A Multidimensional and Multiversion Structure for OLAP Applications

Nov 8 2002 DOLAP 2002McLean USA

A Multidimensional and Multiversion Structure for OLAP Applications

Mathurin Body1,2, Maryvonne Miquel2, Yvan Bédard1,2, Anne Tchounikine2

1 Centre de recherche en Géomatique, Univ Laval, Québec, Canada

2 Institut National des Sciences Appliquées, Lyon, France


Purpose

• Handle evolutions in Multidimensional Structures

• Compare data into static structures

• Provide a new conceptual model

• Define evolution operators

• Give solutions and tools for implementation


Multidimensional Models

Date

Gender

City Country

Static organization of data

Fact Table

Nb of Births

Star or Snowflake representation

Issues


First Case StudyLocation dimension:

2001 2002

D1 100 100

D2 50 100

D3 100 50

Query: « Total number of births per year and city ? »

2001 2002 Evo

C1 150 100

C2 100 150

2001 2002 Evo

C1 150 200

C2 100 50

2001 2002 Evo

C1 100 100

C2 150 150

C1 C2

D1 D2 D3

2001

C1 C2

D1 D2 D3

2002

1. Exact view 2. Mapped info into 2001 organization

3. Mapped info into 2002 organization

Issues


Second Case StudyLocation dimension:

2001 2002

D 100 -

D1 - 150

D2 - 50

Query: « Total number of births per year and district ? »

2001 2002 Evo

D 100 - ?

D1 - 150 ?

D2 - 50 ?

2001 2002 Evo

D 100 200

2001 2002 Evo

D1 40* 150

D2 60** 50

C1

D

2001

1. Exact view 2. First Structure 3. Second Structure

2002

C1

D1 D2

* D1 ~ 40 % of the births of D1** D2 ~ 60 % of the births of D1

Issues


Existing Solutions (1/2)

Related works

• Updating models (M. Blaschka, C. Hurtado, A.O. Mendelzon and A. Vaisman)

+ Pragmatic way+ Allow temporal comparison – Corrupt data– lose data (e.g. deletion of a member)– Hiding evolutions


Existing Solutions (2/2)

Related works

• Tracking history models(R. Bliujute, P. Chamoni and S. Stock, J. Eder and C. Koncilia, R. Kimball, A.O. Mendelzon

and A. Vaisman, T.B. Pedersen, C.S. Jensen and C.E. Dyreson )

+ Temporally consistent representation+ Evolutions kept – Only one representation of data (no comparison across

time)– Limit of data analysis+ Mapping functions (J. Eder and C. Koncilia)

+ Timestamps on the elements of multidimensional database (A.O. Mendelzon and A. Vaisman)


Our Objectives

• For the administrators:– Integrate all kinds of evolution in a

multidimensional structure– Take into account complex dimension structures

• For the users:– Choose between different modes of representation– End user tools for analyse

Proposal


Evolutions in multidimensional structures

Proposal

Dimension schema evolution

•Creation and deletion of a dimension•Creation and deletion of a hierarchy•Creation and deletion of a level•Move of a level in the hierarchical schema structure

Evolution members: simple operations

•Creation of a member•Deletion of a member•Transformation of a member (change of an attribute, its name or meaning…)•Merging of n members into one member•Splitting of one member into n members•Reclassification of a member in the dimension structure

Evolution on members : Exples of complex operations•Decreasing: splitting and deletion•Increasing: creation and merging•Partial annexation: splitting and merging


Conceptual Model: Temporal Multidimensional Schema

Zaïre Dem. Rep. of Congo

[1990 ; 1997] [1997 ; Now]

Temporal Dimension: - Evolution of the hierarchical

structure of the members

Mapping Relationship: - keeping transition links between

member versions D1

D

D2

f : x x

f : x 0.4 x

f : x 0.6 x

f : x x Confidence factor:

- Evaluation of the confidence associated to a mapping

(exact mapping)

(approximated mapping) (exact

mapping)

(approximated mapping)

- Evolution of the members of a dimension

Member Version:

V1

D1 D2 D3

[98 ; 02] [98 ; 02]

[98 ; 00] [01 ; 02]

V2

P1

Proposal


Conceptual Model: MultiVersion Fact Table

Temporal Modes of Presentation: - Modes for the presentation of a

multidimensional request

MultiVersion Fact Table: - Fact Table with different temporal modes of

Presentation - Automatically deduced from the temporally

consistent fact table, the temporal dimensions and the mapping relationship

- a valid, unchanged structure over its given valid time.

Structure Version:

96 99 01

VS.1 VS.2 VS.3

- Temporally consistent mode - version VS.1- version VS.2- version VS.3

Proposal


Logical Model

- Temporal Modes of Presentation integrated in a new dimension.

Date

Gender

City Country

Fact Table

Nb of births

Confidence factor

- Confidence factors integrated as a new measure

T.M.P.

Implementation


Architecture

MultiVersion Data Warehouse

OLAP MultiVersion CubeTemporal Data Warehouse

- Extract the Structure Versions- Compute the transitive… of the

Mapping relation - Mapping of data

- Data Agregation- Multidimensional

Indexation

Implementation


PrototypeSale numbers and production cost per

product, district and month

Table de Faits

FK1 LocIDFK2 The_dateFK3 ProdID

VenteCout

Produits

PK ProdID

TypeProduitProduitFromTo

Time

PK The_date

The_monthThe_yearMonth_of_yearQuarter

Localisation

PK MemberID

MemberNameParentIDLevelNameFromTo

Conversions

MemberFromMemberToFacteurVenteFacteurCoutFiabiliteVenteFiabiliteCoutFromTo

Metadonnees

Star Schema of the temporal Data Warehouse

Implementation


Development ToolsVisual Basic Interface and

Proclarity Components

OLAP MultiVersion Cube(SQL Server Analysis

Service)

Access to the data cube

Data warehouse Repository (SQL Server)

Access to meta data

Implementation


End User Tools for Analysis

Implementation

Example of Metadata: describes the evolutions of the element pointed in the grid.

Grid: presents the values and their confidence factors

Comparative study:Two temporal modes are represented

Dimensions control: used to navigate trough the cube


End User Tools for Navigation:find the « best version »

Implementation

Parameters for each type of confidence

Rank of the temporal modes of presentation


Conclusion

A temporal multidimensional model for supporting evolutions on multidimensional structures

USER

• Navigate through different modes of presentation• Choose the interpretation he wants to give to his request• Be guide to select its best representation• Have access to metadata describing all evolutions of member versions

DESIGNER AND ADMINISTRATOR

• Model different kinds of hierarchical dimensions• Take into account all types of evolutions in the multidimensional structures• Implement this model on commercial OLAP environment

Extension

choose a temporal mode of presentation for each dimension

Documents

A Multidimensional and Multiversion Structure for OLAP Applications