Upload
lynhu
View
225
Download
3
Embed Size (px)
Citation preview
Model Based Technology for Functional TestingBusiness case for BSO Budapest R&D
Company confidential – for internal use only1 © Nokia Siemens Networks
Business case for BSO Budapest R&D
MBT Concept
Model Based Testing (MBT) comes from the following idea:If we have an abstract model of the system’s behavior then we can generate all possible test-cases. The limit is defined only by the abstraction level.
Specification Function Model
Test generation
model from system perspective
Company confidential – for internal use only2 © Nokia Siemens Networks
Test script
Test framework
Test execution engine
System under test
Test generation
Milestones in Model Based Testing Adoption for Test Design• 1990s: 1st generation MBT tools become commercially available
– strong link to Model Driven Development– described system operation from user perspective
• 2006: 2nd generation MBT tools came on market– modeled system operation from specification perspective
• 2007: Start of ITEA2 D-MINT (Deployment of Model Based Technologies to Industrial Testing) project
Company confidential – for internal use only3 © Nokia Siemens Networks
to Industrial Testing) project– 25 partners, 6 countries, 8 case studies, 5 application areas– http://www.d-mint.org/
• 2010: 1st industrial standard on MBT approved at ETSI – ES 202 951 ”Requirements for Modeling Notation”– Active participation from 15 companies incl. MBT tool makers, users &
research institutes• 2011: 1st industrial conferences in EU and India dedicated to MBT
Evolution of Software Testing
FrameworksKeyword Driven
MBTManually written scriptsExecution and analysis are
Generated scriptsExecution and analysis are automatic
Company confidential – for internal use only4 © Nokia Siemens Networks
Manual
Scripts-BasedCapture/Replay
Keyword Driven
No scriptsExecution is manual
Manually written scriptsExecution and analysis are manual
Execution and analysis are automatic
Pilot project in Budapest R&D
Content: Early interface implementation in HLRWhy?• New interface introduced to HLR � we can start from scratch• Well documented: 3GPP standards, specification, fault reports, etc…• Feature is already on the market � passed through all test phasesResources:• 2 cooperative students, with the help of a full-time employee
Company confidential – for internal use only5 © Nokia Siemens Networks
• 2 cooperative students, with the help of a full-time employee• Training and remote support from Conformiq
Comparison parameters
In order to say definitely that MBT is good for us, we need to take the following things into consideration:
• Effort– How many man hours were needed in the original testing, and with MBT?
• Quality
Company confidential – for internal use only6 © Nokia Siemens Networks
• Quality– Number of fault reports: How many new faults were found with MBT, and how
serious are those?– Coverage: full requirement coverage is expected from MBT
• Competence ramp-up• Is accelerated knowledge acquisition possible?
Traditional way of functional testing
Specification writing
Test plan writing
Review OK?
Specification
Test plan
Y
Y
N
N
Company confidential – for internal use only7 © Nokia Siemens Networks
Manual execution
Manual macro writing
Review OK?
TC OK?
Framework writing / updating
Y
Y N
N
Macro
Execution
Automatized way of functional testing with TTCN
Specification writing
Test plan writing
Review OK?
Specification
Test plan
Y
N
N
Company confidential – for internal use only8 © Nokia Siemens Networks
Manual macro writing
Review OK?
TC OK?
Framework writing / updating
Y
Y
N
N
Macro
Execution
Functional testing with MBT
Specification writing
Model writing/updating
Review OK?
Specification
Model
Y
N
N
Company confidential – for internal use only9 © Nokia Siemens Networks
Scripter writing/updating
Exporting macros
Review OK?
Framework writing / updating
Y
N
Macro
Execution
Efforts in man hoursTTCN automation testing MBT
Test plan writing 150 hours Specification study 25 hours
Creating/updating model
225 hours
Model review 79 hours
Creating/updating scripter
62 hours
Manual macro writingand automation
500 hours
Generate test-cases 2 hours for all generations
Company confidential – for internal use only10 © Nokia Siemens Networks
and automation hours generations7 minutes for 1 generation
Validation 100 hours
Validate generatedtest-cases
16 hours
Update test framework
Execution 400 hours Execution 452
Sum 1150 hours Sum 860 hours
25% savings on effort
Product Quality – Fault Reports
32 faults found during original testing (FT + SyVe + Pet + Customer )• Only 9 were found in FT (out of 32)• 8 faults were found later (out of 23) that could have been found in FT
6 additional faults found with MBT (2 critical, 1 i n Implementation Spec)
Company confidential – for internal use only11 © Nokia Siemens Networks
39%
35%
26%
found in FT
could be found in FT, but found later
found with MBT only
Product Quality - Coverage
Requirement coverage:• All explicit requirements are covered by the model.• The most important implicit requirements (e.g.: feature interworking) are
also covered.
Coverage achieved by MBT is greater than by manually written tests, because it is much less effort to increase it after a certain threshold.
Company confidential – for internal use only12 © Nokia Siemens Networks
because it is much less effort to increase it after a certain threshold.
Automated requirement traceability; Forces application level implicit requirement to be covered by the model
Learning from model (Competence ramp -up)
• 1 newcomer:– only basic GSM training before (SYSTRA)
• Training material: – QML model + code– Abbreviations handout
• Learning time: 21 hours• Support time: 6 hours
Company confidential – for internal use only13 © Nokia Siemens Networks
• Support time: 6 hours
• Competence test:– General understanding of the functionality– Detailed knowledge of message sequences and triggering events– Test result: 83%
• Useful productive capacity can be raised to 100% much faster than the usual capacity planning process. The newcomer from the pilot project is already a part of the fault report solving team.
Subjective impressions
• With a few clicks, you can quickly draft many test-cases
• Easy export of test-cases in desired format
• Parameters from model go easily to exported test-cases
• Tool does the boring work of trying out multiple values
• With built-in HTML scripter we can produce test documentation in minutes
• The technology is exciting AND works in the real world
Company confidential – for internal use only14 © Nokia Siemens Networks
• The technology is exciting AND works in the real world
Conclusion
+ shorter time to market: model design and maintainenance time is shorter+ easy to maintain regression set+ full requirement coverage+ better understanding of the product needs and the functionality, increases
quality+ faults can be found in earlier phases (even in IS)+ much better structured documentation, easy to learn, more useful for
Company confidential – for internal use only15 © Nokia Siemens Networks
+ much better structured documentation, easy to learn, more useful for Innovation improvement
- time and competence for TC review and reduction- model creation competence- initial investment: license cost + time of initial model creation- follow up the changed FT set (tc history, QC)- time and competence for model review
Dept. / Author / Date
Let’s put it into use