- Most BI projects usually fail. It is not due to the errors in the construction steps but due to the inappropriate design steps and methodologies.
- Waterfall method is not appropriate for BI. A method like developmental spiral may be more appropriate. Ref: [W. A. Giovinazzo ]
- Steps in developmental spiral: Definition, Analysis, Design, Development, Implementation/Deployment, Maintenance, Evaluation
- Developmental spiral method is effective for BI only when Object Oriented approach is taken
- In developmental spiral, object modeling is used to represent objects in the related business areas. In object modeling, objects are represented as rectangles, the rectangles can be treated as classes. Classes/objects can have sub-classes/sub-objects. Objects have attributes. Objects also have some special properties like: accelerators, stops.
- Definition – first step of developmental spiral: Define a short description what you want to build – like the problem statement in a research proposal. Example: I want to build a data mart that will provide data to support the analysis of customer demographics such as age, gender, and income for all product lines
- Analysis Phase: In this step, analyze the requirements of the BI system, the nature of the BI system, the expectations from the BI system. Interviewing different parties will help a lot:
- Interview the sponsor to get a detail understanding of the mission. The questions will include: Data-related(which objects are related to the business needs?), user-related: who are the users, system evaluation: what parameters will make the project a success?
- Interviewing management: To understand management’s expectation of the data warehouse. Questions may include: User profile related (what types of systems you use/will use to access BI?), System vision: what do you see as the objective of the data warehouse, System Evaluation: what parameters will make the project a success?
- Interviewing Users : Interviewing users who will use the system
- Interviewing information technology staffs: To know the current state of the organization’s information systems
- Design phase: output – object model, implementation model
- Start with the analysis phase information. The design phase is complete when all the objects related to the mission are well defined. An object diagram needs to be drawn showing all objects, relationships, attributes.
- Another chart listing all the attributes and their data types will be useful. Objects will have relations like super classes or sub-classes.
- Cardinality relationships: exactly one, one to many, zero or one, zero to many
- aggregation may be required in some cases. Aggregation: an object is composed of other objects.
- Implementation Model: Data warehouse databases are multi dimensional databases. Objects are represented in more than two dimensions. Common practice is: three dimensional databases. Such as an object/table represented with three dimensions (called cube) like: product, dealer, and time. RDBMS’s are two dimensional.
- Data cubes: have six possible different views [permutation of three dimensions].
- Why not always multidimensional databases: because of space requirements.
- Star schema: Provides a multidimensional flavor in two dimensional relational databases.
- Star schema uses a concept called fact tables to bind dimensions to create a multidimensional space.
- Denormalization of the tables are utilized in star schema to create multi dimension.
- Dimension table: Think about a three dimensional cube. Each wall represents a dimension table. Or think it as a mathematical combination of the dimensions.
- All objects represented in the dimensions can slowly change over time – slowly changing dimensions. There are many approaches to address this issue.
- Snowflakes: normalizing dimensions – not always a great thing
- Implementation considerations: Parallel processing, Bitmapped Indexing, Star Query Optimization, Summation Tables
From: http://sitestree.com/?p=3565
Categories:Data Warehouse, Data Warehouse – 001, Data Warehouse, Data Warehouse Misc
Tags:
Post Data:2016-07-06 15:54:10
Shop Online: https://www.ShopForSoul.com/
(Big Data, Cloud, Security, Machine Learning): Courses: http://Training.SitesTree.com
In Bengali: http://Bangla.SaLearningSchool.com
http://SitesTree.com
8112223 Canada Inc./JustEtc: http://JustEtc.net (Software/Web/Mobile/Big-Data/Machine Learning)
Shop Online: https://www.ShopForSoul.com/
Medium: https://medium.com/@SayedAhmedCanada
