Options Analysis for Reengineering (OAR): A Method for Mining Legacy Assets

Given the set of analyzed candidate components, the team next develops alternatives to mining based on schedule, cost, effort, risk, and resource considerations. The OAR team also screens candidate components one more time and analyzes the impact of different component aggregations.

The Plan Mining Options activity has seven tasks:

Finally, team members select the best mining option or combination of options by balancing program and technical considerations. After evaluating each mining option, they prepare a summary report that presents and justifies their selections.

The Select Mining Option activity has five tasks:

1. Choose Best Option:

• Determine the drivers for selecting from among the options

• Select an option or combination of them.

2. Verify Option:

• Record all the details about each of the options chosen.

3. Identify Component Needs Satisfied:

• Complete the final list of component needs satisfied and not satisfied through the
options selected.

4. Present Findings:

• Prepare findings presentation that provides details about the options selected.

5. Produce Summary:

• Produce a final report detailing the options chosen and the rationales for their
selection.

Each activity also has a set of potential specialized tasks to address circumstances that may otherwise preclude completing the activity. These tasks may apply under the following conditions:

• The exit criteria for the prescribed activity have not been satisfied.
• More work is required than can reasonably be accomplished in the activity wrap-up tasks.
• Additional data is needed to complete a particular task or to address special customer needs.
• There is a need to supplement standard OAR tasks to enhance decision-making and reduce risk.

Section 3 includes examples of specialized tasks.

The activities are structured according to an EITVOX (Entry Criteria, Inputs, Tasks, Validation, Outputs, Exit Criteria) format as follows:

At the end of each activity, examples are given of the specialized tasks that may need to be performed before continuing.

The following sub-sections show how the first activity, Establish Mining Context, is structured according to the EITVOX format. Each of the other activities follows this format. Bergey et al., provide this level of detail [Bergey 01].

The purpose of Establish Mining Context is to understand the organization’s product line or new single system needs, legacy base, and expectations for mining legacy components.

The tasks of Establish Mining Context were developed to address a set of underlying questions that include

• What are management’s expectations for the mining effort?
• What are the requirements (e.g., quality attributes) and primary drivers (e.g., priorities and constraints) for mining components?
• What explicit (and implicit) constraints, ground rules, and assumptions apply?
• What specific product line component needs will be addressed?
• What legacy systems are relevant and what documentation is available?
• What documentation are available describing legacy components (e.g., their functionality, granularity, and interfaces)?
• What is the level of preparedness of the organization in terms of experience, skills, available resources, and OAR timeframe?

The complete set of underlying questions is revisited during validation to make sure they were answered satisfactorily.

The following entry criteria are examined to determine if there is sufficient data and expertise available to perform the activity successfully. The entry criteria for this activity are

• The organization has decided to move to a product line and wants to quickly assess the viability of mining components.
• The product line architecture has been defined and component needs have been identified.
• A set of mining goals and objectives has been established.
• One to three legacy systems have been identified as the source of components for mining effort.
• A "Mining Team" has been chartered and is available for the duration of OAR activities.

If the entry criteria are not satisfied, a specialized task may need to be developed.

Each activity relies on a set of support artifacts. The following artifacts are required for Establish Mining Context:

• goals and objectives for the mining effort
• product line requirements and component needs
• documentation describing product line scope, architecture, and component specifications
• overview of legacy system and component documentation
• experience reports of other mining/reengineering efforts
• typical OAR Scheduling Profile¹

The Establish Mining Context activity is divided into 10 tasks. These tasks are executed in order. Several of the tasks are further divided into subtasks. To help OAR team members perform tasks and sub-tasks, the SEI developed execution and data templates. Execution templates outline the steps, rationale, and assumptions for the task or sub-task. Data templates provide example topics and often serve as a starting point for eliciting information from a customer.

The tasks accomplish the following:

After the tasks are completed, the underlying questions for the activity are reviewed. The following are some of the validation criteria for Establish Mining Context:

• Expectations are outlined and understood.
• All relevant legacy systems are identified.
• A set of 10 to 20 product line component needs have been selected with high potential to be satisfied through mining.
• Drivers and priorities are identified and understood.
• A set of 15 to 30 legacy components have been selected as mining candidates.
• The level of mining preparedness has been evaluated.

Establish Mining Context produces the following output artifacts:

• a set of relevant legacy systems and documentation
• a set of selected product line component needs
• major program and technical drivers
• a set of selected legacy components and associated component documentation
• preparedness evaluation
• list of mining issues and concerns
• revised OAR Scheduling Profile

Before completing the Establish Mining Context, the following exit criteria need to be satisfied:

• The mining team has identified a reasonable set of product line needs and candidate components.
• The organization’s expectations are consistent with its level of preparedness.
• The OAR Scheduling Profile has been reviewed and any needed changes have been accepted.
• All output artifacts of this activity have been produced.
• All underlying questions for this activity have been answered satisfactorily.

If the exit criteria are not satisfied, a specialized task may be appropriate. Examples of specialized tasks for the Establish Mining Context activity follow:

• Further identifying mining effort drivers and resolving programmatic and technical issues.
• Further defining product line requirements and component needs in terms of functionality and interfaces.
• Isolating and identifying the functionality and interfaces of legacy components.
• Developing the required level of documentation for legacy components.

4   An Example of an Application of OAR

OAR was used to explore component mining for a Satellite Tracking Agency. This agency supports efforts to develop, acquire, and deploy satellite tracking systems. The agency had an urgent need to mine components from an existing Satellite Tracking System (STS) for the architecture in their Improved Satellite System (ISS). ISS will integrate space assets with a consolidated ground segment. ISS requires some of the capabilities of STS. Because STS already has a set of reliable and validated models, the agency decided to mine and reuse selected components in ISS.

Members of the technical staff at the SEI and agency representatives formed an OAR team to identify relevant components and to estimate the cost and effort required to rehabilitate them for the ISS. The team first established a set of drivers for making decisions on candidate components. These drivers included flexibility of interfaces, satisfaction of real-time constraints, portability and interoperability, and high level of code quality (i.e., high cohesion, low coupling). After examining the STS system, the team selected six legacy system components as mining candidates: two simulation gateways, a message gateway, two truth model components, and a theater event sub-system. The Component Table that appears in Appendix A describes each component.

Three potential wrapping components were identified (the two simulation gateways and the message gateway). These components would not require modifying their internal programs. The other three components (the two truth model components and the theater event sub-system) were identified as potential white box components that would require changes.

Next, the team estimated the basic types of changes required to rehabilitate these components. The OAR team also estimated the effort and cost to develop candidate components from scratch and compared these figures to the cost of mining them. However, since the interfaces for the new system had not yet been fully defined, the cost estimates were preliminary. Once the ISS interface specifications are complete, the cost estimates will need to be reviewed.

Three options for aggregation were developed. The gateway and message components were aggregated as one option. However, since these components all depend on the interfaces that are still undefined, this aggregation may change. The other two components represent coarse-grained groupings of self-contained functionality. They were each treated as separate options. The team then calculated the comparative cost and effort of each aggregation based on the information captured in the Component Table. The team also determined the aggregated level of difficulty and risk for each option. The Options Table in Appendix B presents this data.

5   Conclusion

OAR is a systematic, architecture-centric, decision-making method for mining existing components. It identifies software product line or new system architecture component needs that can be satisfied through mining, as well as those that cannot be satisfied. Candidate components are screened based on criteria that are relevant for the organization. OAR provides management visibility into this highly complex analysis activity. It also provides insights into implicit stakeholder assumptions, constraints, and other major drivers that impact the mining of components. OAR determines the feasibility of mining components at an early stage of a reengineering project. OAR provides insight into technical issues as well as cost and effort estimates and helps to address the problem that many reengineering projects fail due to lack of early insight into the magnitude of an effort.

As a first step OAR establishes the component needs, identifies candidate legacy components and their related documentation, and determines other aspects that define the mining context. Next it inventories the legacy components that can potentially fill identified product line needs, and identifies their characteristics, types of changes needed, and estimated cost and effort. OAR reflects the elicited needs, priorities and concerns of the organization. Its final outputs are a mining option or combination of options that an organization may pursue and a list of product line component needs that can and cannot be satisfied through mining.

OAR has been applied successfully at the Satellite Tracking Agency, and lessons learned from this effort led to the current Version 2. We are planning to apply OAR on other projects and continue to refine and improve the method. We are currently seeking organizations interested in applying the method. In the near future, we will develop a handbook to accompany the method. Eventually, we plan to transition the method to other organizations.