The TRESE workshop on Quality tradeoff analysis in domain model specializations The ASML Case

1
The TRESE workshop on Quality trade-off analysis
in domain model specializationsThe ASML Case

• Software Engineering Group
• University of Twente
• Monday, December 6, 2004

2
Morning program

• Introduction
• 1030 Description of the case Pascal Dürr
• 1100 Advantages of quality-oriented
• domain engineering Gürcan Gülesir
• 1115 The problems to be discussed Christian
  Bakker
• 1145 A canonical model for quality
• verification and techniques for
• trade-off analysis Mehmet Aksit
• 1215 Discussions
• 1245 Lunch

3
Afternoon program

• Techniques in detail
• 1330 Principles of software measurement Klaas
  van den Berg
• 1410 The state of the art techniques in
• requirements engineering and software
• architectures Bedir Tekinerdogan
• 1450 Break
• 1520 Design algebra Ivan Kurtev
• 1600 Optimization of software products
• and processes Joost Noppen

4
Evening program

• Application to the case
• 1640 Comparison of the techniques All
• 1715 Approach to follow All
• 1815 Dinner
• 2000 End of the day

5
Goal of this workshop

• Investigate possible techniques to select the
  optimal specialization of a domain model, given a
  large solution space and set of requirements.
• Coming up with the optimal solution is NOT OUR
  GOAL.

6
Introduction

• Within Ideals project two cases
• Error handling (EH)
• Supervisory machine control
• Current EH mechanism at ASML is sub-optimal and
  cross-cutting.
• For optimal solution we started with a domain
  analysis.
• Resulted in a common vocabulary and a feature
  diagram

7
Feature diagram

• A feature diagram consists of
• A set of features
• Optional
• Required
• OR choices
• XOR choices
• Relations between features
• Dependencies
• Excluding features

8
Requirements for EH

• Functional requirements derived from
  documentation and indicated by ASML.
• Problems found outside the current synchronous
  calling tree should also be reported.
• Related errors should be presented in a clear
  and convenient manner.

9
Requirements for EH

• Non-functional requirements (qualities) indicated
  by ASML.
• Distributed processing
• Totality
• Maintainability
• Readability
• Reusability
• Prioritized by ASML.

10
Feature selection

• Usually one selects the optimal solution based
  on the set of requirements.
• Works fine for the functional requirements
• The qualities pose the problem
• Solution space is large 50 options to consider
• Number of possible solutions is huge
• Qualities to be considered 20
• Results in 1000 quality decision points
• Requires extensive domain knowledge
• Too many points to do manually

11
Domain Engineering

• Collecting, storing and organizing past
  experience in building (parts of) systems from
  reusable assets (Czarnecki et.al)
• Providing adequate means for reusing these assets
  when building new systems (Czarnecki et.al)

12
Domain Engineering
13
Domain Analysis in EH case

• Building different EH systems is not the focus.
• Domain analysis provides solution space
• What is possible
• Spanning solution space properly is important for
  providing every possible alternative

14
What is wrong with current EH at ASML?

• Return code based EH
• int result OK
• result F()
• if(result OK)
• else //recovery, logging etc.

15
What is wrong with current EH at ASML?

• If 100 properly applied, no functional problem.
• But
• Unreadable (Readability)
• Error prone (Reliability, Robustness)
• Less maintainable code (Maintainability)

16
Lack of Qualities

• Ignoring qualities will lead to suboptimal
  solution
• The solution will be discarded
• New solution will be needed
• Conclusion A solution based on quality factors
  is essential