Evaluating the performance of model transformation styles with Maude @ Symposium FACS 2011
0 likes•537 views
The document evaluates the performance of model transformation techniques using case studies with a focus on two prominent programming styles, analyzing metrics like performance, easiness, and understandability. Experiments conducted show varying results based on different settings and hypotheses related to transformation methods. The conclusion highlights that performance can vary dramatically depending on the chosen representation, and touches upon potential future explorations into different styles and strategies.
Evaluating the performance of model transformation styles with Maude @ Symposium FACS 2011
- 1. Evaluating the performance of model transformation styles with Maude Roberto Bruni Alberto Lluch Lafuente Uni Pisa IMT Lucca Acks to Artur Boronat (Uni Leicester) International Symposium on Formal Aspects of Component Software September 14-16, 2011, Oslo
- 2. ·1· Background and Motivation ·2· Case studies ·3· Some implementation notes ·4· Experiments ·5· Conclusion
- 3. ·1· Background and Motivation ·2· Case studies ·3· Some implementation notes ·4· Experiments ·5· Conclusion
- 4. ·1· Background and Motivation ·2· Case studies ·3· Some implementation notes ·4· Experiments ·5· Conclusion
- 5. ·1· Background and Motivation ·2· Case studies ·3· Some implementation notes ·4· Experiments ·5· Conclusion
- 6. ·1· Background and Motivation ·2· Case studies ·3· Some implementation notes ·4· Experiments ·5· Conclusion
- 7. ·1· Background and Motivation ·2· Case studies ·3· Some implementation notes ·4· Experiments ·5· Conclusion
- 13. Server status: busy Client status: connected
- 14. Server Server status: busy status: free! Client Client status: connected status: disconnected!
- 16. Server status: free Client status: disconnected
- 17. Server Server status: free status: Y Client Client status: disconnected status: connected!
- 18. Server Server status: free status: Y Client Client status: disconnected status: connected! if Server Server status: free status: Y
- 19. Server Server status: free status: Y Client Client status: disconnected status: connected! if Server Server status: free status: Y and Client Client status: disconnected status: connected!
- 21. Locality Server Locality Server Client Client Client
- 22. Locality Server Locality Server Client Client Client
- 23. Locality Server Locality Server Client Client Client
- 24. GT vs SOS?
- 25. GT vs SOS? ● Easiness;
- 26. GT vs SOS? ● Easiness; ● Understandability;
- 27. GT vs SOS? ● Easiness; ● Understandability; ● Verification;
- 28. GT vs SOS? ● Easiness; ● Understandability; ● Verification; ● …
- 29. GT vs SOS? ● Easiness; ● Understandability; ● Verification; ● … ● Performance (this paper!)
- 30. ·1· Background and Motivation ·2· Case studies ·3· Some implementation notes ·4· Experiments ·5· Conclusion
- 31. case study #1: component migration
- 32. case study #1: component migration
- 33. case study #1: component migration
- 34. case study #1: component migration
- 35. case study #1: component migration
- 36. case study #1: component migration
- 37. case study #1: component migration
- 38. case study #1: component migration
- 39. case study #1: component migration
- 40. case study #1: component migration
- 41. case study #1: component migration
- 42. case study #2: class diagrams to relational schemas
- 43. case study #2: class diagrams to relational schemas
- 44. case study #2: class diagrams to relational schemas
- 45. case study #2: class diagrams to relational schemas
- 46. case study #2: class diagrams to relational schemas
- 47. case study #2: class diagrams to relational schemas
- 48. case study #2: class diagrams to relational schemas
- 49. case study #3: pullup refactoring
- 50. case study #3: pullup refactoring
- 51. case study #3: pullup refactoring
- 52. case study #3: pullup refactoring
- 53. case study #3: pullup refactoring
- 54. ·1· Background and Motivation ·2· Case studies ·3· Some implementation notes ·4· Experiments ·5· Conclusion
- 56. a a P → P' Q → Q' τ P|Q → P'|Q'
- 57. A rule like a t → t'
- 58. A rule like a t → t' Following Verdejo is implemented as t => {a}t'
- 59. A rule like a t → t' Following Verdejo is implemented as t => {a}t' Our case studies allow this “top-down” variant {a}t => t'
- 60. ·1· Background and Motivation ·2· Case studies ·3· Some implementation notes ·4· Experiments ·5· Conclusion
- 61. experiment #1
- 63. experiment #1/results Yes! :) No! :( Well... :|
- 64. experiment #1/results Yes! :) No! :( Well... :|
- 65. experiment #1/results Yes! :) No! :( Well... :|
- 66. experiment #2
- 67. experiment #2/hypothesis “SOS degenerates when the number of possible (unnecesary) transformations increases ”
- 68. experiment #2/setting Component Component status: normal status: status1
- 69. experiment #2/setting Component Component status: normal status: status1 Component Component status: normal status: status2
- 70. experiment #2/setting Component Component status: normal status: status1 Component Component status: normal status: status2 ... Component Component status: normal status: statusN
- 72. experiment #3
- 73. experiment #3/hypothesis “the performance of SOS* and GT varies a lot depending on the setting”
- 76. experiment #4
- 77. experiment #4/hypothesis “SOS suffers in presence of cross-layering edges”
- 83. experiment #4/results (before)
- 84. ·1· Background and Motivation ·2· Case studies ·3· Some implementation notes ·4· Experiments ·5· Conclusion
- 85. Summary ● Many ways to program MDT;
- 86. Summary ● Many ways to program MDT; ● We have considered two prominent ones;
- 87. Summary ● Many ways to program MDT; ● We have considered two prominent ones; ● Performance varies drastically;
- 88. Summary ● Many ways to program MDT; ● We have considered two prominent ones; ● Performance varies drastically; ● Each exploits a different representation...
- 89. Summary ● Many ways to program MDT; ● We have considered two prominent ones; ● Performance varies drastically; ● Each exploits a different representation... ● but we can pass from one to the other.
- 90. experiment #5
- 91. experiment #5/hypothesis “flattening is more efficient than nesting”
- 93. Outlook ● Other styles, implementations, examples;
- 94. Outlook ● Other styles, implementations, examples; ● Export to other settings;
- 95. Outlook ● Other styles, implementations, examples; ● Export to other settings; ● (Heuristic) strategies;
- 96. Outlook ● Other styles, implementations, examples; ● Export to other settings; ● (Heuristic) strategies; ● Parallellism;
- 97. Outlook ● Other styles, implementations, examples; ● Export to other settings; ● (Heuristic) strategies; ● Parallellism; ● Non-determinism; ● ...
- 98. thanks a lot!