Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
SlideShare a Scribd company logo

Towards Scalable Validation of Low-Code System Models: Mapping EVL to VIATRA Patterns

•
•

Presented at the LowCode Workshop 2021 at MODELS 2021 by Benedek Horváth. Authors are Qurat ul ain Ali, Benedek Horváth, Dimitris Kolovos, Konstantinos Barmpis and Ákos Horváth.

1 of 16
Download to read offline
Towards Scalable Validation of Low-Code System Models: Mapping EVL to VIATRA Patterns Qurat ul ain Ali1, Benedek Horváth2,3, Dimitris Kolovos1, Konstantinos Barmpis1, Ákos Horváth2 1 University of York, York, UK 2 IncQuery Labs cPlc., Budapest, Hungary 3 Johannes Kepler University Linz, Linz, Austria Contact: Benedek.Horvath@incquerylabs.com
Validation of Low-Code System Models • Low-code platforms adopt Model-Driven Engineering concepts • Large software system models • Model correctness is essential • Scalable cloud resources → $$$ 2 SOTA model validation engines should adapt to these needs
Cyber-Physical Systems Simulation • Integrated CPS simulation for industry 4.0 • Queries on continuously changing runtime models 3 SOTA model query engines should adapt to these needs [1] Márton Búr, Gábor Szilágyi, András Vörös, Dániel Varró: Distributed graph queries over models@run.time for runtime monitoring of cyber-physical systems. Int. J. Softw. Tools Technol. Transf. 22(1): 79-102 (2020) Figure source: [1]
Epsilon 4 • Epsilon – a family of languages for MDE: • EOL: model queries • EVL: model validation • Model connectivity (EMC): to access the models Epsilon live: https://www.eclipse.org/epsilon/live/
Viatra 5 •Declarative graph query language •Transitive closure, negation cond., etc. •Compositional, reusable •Local + Global queries Definition •Incremental evaluation •Cache result set (RETE networks) •Maintain incrementally upon model change •Local search-based evaluation Execution •On-the-fly validation, •Works out-of-the-box with EMF applications, •View generation, •Derived features. Tooling •AUTOSAR case study •3 tools compared (Java, MDT-OCL, IncQuery) •Scalability measurements •Load + query •Manipulate + re-check •Up to 1.5 million elements Experiments https://www.eclipse.org/viatra/
Validation Rule Execution in Sequential EVL • For each Variable Declaration • Check if there is a Single Variable Access • Whose variable reference refers to it • O (|Variable Declaration| x |Single Variable Access| ) • Can be expensive if: • Single Variable Access is not cached • Model changes and rules should be reevaluated 6 Viatra with incremental query engine
Overview of the Approach EVL validation program Translate constraints to VQL patterns Evaluate the constraints by calling Viatra Show results in Epsilon Phase I Phase II Phase III 7
Running Example 8 Java metamodel (MoDisco) excerpt EVL constraint VQL pattern
Proposed Workflow 9 Phase I Phase II Phase III
Query Translations EOL expression VQL pattern Matcher API call Java!NamedElement.all.selectOne(u|u.name="main") Java!NamedElement.all.exists(u|u.name="main") Java!NamedElement.all.one(u|u.name="main") Java!NamedElement.all.none(u|u.name="main") Java!NamedElement.all.count(u|u.name="main") Java!NamedElement.all.nMatch(u|u.name="main",2) Java!NamedElement.all.atLeastNMatch(u|u.name="main",1) Java!NamedElement.all.atMostNMatch(u|u.name="main",1) pattern Java1(ne: NamedElement) { NamedElement.name(ne, "main"); } oneArbitraryMatch hasMatch countMatches == 1 countMatches == 0 countMatches countMatches == 2 countMatches >= 1 countMatches <= 1 Java!NamedElement.all.reject(u|u.name="main") Java!NamedElement.all.forAll(u|u.name="main") pattern Java1(ne: NamedElement) { neg find Java1internal(ne); } pattern Java1internal(ne: NamedElement) { NamedElement.name(ne, "main"); } allMatches hasMatch == false 10
Benchmarks • MoDisco EMF model of the Eclipse source code [4] • FindBugs validation rules in EVL • Compared performance of: • Incremental Viatra engine (RETE) • Local search based Viatra engine (LS) • Sequential EVL engine • Measured execution time, memory use of the engines 11
Evaluation • Shortest execution time: RETE • Longest execution time: Sequential EVL • Local search engine with base index ~ RETE • Highest memory use: RETE • Lowest memory use: Sequential EVL • Local search engine with base index ~ RETE 12 94 92 88 49 141 133 126 80 284 268 243 183 0 50 100 150 200 250 300 RETE LS with base index LS without base index Sequential EVL Memory use (MB) Engine name Memory use of the engines 100k 200k 500k 0.94 1.35 205.72 205.99 1.77 2.49 883.91 955.62 4.32 6.06 93.19 6390.36 0.10 1.00 10.00 100.00 1000.00 10000.00 RETE LS with base index LS without base index Sequential EVL Execution time (s) Engine name Execution time of the engines 100k 200k 500k
Lessons Learnt • Performance of sequential EVL engine can be improved by Viatra • Incremental engine speeded-up the execution by 1481x • Duration of the query translation is negligible • https://github.com/lowcomote/evl-viatra-prototype Limitations: • First-order operations on all instances of model elements • Single-value references, attributes • Integrated to EOL, EVL 13
Future Work • Integrate to Epsilon Transformation and Pattern Languages • Extend supported elements: • Multi-value references • User-defined operations • Performance comparisons with the parallel EVL engine [5] 14
Acknowledgements • This work is partially funded by the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 813884, • and the National Research, Development and Innovation Fund of Hungary, financed under the 2019-2.1.1-EUREKA-2019-00001 funding scheme. • The authors are grateful for the valuable feedback of Géza Kulcsár and the anonymous reviewers about the paper. 15
References [1] Márton Búr, Gábor Szilágyi, András Vörös, Dániel Varró: Distributed graph queries over models@run.time for runtime monitoring of cyber-physical systems. Int. J. Softw. Tools Technol. Transf. 22(1): 79-102 (2020) [2] Dániel Varró, Gábor Bergmann, Ábel Hegedüs, Ákos Horváth, István Ráth, Zoltán Ujhelyi: Road to a reactive and incremental model transformation platform: three generations of the VIATRA framework. Softw. Syst. Model. 15(3): 609-629 (2016) [3] Charles Forgy: Rete: A Fast Algorithm for the Many Patterns/Many Objects Match Problem. Artif. Intell. 19(1): 17-37 (1982) [4] http://atenea.lcc.uma.es/projects/LinTra.html [5] Sina Madani, Dimitrios S. Kolovos, Richard F. Paige: Parallel Model Validation with Epsilon. ECMFA 2018: 115-131 16

More Related Content

Towards Scalable Validation of Low-Code System Models: Mapping EVL to VIATRA Patterns

  • 1. Towards Scalable Validation of Low-Code System Models: Mapping EVL to VIATRA Patterns Qurat ul ain Ali1, Benedek Horváth2,3, Dimitris Kolovos1, Konstantinos Barmpis1, Ákos Horváth2 1 University of York, York, UK 2 IncQuery Labs cPlc., Budapest, Hungary 3 Johannes Kepler University Linz, Linz, Austria Contact: Benedek.Horvath@incquerylabs.com
  • 2. Validation of Low-Code System Models • Low-code platforms adopt Model-Driven Engineering concepts • Large software system models • Model correctness is essential • Scalable cloud resources → $$$ 2 SOTA model validation engines should adapt to these needs
  • 3. Cyber-Physical Systems Simulation • Integrated CPS simulation for industry 4.0 • Queries on continuously changing runtime models 3 SOTA model query engines should adapt to these needs [1] Márton Búr, Gábor Szilágyi, András Vörös, Dániel Varró: Distributed graph queries over models@run.time for runtime monitoring of cyber-physical systems. Int. J. Softw. Tools Technol. Transf. 22(1): 79-102 (2020) Figure source: [1]
  • 4. Epsilon 4 • Epsilon – a family of languages for MDE: • EOL: model queries • EVL: model validation • Model connectivity (EMC): to access the models Epsilon live: https://www.eclipse.org/epsilon/live/
  • 5. Viatra 5 •Declarative graph query language •Transitive closure, negation cond., etc. •Compositional, reusable •Local + Global queries Definition •Incremental evaluation •Cache result set (RETE networks) •Maintain incrementally upon model change •Local search-based evaluation Execution •On-the-fly validation, •Works out-of-the-box with EMF applications, •View generation, •Derived features. Tooling •AUTOSAR case study •3 tools compared (Java, MDT-OCL, IncQuery) •Scalability measurements •Load + query •Manipulate + re-check •Up to 1.5 million elements Experiments https://www.eclipse.org/viatra/
  • 6. Validation Rule Execution in Sequential EVL • For each Variable Declaration • Check if there is a Single Variable Access • Whose variable reference refers to it • O (|Variable Declaration| x |Single Variable Access| ) • Can be expensive if: • Single Variable Access is not cached • Model changes and rules should be reevaluated 6 Viatra with incremental query engine
  • 7. Overview of the Approach EVL validation program Translate constraints to VQL patterns Evaluate the constraints by calling Viatra Show results in Epsilon Phase I Phase II Phase III 7
  • 8. Running Example 8 Java metamodel (MoDisco) excerpt EVL constraint VQL pattern
  • 9. Proposed Workflow 9 Phase I Phase II Phase III
  • 10. Query Translations EOL expression VQL pattern Matcher API call Java!NamedElement.all.selectOne(u|u.name="main") Java!NamedElement.all.exists(u|u.name="main") Java!NamedElement.all.one(u|u.name="main") Java!NamedElement.all.none(u|u.name="main") Java!NamedElement.all.count(u|u.name="main") Java!NamedElement.all.nMatch(u|u.name="main",2) Java!NamedElement.all.atLeastNMatch(u|u.name="main",1) Java!NamedElement.all.atMostNMatch(u|u.name="main",1) pattern Java1(ne: NamedElement) { NamedElement.name(ne, "main"); } oneArbitraryMatch hasMatch countMatches == 1 countMatches == 0 countMatches countMatches == 2 countMatches >= 1 countMatches <= 1 Java!NamedElement.all.reject(u|u.name="main") Java!NamedElement.all.forAll(u|u.name="main") pattern Java1(ne: NamedElement) { neg find Java1internal(ne); } pattern Java1internal(ne: NamedElement) { NamedElement.name(ne, "main"); } allMatches hasMatch == false 10
  • 11. Benchmarks • MoDisco EMF model of the Eclipse source code [4] • FindBugs validation rules in EVL • Compared performance of: • Incremental Viatra engine (RETE) • Local search based Viatra engine (LS) • Sequential EVL engine • Measured execution time, memory use of the engines 11
  • 12. Evaluation • Shortest execution time: RETE • Longest execution time: Sequential EVL • Local search engine with base index ~ RETE • Highest memory use: RETE • Lowest memory use: Sequential EVL • Local search engine with base index ~ RETE 12 94 92 88 49 141 133 126 80 284 268 243 183 0 50 100 150 200 250 300 RETE LS with base index LS without base index Sequential EVL Memory use (MB) Engine name Memory use of the engines 100k 200k 500k 0.94 1.35 205.72 205.99 1.77 2.49 883.91 955.62 4.32 6.06 93.19 6390.36 0.10 1.00 10.00 100.00 1000.00 10000.00 RETE LS with base index LS without base index Sequential EVL Execution time (s) Engine name Execution time of the engines 100k 200k 500k
  • 13. Lessons Learnt • Performance of sequential EVL engine can be improved by Viatra • Incremental engine speeded-up the execution by 1481x • Duration of the query translation is negligible • https://github.com/lowcomote/evl-viatra-prototype Limitations: • First-order operations on all instances of model elements • Single-value references, attributes • Integrated to EOL, EVL 13
  • 14. Future Work • Integrate to Epsilon Transformation and Pattern Languages • Extend supported elements: • Multi-value references • User-defined operations • Performance comparisons with the parallel EVL engine [5] 14
  • 15. Acknowledgements • This work is partially funded by the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 813884, • and the National Research, Development and Innovation Fund of Hungary, financed under the 2019-2.1.1-EUREKA-2019-00001 funding scheme. • The authors are grateful for the valuable feedback of Géza Kulcsár and the anonymous reviewers about the paper. 15
  • 16. References [1] Márton Búr, Gábor Szilágyi, András Vörös, Dániel Varró: Distributed graph queries over models@run.time for runtime monitoring of cyber-physical systems. Int. J. Softw. Tools Technol. Transf. 22(1): 79-102 (2020) [2] Dániel Varró, Gábor Bergmann, Ábel Hegedüs, Ákos Horváth, István Ráth, Zoltán Ujhelyi: Road to a reactive and incremental model transformation platform: three generations of the VIATRA framework. Softw. Syst. Model. 15(3): 609-629 (2016) [3] Charles Forgy: Rete: A Fast Algorithm for the Many Patterns/Many Objects Match Problem. Artif. Intell. 19(1): 17-37 (1982) [4] http://atenea.lcc.uma.es/projects/LinTra.html [5] Sina Madani, Dimitrios S. Kolovos, Richard F. Paige: Parallel Model Validation with Epsilon. ECMFA 2018: 115-131 16