You are on page 1of 28

On-Line Analytical Processing

OLAP - Definition
OLAP enables
analysts, managers, and decision makers to
gain insight into data through
fast, consistent, interactive access in
a variety of possible views of information transformed from
raw data to reflect real dimensionality of the enterprise as
understood by the user

OLAP - Definition

Summarised, specific
operational needs

Data Mart

Integrated, primitive,
historical data

Datawarehouse

OLAP level

Why OLAP
Multi-dimensional analysis
Fast access
Powerful Computations
Summaries and aggregations to support decision making
Moving from summaries to detailed data
Simple calculations: margin=sales-costs
Share calculations: percentages, ratios
Algebraic operations on key business performance indicators
Trend analysis using statistical methods

Why OLAP
Report writers
Point and click to issue SQL query
fixed format output reports
No multidimensionality, no navigation on aggregates etc.
Spreadsheets
can do what if analysis
with add-in tools: some aggregation and calculations
possible
can view rows, columns, and pages
some spreadsheets can cross tabulate, pivot rows/columns
Difficulties with dimensions>3, multidimensionality,

Why OLAP
SQL
Syntax hiding techniques enable NL or point and click
Cannot support complex thought process
a query may translate to many SQL queries
chain of queries: too much effort to write the query
Poor response: queries require full table scans,
multiple joins, aggregations, groupings, sorting

OLAP Operations

Drill Down/Roll-up
Drill down: An operation that allow information retrieval towards
the leaves of the dimension hierarchy
Roll-up: the inverse of drill down
Show the month-wise sales of products for the Northern region
Region Dimension
Radio
Dec
5k
Jan
3k
Feb
7k
Mar
10k

TV
7k
5k
4k
1k

CRO Multimeter
10k
2k
15k
4k
20k
6k
50k
8k

CD
1k
5k
7k
8k

Casette
3k
7k
10k
15k

Example: Roll-up
Roll-up: Show month-wise sales of product type for the Northern region
Region Dimension
Dec
Jan
Feb
Mar

Home
12k
8k
11k
11k

Industrial
12k
19k
26k
58k

Educational
4k
12k
17k
23k

Example: Roll-up
Roll-up: Show season-wise sales of product type for the Northern region
Region Dimension
Home
Winter
Vasant

20k
22k

Industrial
31k
84k

Educational
16k
40k

The drill down can be done by performing the reverse sequence of


operations

OLAP Operations

Drill-across: to another OLAP summarization using a


different set of hierarchies of other dimensions.
Drill-through to the lower levels of granularity, as stored in
the source data warehouse repository.

Slice & Dice/Rotation


To view different versions of slices of cubes

To view the data from many different angles


in order to
understand the facts better
arrive at meaningful conclusions
by
rotating the cube and inquiring along its different dimensions

Slice & Dice/Rotation


product

region

Product along Z-axis, becomes columns


Month along X-axis: becomes rows
Region along Y axis: is static

month
Show the month-wise sales of products for the Northern region
Region = Northern Region
Radio TV
Dec
5k
7k
Jan
3k
5k
Feb
7k
4k
Mar
10k
1k

CRO Multimeter
10k
2k
15k
4k
20k
6k
50k
8k

CD
1k
5k
7k
8k

Casette
3k
7k
10k
15k

Slice & Dice/Rotation


Turn the cube round
Region along Z-axis, becomes columns
Month along X-axis: becomes rows
Product along Y axis: is static
Show the month-wise sales of product = Radio for all regions
Product= Radio
NR
Dec
5k
Jan
3k
Feb
7k
Mar
10k

ER
7k
4k
4k
1k

SR
8k
12k
20k
50k

CR
3k
4k
6k
8k

WR
2k
5k
7k
8k

Hypercubes
Since the representation is difficult, use alternative techniques
Shop
Anika
Kanika
Latika

Month
Dec
Jan
Feb
Mar

Product
Radio
TV
CRO
Multimeter
CD
Casette

Region
NR
ER
CR
SR
WR

How is a result displayed? BY adding additional information

Hypercubes
Show the month-wise sales of products for each shop for the Northern
region
Region = Northern Region
Radio: Anika Radio: Kanika
Dec
1k
2k
Jan
1k
1k
Feb
2k
3k
Mar
3k
3k
Repeat for each product: TV, CRO etc.

Radio: Latika
2k
1k
2k
4k

Hypercubes
Economic
Promotion Shop
Status
Month
LIG
Anika
Type
Dec
Jan
MIG
Media Kanika
Feb
Mar
HIG
Latika
Cost

Product
Radio
TV
CRO
Multimeter
CD
Casette

Region
NR
ER
CR
SR
WR

Hypercubes
Show the month-wise sales of products for each shop for each
economic status for the Northern region
Region = Northern Region
Radio: Anika Radio: Kanika
Dec
1k
2k
Dec
1k
2k
Dec
1k
2k
Jan
1k
1k
Jan
1k
1k
Jan
1k
1k

Radio: Latika
2k
2k
2k
1k
1k
1k

Data Warehouse: What Is It?


An architecture, an environment for corporate decision making

OLAP

Detailed

Aggregate

Current

Small size of data

large size of data

Unpredictable access
Drill down
Drill through
Drill across
Slice and dice

Limited access
Queries
Reports
Browsing
Statistics

Long time horizon data Short time horizon data

OLAP Models
ROLAP

MOLAP
Desktop
Request

Desktop

Proprietary Language

MDDB

OLAP
server

Create/Store cubes
Data Warehouse

Database
server

Request Create cubes


dynamically
OLAP
services
Complex SQL
Data Warehouse

Database
server

Relational Vs. MD Model


Relational

Multidimensional DBMS

Handles large amounts of data

Handles pre-defined set of data

Not Good for

Not Good for

slics and dice/drill down, drill


across

drill through, querying large data

Good for

Good for

dynamic joins, general purpose


update processing, drill through

time-series data, top ten/bottom


ten, slicing/dicing, analysing
aggregate data

Non-optimal performance

High refresh time costs


Inefficient SQL generation

MOLAP VS. ROLAP


Multidimensional OLAP
direct representation of the
data cube
difficulty handling sparsity
efficiently
proprietary solutions -OLAP
Councils API for OLAP
thin client, analytical
processing done on server

Relational OLAP
multidimensional view
built on a Relational DBMS
handles sparsity automatically
depends on efficient
processing of STAR joins and
indexes
analytical processing done on
the client (or middle server)

ROLAP: An analytical processing technique in which data cubes are


created on the fly by the relational database engine

Query perfromance

MOLAP VS. ROLAP

MOLAP

ROLAP

Relational data in DW
Summary data in proprietary MDDBs
Data access from different levels
Pre-calculated cubes
Large library of functions for
complex calculations

Complexity of Analysis
All data as relations
All access from DW
Complex SQL to fetch data
Engine to create data cubes dynamically

You might also like