David Grizzanti Evaluating Software Reuse Alternatives: An Industrial Case Study

Introduction Software reuse results in:  Lower costs  Shorter delivery time. Software reuse factors include the product line approach and the involvement of core assets Product developers decide whether to acquire these “core assets” or develop their own artifacts Decisions should be based on collected data from past reuse scenarios

Reuse takes place when an existing artifact is used in the development of a software product. Two types of reuse: black-box and white-box. Before any reuse can take place, domain analysis must be conducted. Domain Analysis begins with information gathering and results in candidates for reuse. Related Work  Using software reuse alternatives Hewlett-Packard (HP) saw a defect reduction of 15% and a 57% increase in productivity. Reuse Concepts

Cost of reuse software can be divided into 3 general categories: Production Construction: Asset Acquisition Cost Asset Development Cost Product Integration, Verification, and Validation Cost Core-Asset Construction: Asset Acquisition Cost Asset Development Cost Infrastructure Construction: Repository Establishment and Maintenance Cost Repository Storage and Cataloging Cost Domain Analysis (DA) Cost Reuse: Costing Policies

The Model Software construction in the underlying model is described along three axes: Development Maintenance Reuse Artifacts have to be stored and cataloged in the core-assest repository before being transferred between software products.

The Model Mining - Fetching products and copying them into the repository Acquisition - Copying artifacts from the repository into specific products There are two types of Assets:  Repository Assets – cataloged in core-asset repository  Private Assets – contained within a product and are available for mining or acquisition.

Elementary Operations Transformation Operations: Adaptation for Reuse (AR) New for Reuse (NR) White-Box Reuse (WB) New Development (ND) Transition Operations: Cataloged Asset Acquisition (CA) Black-Box Reuse (BB) Mining and Cataloging (MC) Copy and Paste (CP) eXternal Acquisition (XA)

Case Study Example Develop two software products Both need a queue structure One in C++, one in Java Previous product implemented the same structure, except in Ada. Reverse-engineer the Ada package to an object-oriented class design Sequence of operations: MC→ AR → CA → WB A reuse scenario was employed here To determine the best reuse scenario, compare the relative cost of alternative scenarios

Cost Components Operations involve two assets:  Source Asset  Target Asset Cost of Transformation Operations C AR (R,R’)C NR (R’) C WB (P,P’)C ND (P’) Cost of Transition Operations C CA (R,P)C CP (P) C MC (P,R)C XA (R) C BB (R’,P’)

Applying Cost Policies Costing policy table can easily be adapted to accommodate other cost directives. Empty – N/A to operation Full – Full cost to operation Part – Partial cost to operation  Costs usually incurred when activity doesn’t involve resulting artifact

Typical Reuse Scenarios Pure Development (PD)  Target components are developed from scratch Opportunistic Reuse (OR)  Artifacts exist for reuse, but must be searched for Controlled Reuse (CR)  Artifacts are cataloged unchanged then modified for use Systematic Reuse (SR)  Artifacts are mined, cataloged, and modified for future acquisition Alternative Systematic Reuse  Artifacts are created from scratch and then cataloged

Industrial Experience Model was implemented by ISWRIC (Israel SoftWare Reuse Industrial Consortium), a joint project of seven Israeli companies Three are mainly defense contractors The other four produce customizable commercial products Six of the companies defined and evaluated reuse scenarios

Industrial Experience Sytematic Reuse  Implemented by five companies Three implemented systematic reuse of reusable assets Two implemented systematic reuse on existing assets Controlled Reuse  Implemented by one company Reusable assets brought to degree where each product can use these assets for specific needs Use of this model in Industry has demonstrated its potential benefits to various reuse scenarios

Industrial Experience - TES Case Study Tadiran Electronic Systems Ltd., adopted the controlled reuse scenario Established a reuse repository Allocated resources to each “team” to perform Adaptation for Reuse Products were identified as candidates for reuse through Domain Analysis Four of the assets were software modules with sophisticated hardware elements The other three were generic human-computer interface modules

TES Case Study Results Cost, in person-hours (bold numbers were actually measured whereas others were estimated) Except for Asset 5, the implemented scenario achieved the least cover over all alternatives Asset 5, however shows that systematic reuse would have faired better

Alternative Reuse Savings Between Systematic and Controlled Reuse, largest saving was 63% (Asset 7), and lowest savings was -28% (Asset 5) Savings relative to Opportunistic Reuse were between 1-65% and in comparison to Pure Development, it was between 42-81%

Conclusions and Further Work Results based on estimated data are subject to inaccuracies If applied consistently, cumulated data can be analyzed for better estimates Two major contributions:  Cost model provides a practical tool for developers to test different scenarios  Provides a mechanism to analyze and evaluate scenarios at an organizational level Other features, such as time-to-market and product quality also improve with reusing software assets In the future, working to include models for these attributes

References Tomer, Amir; Goldin, Leah; Kuflik, Tsvi; Kimchi, Esther; Schach, Stephen R. "Evaluating Software Reuse Alternatives: A Model and Its Application to an Industrial Case Study." IEEE Transactions on Software Engineering. Volume: 30. Issue: 9. September pp: P. Clements and L.M. Northrop, Software Product Lines: Practices and Patterns. Addison-Wesley, 2001.