•1
Advanced MD Modeling and
MD Database Implementation
2
Overview
•Handling Changes in Dimensions
•Coordinating Data Cubes / Data Marts
•Multidimensional Database Implementation
•2
4
Changing Dimensions
•In the last lecture, we assumed that
dimensions are stable over time
New rows in dimension tables can be inserted
Existing rows do not change
This is not a realistic assumption
•We now study techniques for handling changes in
dimensions
•“Slowly changing dimensions” phenomenon
Dimension information change, but changes are
not frequent
Still assume that the schema is fixed
5
Handling Changes in Dimensions
•Handling change over time
•Changes in dimensions
1. No special handling
2. Versioning dimension values
2A. Special facts
2B. Timestamping
3. Capturing the previous and the current value
4. Split into changing and constant attributes
•3
6
Example
•Attribute values in
dimensions vary over time
A store changes Size
A product changes
Description
Districts are changed
•Problems
Dimensions not updated
DW is not up-to-date
Dimensions updated in a
straightforward way
incorrect information in
historical data
TimeID
StoreID
ProductID
…
ItemsSold
Amount ProductID
Description
Brand
PCategory
StoreID
Address
City
District
Size
SCategory
TimeID
Weekday
Week
Month
Quarter
Year
DayNo
Holiday
timeline
change
??
Sales fact
Time dim.
Store dim.
Product dim.
7
Example
TimeID
StoreID
ProductID
…
ItemsSold
Amount
…
StoreID
Address
City
District
Size
SCategory
…Sales fact
Time dim.
Store dim.
Product dim.
2000
ItemsSold
001
……StoreID
250
Size
001
……StoreID
Sales fact table Store dimension table
The store in Aalborg has
the size of 250 sq. metres.
On a certain day,
customers bought 2000
apples from that store.
8
Solution 1: No Special Handling
2000
ItemsSold
001
……StoreID
250
Size
001
……StoreID
2000
ItemsSold
001
……StoreID
450
Size
001
……StoreID
2000001
3500
ItemsSold
001
……StoreID
450
Size
001
……StoreID
Sales fact table Store dimension table
The size of a store expands
A new fact arrives
What’s the problem here?
Solution 1
