Concept: Agile Practices and RUP
This guideline explains how to apply some of the "best practices" identified in the agile community to RUP-based projects that would like to benefit from some of these practices.
Relationships
Main Description

Topics

Introduction

The Rational Unified Process (RUP) is a process framework that Rational Software has refined over the years which has been widely used for all types of software projects-from small to large. Recently a growing number of "agile" processes-such as eXtreme Programming (XP), SCRUM, Feature-Driven Development (FDD) and the Crystal Clear Methodology-have recently been gaining recognition as effective methods for building smaller systems. (See www.agilealliance.org for further information on the Agile Alliance.)

The following sections are intended to assist those project teams evaluating some of the "agile" practices found in one of these methods to see how they are addressed by the more complete software development process defined by RUP (for more information on RUP, see Introduction to RUP).

Overview

The agile community has synthesized a number of "best practices" that are especially applicable to small, co-located project teams. Although RUP is targeted to project teams of any size, it can be successfully applied to small projects. In general, RUP and the processes of the Agile community have a similar view of the key best practices required to develop quality software-for example, applying iterative development and focusing on the users.

The following sections explain how to apply some of the "best practices" identified in the agile community to RUP-based projects that would like to benefit from some of these practices. In this case, the focus will be specifically on those practices presented by the eXtreme Programming (XP) methodology. (For more information on XP, please refer to the web site: http://www.extremeprogramming.org.)

XP Practices

XP includes four basic "activities" (coding, testing, listening, and designing), which are actually more closely aligned with RUP disciplines. These XP activities are performed using a set of practices that require the performance of additional activities, which map to some of the other disciplines in the RUP. XP's practices, according to Extreme Programming Explained, are:

  • The planning game: Quickly determine the scope of the next release by combining business priorities and technical estimates. As reality overtakes the plan, update the plan.
  • Small releases: Put a simple system into production quickly, then release new versions on a very short cycle.
  • Metaphor: Guide all development with a simple shared story of how the whole system works.
  • Simple design: The system should be designed as simply as possible at any given moment. Extra complexity is removed as soon as it is discovered.
  • Testing: Programmers continually write unit tests, which must run flawlessly for development to continue. Customers write tests demonstrating that features are finished.
  • Refactoring: Programmers restructure the system without changing its behavior to remove duplication, improve communication, simplify, or add flexibility.
  • Pair programming: All production code is written with two programmers at one machine.
  • Collective ownership: Anyone can change any code anywhere in the system at any time.
  • Continuous integration: Integrate and build the system many times a day, every time a task is completed.
  • 40-hour week: Work no more than 40 hours a week as a rule. Never work overtime a second week in a row.
  • On-site customer: Include a real, live user on the team, available full-time to answer questions.
  • Coding standards: Programmers write all code in accordance with rules emphasizing communication through the code.

Activities performed as a result of the "planning game" practice, for example, will mainly map to the RUP's project management discipline. But some RUP topics, such as the deployment of the released software, are outside the scope of XP. Requirements elicitation is largely outside the scope of XP, since the customer defines and provides the requirements. Also, because of simpler development projects it addresses, XP can deal very lightly with the issues the RUP covers in detail in the configuration and change management discipline and the environment discipline.

XP Practices Compatible with RUP

In the disciplines in which XP and the RUP overlap, the following practices described in XP could be-and in some cases already are-employed in the RUP:

  • The planning game: The XP guidance on planning could be used to achieve many of the objectives shown in the Project Management discipline of RUP for a very small project. This is especially useful for low-formality projects that are not required to produce formal intermediate project management artifacts.
  • Test-first design and refactoring: These are good techniques that can be applied in the RUP's implementation discipline. XP's testing practice, which requires test-first design, is in particular an excellent way to clarify requirements at a detailed level. As we'll see in the next section, refactoring may not scale well for larger systems.
  • Continuous integration: The RUP supports this practice through builds at the subsystem and system levels (within an iteration). Unit-tested components are integrated and tested in the emerging system context.
  • On-site customer: Many of the RUP's activities would benefit greatly from having a customer on-site as a team member, which can reduce the number of intermediate deliverables needed-particularly documents. As its preferred medium of customer-developer communication, XP stresses conversation, which relies on continuity and familiarity to succeed; however, when a system-even a small one-has to be transitioned, more than conversation will be needed. XP allows for this as something of an afterthought with, for example, design documents at the end of a project. While it doesn't prohibit producing documents or other artifacts, XP says you should produce only those you really need. The RUP agrees, but it goes on to describe what you might need when continuity and familiarity are not ideal.
  • Coding standards: The RUP has an artifact-programming guidelines-that would almost always be regarded as mandatory. (Most project risk profiles, being a major driver of tailoring, would make it so.)
  • Forty-hour week: As in XP, the RUP suggests that working overtime should not be a chronic condition. XP does not suggest a hard 40-hour limit, recognizing different tolerances for work time. Software engineers are notorious for working long hours without extra reward-just for the satisfaction of seeing something completed-and managers need not necessarily put an arbitrary stop to that. What managers should never do is exploit this practice or impose it. They should always be collecting metrics on hours actually worked, even if uncompensated. If the log of hours worked by anyone seems high over an extended period, this certainly should be investigated; however, these are issues to be resolved in the circumstances in which they arise, between the manager and the individual, recognizing any concerns the rest of the team might have. Forty hours is only a guide-but a strong one.
  • Pair programming: XP claims that pair programming is beneficial to code quality, and that once this skill is acquired it becomes more enjoyable. The RUP doesn't describe the mechanics of code production at such a fine-grained level, although it would certainly be possible to use pair programming in a RUP-based process. Some information on pair programming-as well as test-first design and refactoring-is now provided with the RUP, in the form of white papers. Obviously, it is not a requirement to use any of these practices in the RUP, however in a team environment, with a culture of open communication, we would hazard a guess that the benefits of pair programming (in terms of effect on total lifecycle costs) would be hard to discern. People will come together to discuss and solve problems quite naturally in a team that's working well, without being obliged to do so.

The suggestion that good process has to be enforced at the "micro" level is often unpalatable and may not fit some corporate cultures. Strict enforcement, therefore, is not advocated by RUP. However, in some circumstances, working in pairs-and some of the other team-based practices advocated by XP-is obviously advantageous, as each team member can help the other; for example:

  • the early days of team formation, as people are getting acquainted
  • teams inexperienced in some new technology
  • teams with a mix of experienced staff and novices

XP Practices That Don't Scale Well

The following XP practices don't scale well for larger systems (nor does XP claim they do), so we would make their use subject to this proviso in the RUP.

  • Metaphor: For larger, complex systems, architecture as metaphor is simply not enough. The RUP provides a much richer description framework for architecture that isn't just-as Extreme Programming Explained describes it-"big boxes and connections." Even in the XP community, metaphor has more recently been deprecated. It is no longer one of the practices in XP (until they can figure out how to describe it well-maybe a metaphor would help them).
  • Collective Ownership: It's useful if the members of a team responsible for a small system or a subsystem are familiar with all of its code. But whether you want to have all team members equally empowered to make changes anywhere should depend on the complexity of the code. It will often be faster (and safer) to have a fix made by the individual (or pair) currently working on the relevant code segment. Familiarity with even the best-written code, particularly if it's algorithmically complex, diminishes rapidly over time.
  • Refactoring: In a large system, frequent refactoring is no substitute for a lack of architecture. Extreme Programming Explained says, "XP's design strategy resembles a hill-climbing algorithm. You get a simple design, then you make it a little more complex, then a little simpler, then a little more complex. The problem with hill-climbing algorithms is reaching local optima, where no small change can improve the situation, but a large change could." In the RUP, architecture provides the view and access to the "big hill," to make a large, complex system tractable.
  • Small Releases: The rate at which a customer can accept and deploy new releases will depend on many factors, typically including the size of the system, which is usually correlated with business impact. A two-month cycle may be far too short for some types of system; the logistics of deployment may prohibit it.

XP Practice Requiring Caution

Finally, an XP practice that at first glance sounds potentially usable in the RUP-Simple Design-needs some elaboration and caution when applied generally.

  • Simple Design
    XP is very much functionality driven: user stories are selected, decomposed into tasks, and then implemented. According to Extreme Programming Explained, the right design for the software at any given time is the one that runs all the tests, has no duplicated logic, states every intention important to the programmers, and has the fewest possible classes and methods. XP doesn't believe in adding anything that isn't needed to deliver business value to the customer.

    There's a problem here, akin to the problem of local optimizations, in dealing with what the RUP calls "nonfunctional" requirements. These requirements also deliver business value to the customer, but they're more difficult to express as stories. Some of what XP calls constraints fall into this category. The RUP doesn't advocate designing for more than is required in any kind of speculative way, either, but it does advocate designing with an architectural model in mind-that model being one of the keys to meeting nonfunctional requirements.

    So, the RUP agrees with XP that the "simple design" should include running all the tests, but with the rider that this includes tests that demonstrate that the software will meet the nonfunctional requirements. Again, this only looms as a major issue as system size and complexity increase, or when the architecture is unprecedented or the nonfunctional requirements onerous. For example, the need for marshalling data (to operate in a heterogeneous distributed environment) seems to make code overly complex, but it will still be required throughout the program.

Mapping of Artifacts for a Small Project

When we tailor the RUP for a small project and reduce the Work Product requirements accordingly, how does this compare to the equivalent of artifacts in an XP project? Table 1 shows a typical set of RUP artifacts on a small RUP project.

XP Artifacts
RUP Artifacts
(Example for Small Projects)
Stories
Additional documentation from conversations
Vision
Glossary
Use-Case Model
Constraints Supplementary Specifications
Acceptance tests and unit tests
Test data and test results

Test PlanTest CaseTest Suite (including Test Script, Test Data)
Test Log
Test Evaluation Summary

Software (code) Implementation Model
Releases Product (Deployment Unit)
Release Notes
Metaphor Software Architecture Document
Design (CRC, UML sketch)
Technical tasks and other tasks
Design documents produced at end
Supporting documentation
Design Model
Coding standards Project-Specific Guidelines
Workspace
Testing framework and tools
Development Case
Test Environment Configuration
Release plan
Iteration plan
Story estimates and task estimates
Software Development Plan
Iteration Plan
Overall plan and budget Business Case
Risk List
Reports on progress
Time records for task work
Metrics data (including resources, scope, quality, time)
Results tracking
Reports and notes on meetings
Status Assessment
Iteration Assessment
Review Record
Defects (and associated data) Change Requests
Code management tools Project Repository
Workspace
Spike (solution)

Prototypes
User Interface Prototype
Architectural Proof of Concept

XP itself (it's recommendations and guidance)

Test Ideas List
Project-Specific Guidelines

[Not included in XP]

Data Model
User Support Material.

Table 1: XP-to-RUP mapping of artifacts for a small project

Although the granularity of the artifacts varies on both sides, in general the artifacts in the RUP for small projects (the type XP would comfortably address) map quite well to those of an XP project.

Note that the table also includes a few artifacts which are not covered by XP, but are needed on many projects. These include Data Model, and artifacts related to deployment, such as User Support Material.

Activities

The RUP defines an Task as work performed by a Role-either using and transforming input artifacts or producing new and changed output artifacts. RUP goes on to enumerate these activities and categorize them according to the RUP disciplines. These disciplines include: requirements, analysis and design, deployment, and project management (among others).

Activities are time-related through the artifacts they produce and consume: an activity can logically begin when its inputs are available (and in an appropriately mature state). This means that producer-consumer activity pairs can overlap in time, if the artifact state permits; they need not be rigidly sequenced. Activities are intended to give strong guidance on how an artifact should be produced, and they may also be used to help the project manager with planning.

Woven through the RUP as it's described in terms of lifecycle, artifacts, and activities are "best practices": software engineering principles proven to yield quality software built to predictable schedule and budget. The RUP, through its activities (and their associated artifacts) supports and realizes these best practices - they are themes running through the RUP - which uphold the key principles that underpin RUP. Note that XP uses the notion of "practices" as well, but as we shall see, there is not an exact alignment with RUP's concept of best practice.

XP presents an engagingly simple view of software development as having four basic activities-coding, testing, listening, and designing-which are to be enabled and structured according to some supporting practices (as discussed in Extreme Programming Explained, Chapter 9). Actually, as noted earlier, XP's activities are closer in scope to the RUP's disciplines than to the RUP's activities, and much of what happens on an XP project (in addition to its four basic activities) will come from the elaboration and application of its practices.

So, there is an XP equivalent of the RUP's activities, but XP's "activities" aren't formally identified or described as such. For example, looking at Chapter 4, "User Stories," in Extreme Programming Installed, you'll find the heading, "Define requirements with stories, written on cards," and throughout the chapter there's a mixture of process description and guidance on what user stories are, and how (and by whom) they should be produced. And it goes on that way; in the various sections of the XP books (under headings that are a mixture of artifact-focused and activity-focused), both "things produced" and "things done" are described, to varying degrees of prescription and detail.

RUP's apparently high degree of prescription results from its completeness and greater formality in its treatment of activities and their inputs and outputs. XP does not lack prescription but, perhaps in its attempt to remain lightweight, the formality and detail are simply omitted. Lack of specificity is neither a strength nor a weakness, but the lack of detailed information in XP should not be confused with simplicity. Not having details may be fine for more experienced developers, but in many cases, more details are a great help for new team members, and team members that are still getting up to speed with the team's approach to software development.

With Activities, just as with Artifacts, it is important to keep focus on what we are trying to achieve. Carrying out an activity blindly is never a good practice. Activities and associated guidelines are there to look at when you need them to achieve your objectives, but should not be used as an excuse for not having to figure out what you are trying to achieve. This spirit is well articulated in XP, and we believe it should be applied by every user of RUP

Roles

In the RUP, activities are said to be performed by roles (or, more precisely, by individuals or groups playing roles). Roles also have responsibility for particular artifacts; the responsible role will usually create the artifact and ensure that any changes made by other roles (if allowed at all) don't break the artifact. An individual or group of people may perform just one role or several roles. A role doesn't have to be mapped to a only a single position or "slot" in an organization.

Extreme Programming Explained identifies seven roles applicable to XP-Programmer, Customer, Tester, Tracker, Coach, Consultant, and Big Boss-and describes their responsibilities and the competencies required of the people who will perform them. References are made to these roles in some of the other XP books as well. The difference in the number of roles in XP and the RUP is easy to explain:

  • XP doesn't cover all of the RUP disciplines.
  • XP roles are more comparable to positions within an organization (possibly with multiple responsibilities) than to RUP roles. For example, XP's Programmer actually performs multiple RUP roles-Implementer, Code Reviewer, and Integrator-which require slightly different competencies.

XP and RUP Roles on a Small Project

When RUP roles are mapped to a small project, the number of XP-like roles that they correspond to is reduced considerably in that the number of positions, or job titles, is 5. Table 3 (drawn from the RUP) shows this mapping with the corresponding XP Role.

XP Role Example RUP Small Project Team Member RUP Role
Coach
Consultant
Big Boss
Sally Slalom, Senior Manager Project Manager
Deployment Manager
Technical Reviewer
Configuration Manager
Change Control Manager
Customer Stakeholder (as documented in the Vision)
Management Reviewer
Technical Reviewer (requirements)
Customer
Big Boss
Tracker
Tom Telemark, Senior Software Engineer System Analyst
Requirements Specifier
User Interface Designer
Software Architect
Technical Reviewer
Test Manager
Test Analyst

Developer roles (with less emphasis)

Programmer
Tester

Susan Snow, Software Engineer

Henry Halfpipe, Junior Software Engineer

Designer
Implementer
Technical Reviewer
Integrator
Test Designer
Tester
Technical Writer
Tracker Patrick Powder, Administrative Assistant Responsible for maintaining the Project web site, assisting the Project Manager role in planning/scheduling activities, and assisting the Change Control Manager role in controlling changes to artifacts.  May also provide assistance to other roles as necessary.

Table 3: Mapping XP roles to RUP roles on a small project

Using XP Practices with RUP

The RUP is a process framework from which particular processes can be configured and then instantiated. The RUP must be configured-this is a required step defined in the RUP itself. Strictly speaking then, we should compare a tailored version of the RUP with XP-that is, with the RUP tailored to the project characteristics that XP explicitly establishes (and those that can be inferred). Such a tailored RUP process could accommodate many of XP's practices (such as pair programming, test-first design and refactoring), but it still wouldn't be identical to XP because of RUP's emphasis on the importance of architecture, abstraction (in modeling), and risk, and its different structure in time (phases and iterations).

XP is intentionally directed at implementing a lightweight process for small projects. In doing so, it also includes descriptions (at least in the books) that are not fully elaborated. In an XP implementation there will always be things that will need to be discovered, invented, or defined on the fly. The RUP will accommodate projects that both fit and are beyond the scope of XP in scale and kind. As this roadmap shows, RUP is actually quite compatible with most of the practices described in the XP literature.

Keep in mind that essence of XP is its focus on organization, people, and culture. This is important in all projects and is certainly applicable to those projects using RUP. Small projects could benefit greatly by using these practices together.

Agile Process References

  • eXtreme Programming (XP) (See http://www.extremeprogramming.org/more.html for more information.):
    • Extreme Programming Explained: Embrace Change. Kent Beck explains the concepts and philosophy behind extreme programming. This book teaches what and why but not how.
    • Refactoring Improving the Design of Existing Code. Martin Fowler writes the first authoritative volume on refactoring. Presented as patterns. There are plenty of examples in Java. This book teaches you how to refactor and why.
    • Extreme Programming Installed. By Ron Jeffries, Chet Hendrickson, and Ann Anderson. This book covers specific XP practices in finer detail than Extreme Programming Explained. This book teaches how to program XP style.
    • Planning Extreme Programming. By Kent Beck, and Martin Fowler. This book presents the latest thoughts on how to plan software in a rapid delivery environment. This book teaches how to run an XP project.
    • Extreme Programming Examined. By Giancarlo Succi and Michele Marchesi. Papers presented at XP2000. A well rounded set of papers covers most topics.
    • Extreme Programming in Practice. By Robert C. Martin, James W. Newkirk. A real project which used XP is described in gory detail.
    • Extreme Programming Explored. By William C. Wake. Based on the popular XPlorations web site. Specific subjects are explored in detail.
    • Extreme Programming Applied: Playing to Win. By Ken Auer and Roy Miller. Experiences from pioneers in applying XP. To be published in September.

For information on other members of the Agile Alliance see http://www.agilealliance.org.