In his CM: the Next Generation series, Joe Farah gives us a glimpse into the trends that CM experts will need to tackle and master based upon industry trends and future technology challenges.
Project Management deals with execution. A project is a series of tasks aimed at accomplishing a specific goal or set of objectives. In software development, a project typically transforms one release of a software system into another release. This is done through a series of operations as outlined below:
- Tranforms release requirements into working product
- Transformation done through work breakdown structure (WBS) detailing tasks
- Tasks are defined, prioritized, and assigned
- Efforts are estimated and rolled up to give both a schedule and a plan of completion
- Critical path analysis identifies potential risk areas
- Work effort is tracked for each task, typically using timesheets
- Timesheet effort is rolled up to identify actual effort
- Revised efforts are used for risk identification and schedule adjustment
- Complex tasks are reviewed for risk and monitored
- Resources are managed to ensure timely completion of tasks and adequate risk support.
In an agile project, scheduling is less stringent with prioritization typically sufficient to ensure that each iteration addresses the appropriate content. Risk is dealt with through daily feedback and resources/content moved if necessary. Tasks are generally kept to a reasonably small size so that the tasks can be completed in a single iteration when possible. The result is that we might see a finer level of decomposition in the WBS, and a corresponding improved level of traceability.
Configuration management deals with tracking the artifacts produced by the process and ensuring that a consistent set of baselines march forward toward project completion. The focus is then on auditing the output of the process to ensure that is satisfies the input. Configuration management works hand-in-hand with change management as follows:
- Requirements are translated from customer/market to product requirements (i.e., functional spec), and then to system design requirements (possibly additional levels)
- Design requirements are addressed by changes/updates which modify files
- File revisions are traced back (through the changes) to the requirements
- Changes are implemented by the development team and unit tested prior to check-in
- Changes are reviewed and approved for integration
- Change promotion is used to create new baselines
- Builds are created based on a baseline plus additional changes to be applied against the baseline.
- Builds are tracked to clearly identify the exact software/system content of the deliverables
- CM tools trace contents of builds back to the requirements addressed by each build so that they may be audited
- Verification is performed using the appropriate set of test cases that apply to the specified build. This typically verifies the functional spec.
This is not an exhaustive description of CM, but just a rapid end-to-end overview of the process.
If you look closely, you will see a good level of overlap between project management and configuration management. Each has to transform the initial requirements into more detailed design goals. Traceability has to be maintained, as both converge on the developer’s desktop.
Typically in a PM tool, tasks are identified, scheduled and tracked. In a CM tool, changes are created, requirements/documents/source code are modified and then checked in. The PM tool gives a task management/progress view, while the CM view identifies modified revisions. Trying to relate the two can be a lot of work.
However, consider an approach where the WBS doesn't simply contain tasks, but also identifies task containers with the contents attached. A design task might have a design document attached to it and decompose into 3 changes each with it's own set of files attached. It might also have a white box test plan task attached for testing the design. Higher up the tree, the design task may have as its parent a feature specification which is a portion of the functional spec for the project that is then attached to the FS task which created the FS document in the first place. Along side the design task might be a couple of other tasks under the FS task: a test plan for the feature, and a user guide snippet for the feature (assuming this is different than the functional spec).
Now include a CM tool that works on this WBS. Instead of the WBS being a planning and tracking tool, it is now a living organism. Traverse the WBS and it then becomes the total content of the project: tasks, feature specifications, changes, build records, and so forth. It also includes the changes made to the requirements, documents, source code, and test cases.
A project is typically used to transform a product from one (possibly empty) release to another. So: Product.rel1 + Project.rel2 => Product.rel2. The idea is that each WBS can capture everything, from changes in requirements through to source code and test case changes. On the CM side, intermediate steps collect changes to create baselines along the way, whether for requirements, documentation, source code or test cases. Build records are used to record intermediate builds in terms of a baseline plus a set of changes, as the WBS continues to grow, until finally the project is done.
In this integrated view, not only do we see full traceability enhanced through the WBS, but also progress measured by the actual completion of tasks though the check in of changes for requirements, documents, software and test cases. Furthermore, these changes are promoted as they move through the verification tasks.
One of the major benefits here is that the visibility of the WBS across the team allows the visibility of the entire set of project data. As the WBS grows, clarity is provided for each of the team members. Further, as the WBS grows, a visible measure of progress is provided.
Mitel: The First Test Bed
The first time that I implemented this architecture for an integrated CM/PM tool and process was in the mid-1980s at Mitel (a PBX manufacturer in Ottawa, ON). Even though we were, at the time, restricted to command-line user interfaces (which didn't appear widely until the '90s), the success of the approach was widely evident to all, including upper management. Product managers and VPs would regularly access the system to identify an up-to-date state of each project. Risk management was simplified as all of the data was integrated into the same repository and accessed through the same tool.
The entire team used the same tool and even the marketing people wanted to go into the CM database both to provide product input and also to identify the status and contents of each release. A quick by-product was to generate and circulate regular documents so that the less technical minded did not have to learn a command line interface to the tool. Initially deployed on the SX-2000 project, the approach continued through at least 2 decades.
We made an effort to continue to use integrated CM and PM ever since at Neuma (using CM+). We've reaped the benefits of rapid development and better communication. The result is a more agile, higher quality development environment.
The bottom line: don't treat project management as an add-on function to your development and don't treat CM as an add-on function to your development. Integrate them and reap the benefits.