DATA WAREHOUSING

AND
DATA MINING
 A producer wants to know….
Which
Whichare
areour
our
lowest/highest
lowest/highestmargin
margin
customers
customers??
Who
Whoare
aremymycustomers
customers
What and
andwhat
whatproducts
Whatisisthe
themost
most products
effective are
arethey
theybuying?
effectivedistribution
distribution buying?
channel?
channel?

What
Whatproduct
productprom- Which
prom- Whichcustomers
customers
-otions
-otionshave
havethe
thebiggest are
biggest aremost
mostlikely
likelyto
togo
go
impact
impactononrevenue? to
revenue? tothe
thecompetition
competition??
What
Whatimpact
impactwill
will
new
newproducts/services
products/services
have
haveon
onrevenue
revenue
and
andmargins?
margins? 2
Data, Data everywhere
yet ...  I can’t find the data I need
 data is scattered over the network
 many versions, subtle differences

 I can’t get the data I need

 need an expert to get the data

 I can’t understand the data I

found
 available data poorly documented

 I can’t use the data I found

 results are unexpected
 data needs to be transformed
3
from one form to other
What is a Data Warehouse?

A single, complete and

consistent store of data
obtained from a variety
of different sources
made available to end
users in a what they
can understand and use
in a business context.

[Barry Devlin]
4
What are the users saying...
 Data should be integrated
across the enterprise
 Summary data has a real
value to the organization
 Historical data holds the
key to understanding data
over time
 What-if capabilities are
required

5
What is Data Warehousing?

A process of
Information transforming data into
information and making
it available to users in a
timely enough manner
to make a difference

[Forrester Research, April

1996]
Data
6
Evolution

 60’s: Batch reports

 hard to find and analyze information
 inflexible and expensive, reprogram every new request
 70’s: Terminal-based DSS and EIS (executive
information systems)
 still inflexible, not integrated with desktop tools
 80’s: Desktop data access and analysis tools
 query tools, spreadsheets, GUIs
 easier to use, but only access operational databases
 90’s: Data warehousing with integrated OLAP
engines and tools

7
 Warehouses are Very Large
Databases

35%

30%

25%
Respondents

20%

15%

10%
Initial
5% Projected 2Q96

Source: META Group, Inc.

0%
5GB 10-19GB 50-99GB 250-499GB
5-9GB 20-49GB 100-249GB 500GB-1TB
8
Very Large Data Bases
 Terabytes -- 10^12 bytes: Walmart -- 24 Terabytes

 Petabytes -- 10^15 bytes: Geographic Information

Systems
 Exabytes -- 10^18 bytes: National Medical
Records
 Zettabytes -- 10^21 bytes:
Weather images
 Zottabytes -- 10^24 bytes:
Intelligence Agency
Videos
9
Data Warehousing --
It is a process
 Technique for assembling and
managing data from various
sources for the purpose of
answering business questions.
Thus making decisions that
were not previous possible
 A decision support database
maintained separately from
the organization’s operational
database
10
Data Warehouse

 A data warehouse is a
 subject-oriented
 integrated
 time-varying
 non-volatile

collection of data that is used primarily in

organizational decision making.
-- Bill Inmon, Building the Data Warehouse
1996
11
Explorers, Farmers and Tourists

Tourists: Browse information

harvested by farmers

Farmers: Harvest information

from known access paths

Explorers: Seek out the unknown and

previously unsuspected rewards hiding
in the detailed data

12
 Data Warehouse Architecture
Relational
Databases
Optimized Loader
Extraction
ERP
Systems Cleansing

Data Warehouse
Engine Analyze
Purchased Query
Data

Legacy
Data Metadata Repository
13
Data Warehouse for Decision
Support & OLAP
 Putting Information technology to help the
knowledge worker make faster and better
decisions
 Which of my customers are most likely to go to
the competition?
 What product promotions have the biggest
impact on revenue?
 How did the share price of software
companies correlate with profits over last 10
years?
14
Decision Support

 Used to manage and control business

 Data is historical or point-in-time
 Optimized for inquiry rather than update
 Use of the system is loosely defined and
can be ad-hoc
 Used by managers and end-users to
understand the business and make
judgements
15
Data Mining works with
Warehouse Data

 Data Warehousing provides

the Enterprise with a
memory

 Data Mining provides

the Enterprise with
intelligence
16
We want to know ...
 Given a database of 100,000 names, which persons are the
least likely to default on their credit cards?
 Which types of transactions are likely to be fraudulent
given the demographics and transactional history of a
particular customer?
 If I raise the price of my product by Rs. 2, what is the effect
on my ROI?
 If I offer only 2,500 airline miles as an incentive to purchase
rather than 5,000, how many lost responses will result?
 If I emphasize ease-of-use of the product as opposed to its
technical capabilities, what will be the net effect on my
revenues?
 Which of my customers are likely to be the most loyal?

Data Mining helps extract such information

17
Application Areas

Industry Application
Finance Credit Card Analysis
Insurance Claims, Fraud Analysis
Telecommunication Call record analysis
Transport Logistics management
Consumer goods promotion analysis
Data Service providersValue added data
Utilities Power usage analysis

18
Data Mining in Use
 The US Government uses Data Mining to track
fraud
 A Supermarket becomes an information
broker
 Basketball teams use it to track game
strategy
 Cross Selling
 Warranty Claims Routing
 Holding on to Good Customers
 Weeding out Bad Customers
19
What makes data mining
possible?
Advances in the following areas are
making data mining deployable:
 data warehousing
 better and more data (i.e., operational,
behavioral, and demographic)
 the emergence of easily deployed data
mining tools and
 the advent of new data mining
techniques.
• -- Gartner Group
20
 Why Separate Data Warehouse?
 Performance
 Op dbs designed & tuned for known txs & workloads.
 Complex OLAP queries would degrade perf. for op txs.
 Special data organization, access & implementation
methods needed for multidimensional views & queries.

 Function
 Missing data: Decision support requires historical data, which
op dbs do not typically maintain.
 Data consolidation: Decision support requires consolidation
(aggregation, summarization) of data from many
heterogeneous sources: op dbs, external sources.
 Data quality: Different sources typically use inconsistent data
representations, codes, and formats which have to be 21
reconciled.
What are Operational Systems?
 They are OLTP systems
 Run mission critical
applications
 Need to work with
stringent performance
requirements for
routine tasks
 Used to run a
business!

22
RDBMS used for OLTP

Database Systems have been used

traditionally for OLTP
 clerical data processing tasks
 detailed, up to date data
 structured repetitive tasks
 read/update a few records
 isolation, recovery and integrity are
critical

23
Operational Systems
 Run the business in real time
 Based on up-to-the-second data
 Optimized to handle large
numbers of simple read/write
transactions
 Optimized for fast response to
predefined transactions
 Used by people who deal with
customers, products -- clerks,
salespeople etc.
 They are increasingly used by
customers

24
Examples of Operational Data
Data IndustryUsage Technology Volumes
Customer All Track Legacy application, flat
Small-medium
File Customer files, main frames
Details
Account Finance Control Legacy applications, Large
Balance account hierarchical databases,
activities mainframe
Point-of- Retail Generate ERP, Client/Server, Very Large
Sale data bills, manage
relational databases
stock
Call Telecomm- Billing Legacy application, Very Large
Record unications hierarchical database,
mainframe
ProductionManufact- Control ERP, Medium
Record uring Production relational databases,
AS/400
25
So, what’s different?
 Application-Orientation vs.
Subject-Orientation

Application-Orientation Subject-Orientation

Operation Data
al Warehouse
Database
Credit
Loans Customer
Card
Vendor
Trust Product

Savings Activity
27
OLTP vs. Data Warehouse

 OLTP systems are tuned for known

transactions and workloads while workload
is not known a priori in a data warehouse
 Special data organization, access methods
and implementation methods are needed
to support data warehouse queries
(typically multidimensional queries)
 e.g., average amount spent on phone calls
between 9AM-5PM in Pune during the month of
December

28
OLTP vs Data Warehouse

 OLTP  Warehouse (DSS)

 Application  Subject Oriented
Oriented  Used to analyze
 Used to run business
business  Summarized and
 Detailed data refined
 Current up to date  Snapshot data
 Isolated Data  Integrated Data
 Repetitive access  Ad-hoc access
 Clerical User  Knowledge User
(Manager)
29
OLTP vs Data Warehouse

 OLTP  Data Warehouse

 Performance Sensitive  Performance relaxed
 Few Records accessed  Large volumes accessed
at a time (tens) at a time(millions)
 Mostly Read (Batch
 Read/Update Access Update)
 Redundancy present
 No data redundancy  Database Size 100
 Database Size 100MB GB - few terabytes
-100 GB

30
OLTP vs Data Warehouse

 OLTP  Data Warehouse

 Transaction  Query throughput is
throughput is the the performance
performance metric metric
 Thousands of users  Hundreds of users
 Managed in entirety  Managed by
subsets

31
To summarize ...
 OLTP Systems are
used to “run” a
business

 The Data Warehouse

helps to “optimize”
the business

32
Why Now?
Data is being produced
ERP provides clean data
The computing power is available
The computing power is affordable
The competitive pressures are strong
Commercial products are available

33
Myths surrounding OLAP Servers
and Data Marts
 Data marts and OLAP servers are departmental
solutions supporting a handful of users
 Million dollar massively parallel hardware is needed
to deliver fast time for complex queries
 OLAP servers require massive and unwieldy indices
 Complex OLAP queries clog the network with data
 Data warehouses must be at least 100 GB to be
effective
– Source -- Arbor Software Home Page

34
Wal*Mart Case Study
Founded by Sam Walton
One the largest Super Market Chains
in the US

Wal*Mart: 2000+ Retail Stores

SAM's Clubs 100+Wholesalers Stores
 This case study is from Felipe Carino’s (NCR Teradata)
presentation made at Stanford Database Seminar

35
Old Retail Paradigm
 Wal*Mart  Suppliers
 Inventory Management  Accept Orders
 Merchandise Accounts  Promote Products
Payable  Provide special
 Purchasing Incentives
 Supplier Promotions:  Monitor and Track
National, Region, Store The Incentives
Level  Bill and Collect
Receivables
 Estimate Retailer
Demands
36
New (Just-In-Time) Retail
Paradigm
 No more deals
 Shelf-Pass Through (POS Application)
 One Unit Price
 Suppliers paid once a week on ACTUAL items sold
 Wal*Mart Manager
 Daily Inventory Restock
 Suppliers (sometimes SameDay) ship to Wal*Mart
 Warehouse-Pass Through
 Stock some Large Items
 Delivery may come from supplier
 Distribution Center
 Supplier’s merchandise unloaded directly onto Wal*Mart
Trucks

37
Wal*Mart System

 NCR 5100M 96 24 TB Raw Disk; 700 -

Nodes; 1000 Pentium CPUs
 Number of Rows: > 5 Billions
 Historical Data: 65 weeks (5 Quarters)
 New Daily Volume: Current Apps: 75 Million
New Apps: 100 Million +
 Number of Users: Thousands
 Number of Queries: 60,000 per week

38
 Course Overview

 0. Introduction
 I. Data Warehousing
 II. Decision Support
and OLAP
 III. Data Mining
 IV. Looking Ahead

 Demos and Labs

39
I. Data Warehouses:
Architecture, Design &
Construction
 DW Architecture
 Loading, refreshing
 Structuring/Modeling
 DWs and Data Marts
 Query Processing

 demos, labs

40
 Data Warehouse Architecture
Relational
Databases
Optimized Loader

ERP Extraction
Systems Cleansing

Data Warehouse
Engine Analyze
Purchased Query
Data

Legacy
Data Metadata Repository
41
Components of the Warehouse
Data Extraction and Loading
The Warehouse
Analyze and Query -- OLAP Tools
Metadata

Data Mining tools

42
Loading the Warehouse

Cleaning the data

before it is loaded
Source Data

Operational/ Sequential
Source Data Legacy Relational External

Typically host based, legacy

applications
 Customized applications, COBOL, 3GL, 4GL
Point of Contact Devices
 POS, ATM, Call switches
External Sources
 Nielsen’s, Acxiom, CMIE, Vendors, Partners

44
Data Quality - The Reality

Tempting to think creating a data

warehouse is simply extracting
operational data and entering into a
data warehouse

Nothing could be farther from the truth

Warehouse data comes from disparate
questionable sources

45
Data Quality - The Reality

 Legacy systems no longer documented

 Outside sources with questionable quality
procedures
 Production systems with no built in integrity
checks and no integration
 Operational systems are usually designed to
solve a specific business problem and are rarely
developed to a a corporate plan
 “And get it done quickly, we do not have time to worry
about corporate standards...”

46
 Data Integration Across Sources

Savings Loans Trust Credit card

Same data Different data Data found here Different keys

different name Same name nowhere else same data

47
 Data Transformation Example

Data Warehouse
encoding

appl A - m,f
appl B - 1,0
appl C - x,y
appl D - male, female

appl A - pipeline - cm
unit

appl B - pipeline - in
appl C - pipeline - feet
appl D - pipeline - yds

appl A - balance
field

appl B - bal
appl C - currbal
appl D - balcurr
48
Data Integrity Problems

 Same person, different spellings

 Agarwal, Agrawal, Aggarwal etc...
 Multiple ways to denote company name
 Persistent Systems, PSPL, Persistent Pvt. LTD.
 Use of different names
 mumbai, bombay
 Different account numbers generated by different
applications for the same customer
 Required fields left blank
 Invalid product codes collected at point of sale
 manual entry leads to mistakes
 “in case of a problem use 9999999”

49
Data Transformation Terms

Extracting Enrichment
Conditioning Scoring
Scrubbing Loading
Merging Validating
Householding Delta Updating

50
Data Transformation Terms

 Extracting
 Capture of data from operational source in
“as is” status
 Sources for data generally in legacy
mainframes in VSAM, IMS, IDMS, DB2; more
data today in relational databases on Unix
 Conditioning
 The conversion of data types from the source
to the target data store (warehouse) -- always
a relational database
51
Data Transformation Terms

Householding
 Identifying all members of a household
(living at the same address)
 Ensures only one mail is sent to a
household
 Can result in substantial savings: 1 lakh
catalogues at Rs. 50 each costs Rs. 50
lakhs. A 2% savings would save Rs. 1
lakh.

52
Data Transformation Terms

Enrichment
 Bring data from external sources to
augment/enrich operational data. Data
sources include Dunn and Bradstreet, A.
C. Nielsen, CMIE, IMRA etc...
Scoring
 computation of a probability of an
event. e.g..., chance that a customer will
defect to AT&T from MCI, chance that a
customer is likely to buy a new product

53
Loads

After extracting, scrubbing, cleaning,

validating etc. need to load the data
into the warehouse
Issues
 huge volumes of data to be loaded
 small time window available when warehouse can be
taken off line (usually nights)
 when to build index and summary tables
 allow system administrators to monitor, cancel, resume,
change load rates
 Recover gracefully -- restart after failure from where
you were and without loss of data integrity

54
Load Techniques

Use SQL to append or insert new

data
 record at a time interface
 will lead to random disk I/O’s
Use batch load utility

55
Load Taxonomy

Incremental versus Full loads

Online versus Offline loads

56
Refresh

Propagate updates on source data to

the warehouse
Issues:
 when to refresh
 how to refresh -- refresh techniques

57
When to Refresh?
 periodically (e.g., every night, every week)
or after significant events
 on every update: not warranted unless
warehouse data require current data (up
to the minute stock quotes)
 refresh policy set by administrator based
on user needs and traffic
 possibly different policies for different
sources
58
Refresh Techniques

Full Extract from base tables

 read entire source table: too expensive
 maybe the only choice for legacy
systems

59
How To Detect Changes

Create a snapshot log table to record

ids of updated rows of source data
and timestamp
Detect changes by:
 Defining after row triggers to update
snapshot log when source table
changes
 Using regular transaction log to detect
changes to source data
60
Data Extraction and Cleansing
Extract data from existing
operational and legacy data
Issues:
 Sources of data for the warehouse
 Data quality at the sources
 Merging different data sources
 Data Transformation
 How to propagate updates (on the sources) to
the warehouse
 Terabytes of data to be loaded

61
Scrubbing Data
 Sophisticated
transformation tools.
 Used for cleaning the
quality of data
 Clean data is vital for the
success of the warehouse
 Example
 Seshadri, Sheshadri,
Sesadri, Seshadri S.,
Srinivasan Seshadri, etc.
are the same person

62
Scrubbing Tools
Apertus -- Enterprise/Integrator
Vality -- IPE
Postal Soft

63
Structuring/Modeling Issues
Data -- Heart of the Data
Warehouse
Heart of the data warehouse is the
data itself!
Single version of the truth
Corporate memory
Data is organized in a way that
represents business -- subject
orientation

65
Data Warehouse Structure

Subject Orientation -- customer,

product, policy, account etc... A
subject may be implemented as a
set of related tables. E.g.,
customer may be five tables

66
 Data Warehouse Structure
 base customer (1985-87)
 custid, from date, to date, name, phone, dob

Time is  base customer (1988-90)

part of  custid, from date, to date, name, credit rating,
key of employer
each table

customer activity (1986-89) -- monthly
summary
 customer activity detail (1987-89)
 custid, activity date, amount, clerk id, order no
 customer activity detail (1990-91)
 custid, activity date, amount, line item no, order no

67
Data Granularity in Warehouse

Summarized data stored

 reduce storage costs
 reduce cpu usage
 increases performance since smaller
number of records to be processed
 design around traditional high level
reporting needs
 tradeoff with volume of data to be
stored and detailed usage of data
68
Granularity in Warehouse

Can not answer some questions with

summarized data
 Did Anand call Seshadri last month? Not
possible to answer if total duration of
calls by Anand over a month is only
maintained and individual call details
are not.
Detailed data too voluminous

69
Granularity in Warehouse
Tradeoff is to have dual level of
granularity
 Store summary data on disks
95% of DSS processing done against this
data
 Store detail on tapes
5% of DSS processing against this data

70
Vertical Partitioning

Acct. Interest
Name BalanceDate Opened Address
No Rate

Frequently
accessed Rarely
accessed
Acct. Acct. Interest
Balance Name Date Opened Address
No No Rate

Smaller table
and so less
I/O
71
Derived Data
Introduction of derived (calculated
data) may often help
Have seen this in the context of dual
levels of granularity
Can keep auxiliary views and
indexes to speed up query
processing

72
Schema Design

Database organization
 must look like business
 must be recognizable by business user
 approachable by business user
 Must be simple
Schema Types
 Star Schema
 Fact Constellation Schema
 Snowflake schema

73
Dimension Tables

Dimension tables
 Define business in terms already familiar
to users
 Wide rows with lots of descriptive text
 Small tables (about a million rows)
 Joined to fact table by a foreign key
 heavily indexed
 typical dimensions
time periods, geographic region (markets,
cities), products, customers, salesperson, etc.

74
Fact Table

Central table
 mostly raw numeric items
 narrow rows, a few columns at most
 large number of rows (millions to a
billion)
 Access via dimensions

75
Star Schema

 A single fact table and for each

dimension one dimension table
 Does not capture hierarchies directly
T p
date, custno, prodno, cityname, ...
i r
m o
e f d
a
c c c
u t i
s t
t y 76
Snowflake schema

 Represent dimensional hierarchy

directly by normalizing tables.
 Easy to maintain and saves storage
T p
date, custno, prodno, cityname, ...
i r
m o
e f d
a
c c c r
u t i e
s g
t i
t y o
77
n
Fact Constellation

Fact Constellation
 Multiple fact tables that share many
dimension tables
 Booking and Checkout may share many
dimension tables in the hotel industry
Promotion
Hotels
Booking
Checkout
Travel Agents Room Type
Customer 78
De-normalization
Normalization in a data warehouse
may lead to lots of small tables
Can lead to excessive I/O’s since
many tables have to be accessed
De-normalization is the answer
especially since updates are rare

79
Creating Arrays

 Many times each occurrence of a sequence of

data is in a different physical location
 Beneficial to collect all occurrences together
and store as an array in a single row
 Makes sense only if there are a stable number
of occurrences which are accessed together
 In a data warehouse, such situations arise
naturally due to time based orientation
 can create an array by month

80
Selective Redundancy
Description of an item can be stored
redundantly with order table -- most
often item description is also
accessed with order table
Updates have to be careful

81
Partitioning
 Breaking data into several
physical units that can be
handled separately
 Not a question of whether
to do it in data
warehouses but how to do
it
 Granularity and
partitioning are key to
effective implementation
of a warehouse
82
Why Partition?
Flexibility in managing data
Smaller physical units allow
 easy restructuring
 free indexing
 sequential scans if needed
 easy reorganization
 easy recovery
 easy monitoring

83
Criterion for Partitioning
Typically partitioned by
 date
 line of business
 geography
 organizational unit
 any combination of above

84
Where to Partition?
Application level or DBMS level
Makes sense to partition at
application level
 Allows different definition for each year
Important since warehouse spans many
years and as business evolves definition
changes
 Allows data to be moved between
processing complexes easily

85
Data Warehouse vs. Data Marts

What comes first

 From the Data Warehouse to
Data Marts

Information

Individually Less
Structured

Departmentally History
Structured Normalized
Detailed

Organizationally More
Structured Data Warehouse

Data
87
Data Warehouse and Data Marts
OLAP
Data Mart
Lightly summarized
Departmentally structured

Organizationally structured
Atomic
Detailed Data Warehouse Data

88
Characteristics of the
Departmental Data Mart
 OLAP
 Small
 Flexible
 Customized by
Department
 Source is
departmentally
structured data
warehouse

89
Techniques for Creating
Departmental Data Mart

OLAP
Sales Finance Mktg.
Subset
Summarized
Superset
Indexed
Arrayed

90
Data Mart Centric

Data Sources

Data Marts

Data Warehouse

91
Problems with Data Mart Centric
Solution

If you end up creating multiple

warehouses, integrating them is a
problem
92
 True Warehouse

Data Sources

Data Warehouse

Data Marts

93
Query Processing

 Indexing

 Pre computed
views/aggregates
 SQL extensions
94
Indexing Techniques
Exploiting indexes to reduce scanning
of data is of crucial importance
Bitmap Indexes
Join Indexes
Other Issues
 Text indexing
 Parallelizing and sequencing of index
builds and incremental updates

95
Indexing Techniques

Bitmap index:
 A collection of bitmaps -- one for each
distinct value of the column
 Each bitmap has N bits where N is the
number of rows in the table
 A bit corresponding to a value v for a
row r is set if and only if r has the value
for the indexed attribute

96
BitMap Indexes
 An alternative representation of RID-list
 Specially advantageous for low-cardinality
domains
 Represent each row of a table by a bit and
the table as a bit vector
 There is a distinct bit vector Bv for each
value v for the domain
 Example: the attribute sex has values M
and F. A table of 100 million people needs
2 lists of 100 million bits
97
Bitmap Index

M Y 0 1 0

F Y 1 1 1

F N 1 0 0

M N 0 0 0

F Y 1 1 1

F N 1 0 0

Customer Query : select * from customer where

gender = ‘F’ and vote = ‘Y’ 98
Bit Map Index
Base Region
Table Index Rating Index
Cust Region Rating Row ID N S E W Row ID H M L
C1 N H 1 1 0 0 0 1 1 0 0
C2 S M 2 0 1 0 0 2 0 1 0
C3 W L 3 0 0 0 1 3 0 0 0
C4 W H 4 0 0 0 1 4 0 0 0
C5 S L 5 0 1 0 0 5 0 1 0
C6 W L 6 0 0 0 1 6 0 0 0
C7 N H 7 1 0 0 0 7 1 0 0
Region = Rating =
Customers whereW And M

99
 BitMap Indexes
 Comparison, join and aggregation operations
are reduced to bit arithmetic with dramatic
improvement in processing time
 Significant reduction in space and I/O (30:1)
 Adapted for higher cardinality domains as well.
 Compression (e.g., run-length encoding)
exploited
 Products that support bitmaps: Model 204,
TargetIndex (Redbrick), IQ (Sybase), Oracle 7.3

100
Join Indexes

 Pre-computed joins
 A join index between a fact table and a
dimension table correlates a dimension
tuple with the fact tuples that have the
same value on the common dimensional
attribute
 e.g., a join index on city dimension of calls
fact table
 correlates for each city the calls (in the calls
table) from that city

101
Join Indexes
Join indexes can also span multiple
dimension tables
 e.g., a join index on city and time
dimension of calls fact table

102
Star Join Processing

Use join indexes to join dimension and fact table

Calls
C+T

Time C+T+L

Loca-
tion C+T+L
Plan +P
103
Optimized Star Join Processing

Time Apply Selections

Loca- Calls
tion
Virtual Cross Product
Plan of T, L and P

104
Bitmapped Join Processing

Bitmaps
1
Time Calls
0
1

Loca- 0

tion Calls 0
1 AND

Plan Calls
1
1
0

105
Intelligent Scan

Piggyback multiple scans of a

relation (Redbrick)
 piggybacking also done if second scan
starts a little while after the first scan

106
Parallel Query Processing

Three forms of parallelism

 Independent
 Pipelined
 Partitioned and “partition and replicate”
Deterrents to parallelism
 startup
 communication

107
Parallel Query Processing

 Partitioned Data
 Parallel scans
 Yields I/O parallelism
 Parallel algorithms for relational operators
 Joins, Aggregates, Sort
 Parallel Utilities
 Load, Archive, Update, Parse, Checkpoint,
Recovery
 Parallel Query Optimization

108
Pre-computed Aggregates

Keep aggregated data for

efficiency (pre-computed queries)
Questions
 Which aggregates to compute?
 How to update aggregates?
 How to use pre-computed aggregates
in queries?

109
Pre-computed Aggregates
Aggregated table can be maintained
by the
 warehouse server
 middle tier
 client applications
Pre-computed aggregates -- special
case of materialized views -- same
questions and issues remain

110
SQL Extensions

Extended family of aggregate

functions
 rank (top 10 customers)
 percentile (top 30% of customers)
 median, mode
 Object Relational Systems allow
addition of new aggregate functions

111
SQL Extensions
Reporting features
 running total, cumulative totals
Cube operator
 group by on all subsets of a set of
attributes (month,city)
 redundant scan and sorting of data can
be avoided

112
Red Brick has Extended set of
Aggregates
 Select month, dollars, cume(dollars) as
run_dollars, weight, cume(weight) as
run_weights
from sales, market, product, period t
where year = 1993
and product like ‘Columbian%’
and city like ‘San Fr%’
order by t.perkey

113
RISQL (Red Brick Systems)
Extensions
 Aggregates  Calculating Row
 CUME Subtotals
 MOVINGAVG  BREAK BY
 MOVINGSUM  Sophisticated Date
 RANK Time Support
 TERTILE  DATEDIFF
 RATIOTOREPORT  Using SubQueries
in calculations

114
Using SubQueries in Calculations
select product, dollars as jun97_sales,
(select sum(s1.dollars)
from market mi, product pi, period, ti, sales si
where pi.product = product.product
and ti.year = period.year
and mi.city = market.city) as total97_sales,
100 * dollars/
(select sum(s1.dollars)
from market mi, product pi, period, ti, sales si
where pi.product = product.product
and ti.year = period.year
and mi.city = market.city) as percent_of_yr
from market, product, period, sales
where year = 1997
and month = ‘June’ and city like ‘Ahmed%’
order by product;

115
Course Overview
 The course: what and
how

 0. Introduction
 I. Data Warehousing
 II. Decision Support and
OLAP
 III. Data Mining
 IV. Looking Ahead

 Demos and Labs

116
II. On-Line Analytical Processing
(OLAP)

Making Decision
Support Possible
Limitations of SQL

“A Freshman in
Business needs
a Ph.D. in SQL”

-- Ralph Kimball
118
Typical OLAP Queries

 Write a multi-table join to compare sales for each

product line YTD this year vs. last year.
 Repeat the above process to find the top 5
product contributors to margin.
 Repeat the above process to find the sales of a
product line to new vs. existing customers.
 Repeat the above process to find the customers
that have had negative sales growth.

119
What Is OLAP?
 Online Analytical Processing - coined by
EF Codd in 1994 paper contracted by
Arbor Software*
 Generally synonymous with earlier terms such as
Decisions Support, Business Intelligence, Executive
Information System
 OLAP = Multidimensional Database
 MOLAP: Multidimensional OLAP (Arbor Essbase,
Oracle Express)
 ROLAP: Relational OLAP (Informix MetaCube,
Microstrategy DSS Agent)
* Reference: http://www.arborsoft.com/essbase/wht_ppr/coddTOC.html

120
The OLAP Market
 Rapid growth in the enterprise market
 1995: $700 Million
 1997: $2.1 Billion
 Significant consolidation activity among
major DBMS vendors
 10/94: Sybase acquires ExpressWay
 7/95: Oracle acquires Express
 11/95: Informix acquires Metacube
 1/97: Arbor partners up with IBM
 10/96: Microsoft acquires Panorama
 Result: OLAP shifted from small vertical
niche to mainstream DBMS category
121
Strengths of OLAP

 It is a powerful visualization paradigm

 It provides fast, interactive response times
 It is good for analyzing time series
 It can be useful to find some clusters and
outliers
 Many vendors offer OLAP tools

122
OLAP Is FASMI
Fast
Analysis
Shared
Multidimensional
Information

Nigel Pendse, Richard Creath - The OLAP Report

123
 Multi-dimensional Data

“Hey…I sold $100M worth of goods”

Dimensions: Product, Region, Time
Hierarchical summarization paths
n
io

W
eg

S
R

N Product Region Time

Product

Juice Industry Country Year

Cola
Milk
Cream Category Region Quarter
Toothpaste
Soap
1 2 34 5 6 7 Product City Month Week

Month
Office Day124
Data Cube Lattice
 Cube lattice
 ABC
AB AC BC
A B C
none
 Can materialize some groupbys, compute others
on demand
 Question: which groupbys to materialze?
 Question: what indices to create
 Question: how to organize data (chunks, etc)

125
Visualizing Neighbors is simpler
Month Store Sales
1 2 3 4 5 6 7 8 Apr 1
Apr Apr 2
Apr 3
May Apr 4
J un Apr 5
J ul Apr 6
Apr 7
Aug Apr 8
Sep May 1
Oct May 2
May 3
Nov May 4
Dec May 5
J an May 6
May 7
Feb May 8
Mar J un 1
J un 2

126
A Visual Operation: Pivot
(Rotate)
NY
LA

n th
SF

Juice M
o
10
Cola

Region
47
Milk
Crea 30
m 12 Product
3/1 3/2 3/3
3/4
Date 127
 “Slicing and Dicing”

The Telecomm Slice

Product

Household

Telecomm ns
i o
eg
Video R Europe
Far East
Audio India

Retail Direct Special Sales Channel

128
Roll-up and Drill Down
Higher Level of
Aggregation
 Sales Channel

Drill-Down
 Region
Roll Up

 Country
 State
 Location Address
 Sales
Representative
Low-level
Details
129
Nature of OLAP Analysis
 Aggregation -- (total sales,
percent-to-total)
 Comparison -- Budget vs.
Expenses
 Ranking -- Top 10, quartile
analysis
 Access to detailed and
aggregate data
 Complex criteria
specification
 Visualization
130
Organizationally Structured Data
 Different Departments look at the same
detailed data in different ways. Without
the detailed, organizationally structured
data as a foundation, there is no
reconcilability of data

marketing

sales
finance
manufacturing
131
Multidimensional Spreadsheets
 Analysts need
spreadsheets that support
 pivot tables (cross-tabs)
 drill-down and roll-up
 slice and dice
 sort
 selections
 derived attributes
 Popular in retail domain

132
OLAP - Data Cube

 Idea: analysts need to group data in many different

ways
 eg. Sales(region, product, prodtype, prodstyle,
date, saleamount)
 saleamount is a measure attribute, rest are
dimension attributes
 groupby every subset of the other attributes
materialize (precompute and store) groupbys to
give online response
 Also: hierarchies on attributes: date -> weekday,
date -> month -> quarter -> year

133
SQL Extensions
Front-end tools require
 Extended Family of Aggregate Functions
rank, median, mode
 Reporting Features
running totals, cumulative totals
 Results of multiple group by
total sales by month and total sales by
product
 Data Cube

134
 Relational OLAP: 3 Tier DSS
Data Warehouse ROLAP Engine Decision Support Client

Database Layer Application Logic Layer Presentation Layer

Store atomic Generate SQL Obtain multi-

data in execution plans dimensional
industry in the ROLAP reports from
standard engine to obtain the DSS Client.
RDBMS. OLAP
functionality. 135
MD-OLAP: 2 Tier DSS
MDDB Engine MDDB Engine Decision Support Client

Database Layer Application Logic Layer Presentation Layer

Store atomic data in a Obtain multi-

proprietary data structure dimensional
(MDDB), pre-calculate as many reports from the
outcomes as possible, obtain DSS Client.
OLAP functionality via proprietary
algorithms running against this 136
data.
 Typical OLAP Problems
Data Explosion

Data Explosion Syndrome

Number of Aggregations

70000
65536
60000
50000
40000
30000
20000
16384
10000
4096
0 16 81 256 1024
2 3 4 5 6 7 8

(4 levels in each dimension) Number of Dimensions

137
Microsoft TechEd’98
Metadata Repository
 Administrative metadata
 source databases and their contents
 gateway descriptions
 warehouse schema, view & derived data definitions
 dimensions, hierarchies
 pre-defined queries and reports
 data mart locations and contents
 data partitions
 data extraction, cleansing, transformation rules, defaults
 data refresh and purging rules
 user profiles, user groups
 security: user authorization, access control

138
Metdata Repository .. 2

 Business data
 business terms and definitions
 ownership of data
 charging policies
 operational metadata
 data lineage: history of migrated data and
sequence of transformations applied
 currency of data: active, archived, purged
 monitoring information: warehouse usage
statistics, error reports, audit trails.
139
Recipe for a Successful
Warehouse
 For a Successful Warehouse

From Larry Greenfield, http://pwp.starnetinc.com/larryg/index.htm

 From day one establish that warehousing
is a joint user/builder project
 Establish that maintaining data quality will
be an ONGOING joint user/builder
responsibility
 Train the users one step at a time
 Consider doing a high level corporate data
model in no more than three weeks
141
For a Successful Warehouse
 Look closely at the data extracting,
cleaning, and loading tools
 Implement a user accessible automated
directory to information stored in the
warehouse
 Determine a plan to test the integrity of
the data in the warehouse
 From the start get warehouse users in the
habit of 'testing' complex queries

142
For a Successful Warehouse
 Coordinate system roll-out with network
administration personnel
 When in a bind, ask others who have done
the same thing for advice
 Be on the lookout for small, but strategic,
projects
 Market and sell your data warehousing
systems

143
Data Warehouse Pitfalls
 You are going to spend much time extracting,
cleaning, and loading data
 Despite best efforts at project management, data
warehousing project scope will increase
 You are going to find problems with systems
feeding the data warehouse
 You will find the need to store data not being
captured by any existing system
 You will need to validate data not being validated
by transaction processing systems

144
Data Warehouse Pitfalls
 Some transaction processing systems feeding the
warehousing system will not contain detail
 Many warehouse end users will be trained and
never or seldom apply their training
 After end users receive query and report tools,
requests for IS written reports may increase
 Your warehouse users will develop conflicting
business rules
 Large scale data warehousing can become an
exercise in data homogenizing

145
Data Warehouse Pitfalls
 'Overhead' can eat up great amounts of disk space
 The time it takes to load the warehouse will expand
to the amount of the time in the available window...
and then some
 Assigning security cannot be done with a
transaction processing system mindset
 You are building a HIGH maintenance system
 You will fail if you concentrate on resource
optimization to the neglect of project, data, and
customer management issues and an understanding
of what adds value to the customer

146
DW and OLAP Research Issues
 Data cleaning
 focus on data inconsistencies, not schema differences
 data mining techniques
 Physical Design
 design of summary tables, partitions, indexes
 tradeoffs in use of different indexes
 Query processing
 selecting appropriate summary tables
 dynamic optimization with feedback
 acid test for query optimization: cost estimation, use of
transformations, search strategies
 partitioning query processing between OLAP server and
backend server.

147
DW and OLAP Research Issues ..
2
 Warehouse Management
 detecting runaway queries
 resource management
 incremental refresh techniques
 computing summary tables during load
 failure recovery during load and refresh
 process management: scheduling queries, load
and refresh
 Query processing, caching
 use of workflow technology for process
management
148
Products, References, Useful
Links
Reporting Tools
 Andyne Computing -- GQL
 Brio -- BrioQuery
 Business Objects -- Business Objects
 Cognos -- Impromptu
 Information Builders Inc. -- Focus for Windows
 Oracle -- Discoverer2000
 Platinum Technology -- SQL*Assist, ProReports
 PowerSoft -- InfoMaker
 SAS Institute -- SAS/Assist
 Software AG -- Esperant
 Sterling Software -- VISION:Data

150
OLAP and Executive Information
Systems
 Andyne Computing -- Pablo  Microsoft -- Plato
 Arbor Software -- Essbase  Oracle -- Express
 Cognos -- PowerPlay  Pilot -- LightShip
 Comshare -- Commander  Planning Sciences --
OLAP Gentium
 Holistic Systems -- Holos  Platinum Technology --
 Information Advantage -- ProdeaBeacon, Forest &
AXSYS, WebOLAP Trees

 Informix -- Metacube  SAS Institute -- SAS/EIS,

OLAP++
 Microstrategies --DSS/Agent
 Speedware -- Media

151
Other Warehouse Related
Products
Data extract, clean, transform,
refresh
 CA-Ingres replicator
 Carleton Passport
 Prism Warehouse Manager
 SAS Access
 Sybase Replication Server
 Platinum Inforefiner, Infopump

152
Extraction and Transformation
Tools
 Carleton Corporation -- Passport
 Evolutionary Technologies Inc. -- Extract
 Informatica -- OpenBridge
 Information Builders Inc. -- EDA Copy Manager
 Platinum Technology -- InfoRefiner
 Prism Solutions -- Prism Warehouse Manager
 Red Brick Systems -- DecisionScape Formation

153
Scrubbing Tools
Apertus -- Enterprise/Integrator
Vality -- IPE
Postal Soft

154
Warehouse Products
 Computer Associates -- CA-Ingres
 Hewlett-Packard -- Allbase/SQL
 Informix -- Informix, Informix XPS
 Microsoft -- SQL Server
 Oracle -- Oracle7, Oracle Parallel Server
 Red Brick -- Red Brick Warehouse
 SAS Institute -- SAS
 Software AG -- ADABAS
 Sybase -- SQL Server, IQ, MPP
155
Warehouse Server Products
Oracle 8
Informix
 Online Dynamic Server
 XPS --Extended Parallel Server
 Universal Server for object relational
applications
Sybase
 Adaptive Server 11.5
 Sybase MPP
 Sybase IQ

156
Warehouse Server Products
Red Brick Warehouse
Tandem Nonstop
IBM
 DB2 MVS
 Universal Server
 DB2 400
Teradata

157
Other Warehouse Related
Products
Connectivity to Sources
 Apertus
 Information Builders EDA/SQL
 Platimum Infohub
 SAS Connect
 IBM Data Joiner
 Oracle Open Connect
 Informix Express Gateway

158
Other Warehouse Related
Products
Query/Reporting Environments
 Brio/Query
 Cognos Impromptu
 Informix Viewpoint
 CA Visual Express
 Business Objects
 Platinum Forest and Trees

159
4GL's, GUI Builders, and PC
Databases
Information Builders --Focus
Lotus -- Approach
Microsoft -- Access, Visual Basic
MITI -- SQR/Workbench
PowerSoft -- PowerBuilder
SAS Institute -- SAS/AF

160
Data Mining Products
DataMind -- neurOagent
Information Discovery -- IDIS
SAS Institute -- SAS/Neuronets

161
Data Warehouse
 W.H. Inmon, Building the Data Warehouse,
Second Edition, John Wiley and Sons, 1996
 W.H. Inmon, J. D. Welch, Katherine L.
Glassey, Managing the Data Warehouse,
John Wiley and Sons, 1997
 Barry Devlin, Data Warehouse from
Architecture to Implementation, Addison
Wesley Longman, Inc 1997

162
Data Warehouse
 W.H. Inmon, John A. Zachman, Jonathan G.
Geiger, Data Stores Data Warehousing
and the Zachman Framework, McGraw Hill
Series on Data Warehousing and Data
Management, 1997
 Ralph Kimball, The Data Warehouse
Toolkit, John Wiley and Sons, 1996

163
OLAP and DSS
 Erik Thomsen, OLAP Solutions, John Wiley
and Sons 1997
 Microsoft TechEd Transparencies from
Microsoft TechEd 98
 Essbase Product Literature
 Oracle Express Product Literature
 Microsoft Plato Web Site
 Microstrategy Web Site

164
Data Mining
 Michael J.A. Berry and Gordon Linoff, Data
Mining Techniques, John Wiley and Sons
1997
 Peter Adriaans and Dolf Zantinge, Data
Mining, Addison Wesley Longman Ltd.
1996
 KDD Conferences

165
Other Tutorials
 Donovan Schneider, Data Warehousing Tutorial,
Tutorial at International Conference for
Management of Data (SIGMOD 1996) and
International Conference on Very Large Data
Bases 97
 Umeshwar Dayal and Surajit Chaudhuri, Data
Warehousing Tutorial at International Conference
on Very Large Data Bases 1996
 Anand Deshpande and S. Seshadri, Tutorial on
Datawarehousing and Data Mining, CSI-97

166
Useful URLs
 Ralph Kimball’s home page
 http://www.rkimball.com
 Larry Greenfield’s Data Warehouse
Information Center
 http://pwp.starnetinc.com/larryg/
 Data Warehousing Institute
 http://www.dw-institute.com/
 OLAP Council
 http://www.olapcouncil.com/

167