Evolutionary Database Development
Waterfalls are wonderful tourist attractions. They are spectacularly bad strategies for organizing software development projects.
Modern software processes, also called methodologies, are all evolutionary in nature, requiring you to work both iteratively and incrementally. Examples of such processes include Rational Unified Process (RUP), Extreme Programming (XP), Scrum, Dynamic System Development Method (DSDM), the Crystal family, Team Software Process (TSP), Agile Unified Process (AUP), Enterprise Unified Process (EUP), Feature-Driven Development (FDD), and Rapid Application Development (RAD), to name a few. Working iteratively, you do a little bit of an activity such as modeling, testing, coding, or deployment at a time, and then do another little bit, then another, and so on. This process differs from a serial approach in which you identify all the requirements that you are going to implement, then create a detailed design, then implement to that design, then test, and finally deploy your system. With an incremental approach, you organize your system into a series of releases rather than one big one.
Furthermore, many of the modern processes are agile, which for the sake of simplicity we will characterize as both evolutionary and highly collaborative in nature. When a team takes a collaborative approach, they actively strive to find ways to work together effectively; you should even try to ensure that project stakeholders such as business customers are active team members. Cockburn (2002) advises that you should strive to adopt the "hottest" communication technique applicable to your situation: Prefer face-to-face conversation around a whiteboard over a telephone call, prefer a telephone call over sending someone an e-mail, and prefer an e-mail over sending someone a detailed document. The better the communication and collaboration within a software development team, the greater your chance of success.
Although both evolutionary and agile ways of working have been readily adopted within the development community, the same cannot be said within the data community. Most data-oriented techniques are serial in nature, requiring the creation of fairly detailed models before implementation is "allowed" to begin. Worse yet, these models are often baselined and put under change management control to minimize changes. (If you consider the end results, this should really be called a change prevention process.) Therein lies the rub: Common database development techniques do not reflect the realities of modern software development processes. It does not have to be this way.
Our premise is that data professionals need to adopt the evolutionary techniques similar to those of developers. Although you could argue that developers should return to the "tried-and-true" traditional approaches common within the data community, it is becoming more and more apparent that the traditional ways just do not work well. In Chapter 5 of Agile & Iterative Development, Craig Larman (2004) summarizes the research evidence, as well as the overwhelming support among the thought leaders within the information technology (IT) community, in support of evolutionary approaches. The bottom line is that the evolutionary and agile techniques prevalent within the development community work much better than the traditional techniques prevalent within the data community.
It is possible for data professionals to adopt evolutionary approaches to all aspects of their work, if they choose to do so. The first step is to rethink the "data culture" of your IT organization to reflect the needs of modern IT project teams. The Agile Data (AD) method (Ambler 2003) does exactly that, describing a collection of philosophies and roles for modern data-oriented activities. The philosophies reflect how data is one of many important aspects of business software, implying that developers need to become more adept at data techniques and that data professionals need to learn modern development technologies and skills. The AD method recognizes that each project team is unique and needs to follow a process tailored for their situation. The importance of looking beyond your current project to address enterprise issues is also stressed, as is the need for enterprise professionals such as operational database administrators and data architects to be flexible enough to work with project teams in an agile manner.
The second step is for data professionals, in particular database administrators, to adopt new techniques that enable them to work in an evolutionary manner. In this chapter, we briefly overview these critical techniques, and in our opinion the most important technique is database refactoring, which is the focus of this book. The evolutionary database development techniques are as follows:
Database refactoring. Evolve an existing database schema a small bit at a time to improve the quality of its design without changing its semantics.
Evolutionary data modeling. Model the data aspects of a system iteratively and incrementally, just like all other aspects of a system, to ensure that the database schema evolves in step with the application code.
Database regression testing. Ensure that the database schema actually works.
Configuration management of database artifacts. Your data models, database tests, test data, and so on are important project artifacts that should be managed just like any other artifact.
Developer sandboxes. Developers need their own working environments in which they can modify the portion of the system that they are building and get it working before they integrate their work with that of their teammates.
Let’s consider each evolutionary database technique in detail.
1.1 Database Refactoring
Refactoring (Fowler 1999) is a disciplined way to make small changes to your source code to improve its design, making it easier to work with. A critical aspect of a refactoring is that it retains the behavioral semantics of your code—you neither add nor remove anything when you refactor; you merely improve its quality. An example refactoring would be to rename the getPersons() operation to getPeople(). To implement this refactoring, you must change the operation definition, and then change every single invocation of this operation throughout your application code. A refactoring is not complete until your code runs again as before.
Similarly, a database refactoring is a simple change to a database schema that improves its design while retaining both its behavioral and informational semantics. You could refactor either structural aspects of your database schema such as table and view definitions or functional aspects such as stored procedures and triggers. When you refactor your database schema, not only must you rework the schema itself, but also the external systems, such as business applications or data extracts, which are coupled to your schema. Database refactorings are clearly more difficult to implement than code refactorings; therefore, you need to be careful. Database refactoring is described in detail in Chapter 2, and the process of performing a database refactoring in Chapter 3.