Consider an example where business analyst uses the systems containing operational data (the data that runs the daily transactions of your business). Analysts can use information about, which products were sold in which regions at what time of the year, to look for anomalies or to project future sales. However, there are several problems if analyst accesses operational data directly:
- He might not have the expertise to query the operational database. For example, querying IMS databases requires an application program that uses a specialized type of data manipulation language. In general, those programmers who have the expertise to query the operational database have a full-time job in maintaining the database and its applications.
- Performance is critical for many operational databases, such as databases for a bank. The system cannot handle users making ad-hoc queries.
- The operational data generally is not in the best format to be used for reporting queries
Data warehousing solves these problems. In data warehousing, you create stores of informational data data that is extracted from the operational data and then transformed for reporting and decision making. For example, a data warehousing tool might copy all the sales data from the operational database, perform calculations to summarize the data, and write it to a new database. End-users can query the new database (the warehouse) without impacting the operational databases.
- The purpose of data warehouse is to store data consistently across the organization and to make the organizational information accessible.
- It is adaptive and resilient source of information. When new data is added to the Data Warehouse, the existing data and technologies are not disrupted. The design of separate data marts that make up the data warehouse must be distributed and incremental. Anything else is a compromise.
- The data warehouse not only controls the access to the data, but gives its owners great visibility into the uses and abuses of the data, even after it has left the data warehouse.
- Data warehouse is the foundation for decision-making.
Fathers of Data Warehousing
William H. Inmon Biography
Bill Inmon, is recognized as the “father of the data warehouse” and co-creator of the “Corporate Information Factory.” He has 35 years of experience in database technology management and data warehouse design. He is known globally for his seminars on developing data warehouses and has been a keynote speaker for every major computing association and many industry conferences, seminars, and tradeshows.
As an author, Bill has written about a variety of topics on the building, usage, and maintenance of the data warehouse and the Corporate Information Factory. He has written more than 650 articles, many of them have been published in major computer journals such as Datamation, ComputerWorld, and Byte Magazine. Bill is currently a columnist with Data Management Review, and has been since its inception. He has published 45 books; one sold over half a million copies, 21 have been book club selections with publishers such as Prentice-Hall, John Wiley, and QED. Translations of various books have been done in Chinese, Dutch, French, German, Japanese, Korean, Portuguese, Russian, and Spanish.
Ralph Kimball Biography
Ralph Kimball is known worldwide as an innovator, writer, educator, speaker and consultant in the field of data warehousing. He has remained steadfast in his long-term conviction that data warehouses must be designed to be understandable and fast. His books on dimensional design techniques have become the all time best sellers in data warehousing. To date Ralph has written more than 100 articles and columns for Intelligent Enterprise and its predecessors, winning the Readers Choice Award five years in a row.
After receiving a Ph.D. in 1972 from Stanford in electrical engineering (specializing in man-machine systems), Ralph joined the Xerox Palo Alto Research Center (PARC). At PARC Ralph co-invented the Xerox Star Workstation, the first commercial product to use mice, icons and windows.
Ralph then became vice president of applications at Metaphor Computer Systems, pioneering decision support software and services provider. As a hands-on manager, he developed the Capsule Facility in 1982. The Capsule was a graphical programming technique which connected icons together in a logical flow, allowing a very visual style of programming for non-programmers. The Capsule was used to build reporting and analysis applications at Metaphor.
Ralph founded Red Brick Systems in 1986, serving as CEO until 1992. Red Brick Systems, now owned by IBM, was known for its lightning fast relational database optimized for data warehousing. Ralph Kimball Associates incorporated in 1992 to provide data warehouse consulting and education.
Elements of Data Warehousing
Typically in any organization the data is stored in various databases, usually divided up by the systems. There may be data for marketing, sales, payroll, engineering, etc. These systems might be legacy/mainframe systems or relational database systems.
The data coming from various source systems is first kept in a staging area. The staging area is used to clean, transform, combine, de-duplicate, household, archive, and to prepare source data for use in data warehouse. The data coming from source system is kept as it is in this area. This need not be based on relational terminology. Sometimes managers of the data are comfortable with normalized set of data. In these cases, normalized structure of the data staging storage is certainly acceptable. Also, staging area doesnt provide querying/presentation services.
Once the data is in staging area, it is cleansed, transformed and then sent to Data warehouse. You may or may not have ODS before transferring data to Data Warehouse.
The data in Data Warehouse has to be easily manipulated in order to answer the business questions from management and other users. This is accomplished by connecting the data to fast and easy-to-use tools known as Online Analytical Processing (OLAP) tools. OLAP tools can be thought of as super high-speed forklifts that have knowledge of the warehouse and the operators built into them in order to allow ordinary people off the street to jump in and quickly find products by asking English-like questions. Within the OLAP server, data is reorganized to meet the reporting and analysis requirements of the business, including:
- Exception reporting
- Ad-hoc analysis
- Actual vs. budget reporting
- Data mining (looking for trends or anomalies in the data)
In order to process business queries at high speed, answers to common questions are preprocessed in some OLAP servers, resulting in exceptional query responses at the cost of having an OLAP database that may be several times bigger than the data warehouse itself.
Data mart is a logical subset of complete data warehouse. It is often viewed as the restriction of data warehouse to a single business process or to a group of related business processes targeted toward a particular business group. For example an organization may have a data mart for Sales or Inventory.
Quick Reference Guide to Dimensional Modeling
Dimensional modeling is the design concept used by many data warehouse designers to build their data warehouse. Dimensional model is the underlying data model used by many of the commercial OLAP products available today in the market. Designing a data warehouse is very different from designing an online transaction processing (OLTP) system. In contrast to an OLTP system in which the purpose is to capture high rates of data changes and additions, the purpose of a data warehouse is to organize large amounts of stable data for ease of analysis and retrieval. Because of these differing purposes, there are many considerations in data warehouse design that differ from OLTP database design. In dimensional model, all data is contained in two types of tables called Fact Table and Dimension Table.
Each data warehouse or data mart includes one or more fact tables. The fact table captures the data that measures the organizations business operations. A fact table might contain business sales events such as cash register transactions or the contributions and expenditures of a nonprofit organization. Fact tables usually contain large numbers of rows, sometimes in the hundreds of millions of records when they contain one or more years of history for a large organization. A key characteristic of a fact table is that it contains numerical data (facts) that can be summarized to provide information about the history of the operation of the organization. Each fact table also includes a multipart index that contains as foreign keys the primary keys of related dimension tables, which contain the attributes of the fact records. Fact tables should not contain descriptive information or any data other than the numerical measurement fields and the index fields that relate the facts to corresponding entries in the dimension tables. An example of fact table is Sales_Fact table that might contain the information like sale_amount, unit_price, discount, etc.
Dimension tables contain attributes that describe fact records in the fact table. Some of these attributes provide descriptive information; others are used to specify how fact table data should be summarized to provide useful information to the analyst. Dimension tables contain hierarchies of attributes that aid in summarization. For example, a dimension containing product information would often contain a hierarchy that separates products into categories such as food, drink, and non-consumable items, with each of these categories further subdivided a number of times until the individual product is reached at the lowest level.
Dimensional modeling produces dimension tables in which each table contains fact attributes that are independent of those in other dimensions. For example, a customer dimension table contains data about customers, a product dimension table contains information about products, and a store dimension table contains information about stores. Queries use attributes in dimensions to specify a view into the fact information. For example, a query might use the product, store, and time dimensions to ask the question “What was the cost of non-consumable goods sold in the northeast region in 1999?” Subsequent queries might drill down along one or more dimensions to examine more detailed data, such as “What was the cost of kitchen products in New York City in the third quarter of 1999?” In these examples, the dimension tables are used to specify how a measure (sale_amount) in the fact table is to be summarized.
Consider an example of Sales_Fact table and the various attributes that describe this fact are Store, Product, Time and say Sales Person. In this case we will have four dimension tables, viz. Store_Dimension, Product_Dimension, Time_Dimension and Sales_Person_Dimension.
You may notice that all of these dimensions contain a Key field. This is called Surrogate Key. This key is substitute for a natural key in dimensions (e.g., in Sales_Person_Dimension, we have natural key as ID). In a data warehouse a surrogate key is a generalization of the natural production key and is one of the basic elements of data warehouse.
As a fact table is described by the four dimension tables described above, it will contain the Surrogate Keys of all these dimensions. This is how the Sales_Fact table will look like:
Now if you carefully look at the structure of above tables and how they are linked the schema will look like this:
You can easily tell that this looks like a STAR. Hence its known as Star Schema.
Advantages of having Star Schema
- Star Schema is very easy to understand, even for non technical business managers
- Star Schema provides better performance and smaller query times
- Star schema is easily extensible and will handle future changes easily
Slowly Changing Dimensions
Handling changes to dimensional data across time is the most important aspect in designing a data warehouse. In dimensional modeling, there is a very rare chance that a dimension will remain static over time. For example, a customer address may change; a company may phase out old products and introduce new products. What if a customer name changes, sales person changes his region of sale or a company assigns new sales territory. How to record the history or preserve the old version of history? Here comes the concept of Slowly Changing Dimensions. The term Slowly Changing Dimension is about variation in dimensional attributes over time. The word slowly, in this context, might seem incorrect. A sales person may change his territory rapidly. But in general, when compared to measures in fact table, the changes in dimensions occur slowly.
Types of Slowly Changing Dimensions
In reference to Figure 3 above, lets say a sales person changes his region of sale. We may handle this change in several ways. These methods fall in various categories based on companys need to preserve an accurate history of dimensional changes. Ralph Kimball categorized the dimensional changes into three categories
- Type One: Changes that overwrite history
- Type Two: Preserve history
- Type Three: Preserve a version of history
Type One (Overwrite History)
A type one change overwrites existing dimensional attribute with new information. In Sales Person Region change example, the old region name will be overwritten by the new region. Say, a sales person Rob, has territory as ASIA.
Now, if he starts looking after NorthWest Region, by implementing Type 1 dimension, the dimension table will look like:
- This is the easiest way to handle the Slowly Changing Dimension problem, since there is no need to keep track of the old information.
- All history is lost. By applying this methodology, it is not possible to trace back in history. For example, in this case, the company would not be able to know that Christina lived in Illinois before.
Type Two (Preserve History)
A Type Two change writes a record with the new attribute information and preserves a record of the old dimensional data. Type Two changes let you preserve historical data. Implementing Type Two changes within a data warehouse might require significant analysis and development. Type Two changes accurately partition history across time more effectively than other types. However, because Type Two changes add records, they can significantly increase the database’s size.
In our example, lets say we identify Region as Type Two attribute. This can be handled in this way using:
- This allows us to accurately keep all historical information.
- This will cause the size of the table to grow fast. In cases where the number of rows for the table is very high to start with, storage and performance can become a concern.
- This necessarily complicates the ETL process.
Type Three (Preserve a Version of History)
You usually implement Type Three changes only if you have a limited need to preserve and accurately describe history, such as when someone gets married and you need to retain the previous name. Instead of creating a new dimensional record to hold the attribute change, a Type Three change places a value for the change in the original dimensional record. You can create multiple fields to hold distinct values for separate points in time. In the case of a region change example, you could create an OLD_REGION and NEW_REGION field and a REGION_CHANGE_EFF_DATE field to record when the change occurs. This method preserves the change. But how would you handle a second name change, or a third, and so on? The side effects of this method are increased table size and, more important, increased complexity of the queries that analyze historical values from these old fields. After more than a couple of iterations, queries become impossibly complex, and ultimately you’re constrained by the maximum number of attributes allowed on a table.
This is how the table will look like in Type Three change:
|Sales_Person_Key||ID||Name||Old Region||New Region||…|
- This does not increase the size of the table, since new information is updated.
- This allows us to keep some part of history.
- Type 3 will not be able to keep all history where an attribute is changed more than once. For example, if Christina later moves to Texas on December 15, 2003, the California information will be lost.
Because most business requirements include tracking changes over time, data warehouse architects commonly implement Type Two changes. A data warehouse might use Type Two changes for all attributes in all tables. As an alternative, you can implement a mix of Type One and Type Two changes at an attribute level by implementing Type 2 changes for only attributes whose historical values are important when you’re slicing and dicing. For example, users might not need to know an individual’s previous name if a name change occurs, so a Type One change would suffice. Users might want the system to show only the person’s current name. However, if the company reassigns sales territories, users might need to track who sold what, at what time, and in what territory, necessitating a Type Two change.
Although most data warehouses include Type Two changes, you need to seriously examine the business need to record historical data. Implementing Type Two changes might be necessary, but those changes will increase the database size, degrade performance, and lengthen the development time. You need to carefully evaluate using a Type Two implementation, a Type One implementation, or a hybrid implementation.