restore scenarios. Improved Storage Structure The new storage structure provides a more robust repository for the archived database. By using the new storage structure, there is no practical limit to the size of the database file, nor is there a limit to the number of files that a database can have. Improved Performance The new backup and restore functionality achieves increased performance. Tests on different sized databases and with various numbers of files have shown significant performance improvements.

What's New (Analysis Services - Multidimensional Database)

Dynamic Management Views Monitoring Connections, Sessions, and Commands Discover_Connections, Discover_Sessions, and Discover_Commands. select * from $system.discover_connections

Fetching Data from Cube

What Is MDX Testing MDX with the Query Tool in SQL Server Management Studio The Basic Elements of an MDX Query

What Is MDX
An Extension of SQL Syntax That: Queries and manipulates multidimensional data in OLAP cubes Defines calculations based on information in the cube Defines and populates local cubes
Not a True Extension Syntax Deviates Significantly from SQL

Testing MDX with Management Studio

Background
Select on axis (x), on axis (y), on axis (z) From [cubeName]

Every cell has a name...

Components Clothing Bikes 1997 1998 Time 1999 2000

2001
Measures

Every cell has a name...

(Products.Bikes, Measures.Units, Time.[2000])
Components Clothing Bikes 1997 1998 Time 1999 2000 2001

Measures

Every cell has a name...

(Products.Bikes, Measures.Units, Time.[2000])

(Products.Bikes, Measures.Sales, Time.[1999])

Components Clothing Bikes 1997 1998 Time 1999 2000 2001 Measures

A Cell is referenced by all the dimensions

What if I only specify this? (Products.Bikes, Measures.Units)
Components Clothing Bikes 1997 1998 Time 1999 2000 2001 Measures

Default Member
What if I only specify this? (Products.Bikes, Measures.Units) If Times default member is [1997] Ans: (Products.Bikes, Measures.Units, Time.[1997])
Computer Monitor Printer 1997 1998 Time 1999

2000
2001 Measures

The Basic Elements of an MDX Query

Select {[Ship Date].[Calendar]} on columns, {[Product].[Product Categories]} on rows from [Adventure Works]

Using Braces { }
Braces Denote a Set Braces Can be Omitted when the Set is Unambiguous. In SSAS 2005 / 2008:
SELECT [Ship Date].[Calendar] ON COLUMNS, [Product].[Product Categories] ON ROWS
FROM [Adventure Works]

In AS 2000:
SELECT {[Ship Date].[Calendar]} ON COLUMNS, {[Product].[Product Categories]} ON ROWS FROM [Adventure Works]

Using Brackets [ ]
Brackets Enclose a String Value Necessary for: Field names with spaces: [New York], [Mary Lo] Numbers as field names: [2007], [2008]
Otherwise, the SSAS will treat them as numerous constants

Default Members
Every Dimension has a Default Member Usually the All member is the default member. Default Measures The measures dimension also has a default measure In our sample cube [Adventure Works], the default member for the cube is [Reseller Sales Amount]

Members
You want to query more than a single cell. Use Members function Members function returns the set of members in a dimension, level, or hierarchy.
select [Ship Date].[Calendar] on columns, [Product].[Product Categories].members on rows from [Adventure Works]

Test Yourself: Number 1

[Ship Date].[Calendar] also has a membership; that is, it is made up of more granular information. Modify the query to return the membership of the [Ship Date].[Calendar]dimension.
select [Ship Date].[Calendar] on columns, [Product].[Product Categories].members on rows from [Adventure Works]

Desired result:

Naming Additional Dimensions

Number AXIS(0) Name COLUMNS

AXIS(1)
AXIS(2) AXIS(3) AXIS(4)

ROWS
PAGES SECTIONS CHAPTERS

Retrieving Data from a Cube

select [Ship Date].[Calendar].[Calendar Year].[CY 2004] on axis(0), [Promotion].[Promotions].[reseller] on axis(1) from [Adventure Works] [Promotion].[Promotions]

No Discunt Reseller
2001 2002 2003 2004

[Ship Date].[Calendar]

Test Yourself: Number 2

Modify the query to return the sales of Bikes with No Discount
select [Ship Date].[Calendar].[Calendar Year].[CY 2004] on axis(0), [Promotion].[Promotions].[reseller] on axis(1) from [Adventure Works]

Expect Result

Fully Qualified Names

[CY 2001] below could be [Delivery Date].[Calendar].[CY 2001] or [Ship Date].[Calendar].[CY 2001]
select [CY 2001] on axis(0) from [Adventure Works]
[Product].[Product Categories].[bikes] is the same as [Product].[Product Categories].[All Products].[bikes]

Two Dimensions with Where Clause

select [Ship Date].[Calendar].[Calendar Year].members on axis(0), [Promotion].[Promotions].[reseller] on axis(1) from [Adventure Works] where [Product].[Product Categories].[bikes]
[Promotion].[Pr omotions] No Discount Reseller

2001

Components Clothing [Product]. Bikes [Product Categories] 2002 2003 2004

[Ship Date].[Calendar]

Demo
Lab MDX Query

Few Useful References

www.microsoft.com/sqlserver/2008/en/us/analysisservices.aspx All BI WebCasts http://www.microsoft.com/events/series/bi.aspx?tab=webc asts&id=all MDX References msdn.microsoft.com/en-us/library/ms145506.aspx

Thank You pundit.prem@gmail.com