Designing Data Marts for Data Warehouses

1
Designing Data Marts for Data Warehouses

• ACM Transactions on software Engineering and
  Methodology, October,2001
• Present by
  L.W.Lu

2
Outline

• Introduction
• Design Data Warehouse Schemas
• Conclusion

3
Introduction

• This paper presents a method to support the
  identification and design of data marts. The
  method is based on three basic steps. The first
  top-down step makes it possible to elicit and
  consolidate user requirements and expectations.

4
Introduction (conti.)

• The second bottom-up step extracts candidate data
  marts from the conceptual schema of the
  information system. The final step compares idel
  and candidate data marts to derive a collection
  of data marts that are supported by the
  underlying information system and maximally
  satisfy user requirements.

5
Top-Down Phase

• Collect user requirements through interviews.

6
Top-Down Phase (conti.)
7
Top-Down Phase (conti.)

• Represent requirements as GQM goals
• (a) For each goal, fill in the goal definition,
  i.e., define the object of the study, the
  purpose, the quality focus, the viewpoint, and
  the environment.

8
Top-Down Phase (conti.)

• (b) For each goal, detail the goal by filling in
  the abstraction sheet define the quality focus,
  the variation factors, the baseline hypotheses,
  and the impact on baseline hypotheses.

9
(No Transcript)
10
Top-Down Phase (conti.)
11
Top-Down Phase (conti.)

• (c) Compare and assess the goals identify
  similarities and implications among goals to
  reduce them to a manageable number, i.e., merge
  related goals and drop goals that are subsumed by
  others

12
Top-Down Phase (conti.)

• Derive ideal schema fragments for each goal.

13
Bottom-up Analysis(Derivation of candidate star
schemas from DB conceptual schemas)

• Obtain the Star Join Graphs
• (a) Map the E/R schema into a connectivity graph.

14
(No Transcript)
15
Bottom-up Analysis (conti.)

• (b) Run the Snowflake Graph Algorithm, to find
  all the possible snowflake graphs.

16
Bottom-up Analysis (conti.)
17
Integration and Ranking

• Match the ideal schema with the star join graphs
• (a) Identify common terms between the ideal
  schema and the star join graphs, through
  direct/indirect mapping.
• (b) Match ideal and candidate schemas, taking
  into account the matching attributes of the
  schemas, the matching dimensions, the additional
  attributes, and the additional dimensions .
• (c) Rank the solutions.

18
Conclusion

• The combination of top-down and bottom-up steps
  helps in coherently evaluating both the coverage
  of users requirements and the feasibility of
  data warehouse design.