successive refinementof models for model based … · successive refinementof models for...

14
Budapest, 26-28 October 2016 SUCCESSIVE REFINEMENT OF MODELS FOR MODELBASED TESTING TO INCREASE SYSTEM TEST EFFECTIVENESS Presented by Ceren Şahin Gebizli © All rights reserved

Upload: phungkhanh

Post on 16-Jul-2019

228 views

Category:

Documents


0 download

TRANSCRIPT

Budapest, 26-28 October 2016

SUCCESSIVE REFINEMENT OF MODELS FOR MODEL‐BASED TESTING TO INCREASE SYSTEM TEST EFFECTIVENESSPresented by Ceren Şahin Gebizli 

© All rights reserved

Outline

2

• Testing Challenges in Consumer Electronics Domain

• Model‐based Testing and System Models

• Overall Approach

• Model Updates and Case Study

• Results & Conclusions

©All rights reserved

Challenges

3

• Short time‐to‐market• Limited resources• Large code base• Large models• Importance of User Perception

©All rights reserved

Model‐based Testing (MBT)

4

Effective test case generation;• Focus on features that are mostly used• Focus on scenarios that are mostly error‐prone• Focus on scenarios that reveal different failures

©All rights reserved

Test Model

AB

C

DE

System Models used for MBT

5

*Hierarchical Markov chains defined with the MaTeLo tool  (http://www.all4tec.net)

© All rights reserved

transitionprobabilities

states that can comprise sub‐models

finish state

start state

Overall Approach

6

• Update system models based on;• Frequency of usage by the end‐users• Estimated risk of failure based on static analysis

• Estimated risk of failure based on dynamic analysis

• (Re)generate and execute test cases

©All rights reserved

System Model Updates

7

• First assignments of transition probabilities based on number of visits recorded in the usage profile.

.• Next: second & third updates .based on estimated risk of error

©All rights reserved

vn‐1

v0

v1s

ti

Update based on Risk of Error

8

• Risk estimations:• Static analysis: Ratio of static code analysis alerts• Dynamic analysis: Ratio of memory leaks

• Example:  Update of the system model after the probability of error for state s is calculated as 0.2

©All rights reserved

Industrial Case Study

9

• Initial model was previously developed by the software test group in the company.

Data Collection and Estimations;• Usage Profile• Static Analysis*• Memory Profile

©All rights reserved

*Performed with the Klockwork tool(http://www.klocwork.com/)

Model Updates

10 © All rights reserved

Iterations

11 © All rights reserved

Results

12

• Reduction in the number of test cases• Detection of new faults

©All rights reserved

Conclusions

13

• Consumer electronics domain

• Context of an industrial case study for MBT of a Smart TV system

• An iterative model refinement approach• New faults were detected in each iteration

©All rights reserved

QUESTIONS?Ceren Şahin Gebizli

© All rights reserved