Model-Driven Instrumentation for Dynamic Analysis of ... · Software Engineering Group, Univ. Kiel...

Model-Driven Instrumentationfor Dynamic Analysis

of Legacy Software Systems

André van Hoorn1, Holger Knoche2,Wolfgang Goerigk2, and Wilhelm Hasselbring1

1Software Engineering Group, University of Kiel, Germany2b+m Informatik AG, Melsdorf, Germany

May 03, 2011 @ WSR 2011, Bad Honnef

May 03, 2011 @ WSR, Bad Honnef

hc

Schreibmaschinentext

hc


Preprint: van Hoorn, A., Knoche, H., Goerigk, W. and Hasselbring, W. (2011) Model-Driven Instrumentation for Dynamic Analysis of Legacy Software Systems [Paper] In: 13th Workshop Software-Reengineering(WSR 2011), May 2-4, 2011, Bad Honnef, Germany.

hc


hc


hc


hc


hc


hc


hc

Rechteck

Context — DynaMod ProjectContext & Motivation

Dynamic Analysis forModel-Driven Software Modernization

WP 7 Project Management WP 8 Publication of Results

Static Analysis

Dynamic Analysis

Model-BasedTesting

Code Generation

Definition of Transformations

Evaluation

v. Hoorn, Knoche, Goerigk, Hasselbring Model-Driven Instrumentation for Dynamic Analysis May 03, 2011 2 / 15



Project Consortium:

1 b+m Informatik AG(Development partner, consortium leader)

• Comprehensive MDSD know-how• Initiated openArchitectureWare (oAW)

2 Software Engineering Group, Univ. Kiel(Research partner)

3 Dataport(Associated partner)

4 HSH Nordbank AG(Associated partner)

b+m Informatik AG, Melsdorfhttp://www.bmiag.de/

Software Engineering GroupUniversity of Kiel, Kielhttp://se.informatik.uni-kiel.de/

HSH Nordbank AG, Kielhttp://www.hsh-nordbank.de/

Dataport, Altenholzhttp://www.dataport.de/




Project Consortium:




• Model-driven engineering, operation,and evolution of software systems

• Emphasis on software quality (of service)










Project Consortium:







• Provides ICT services for public/taxadministrations









Project Consortium:







• Provides ICT services for public/taxadministrations


• Leading bank for corporate and private clients innorthern Germany








Project Consortium:





Funding:

• BMBF “KMU-innovativ”

• 2 years (01/11–12/12)

Under grant no. 01IS10051

Read More:• http://kosse-sh.de/dynamod/

• MDSM @ CSMR 2011 [vHFG+11]






http://kosse-sh.de/dynamod/



WP 7 Project Management WP 8 Publication of Results

Static Analysis

Dynamic Analysis

Model-BasedTesting

Code Generation

Definition of Transformations

Evaluation






Dynamic Analysis of Software Runtime Behavior

Context & Motivation

Motivation

Static analysis is not sufficient to study a system’s architecture comprehensively

How to Gather Runtime Data from Executing Systems? — Instrumentation

• Profiling — employed in development environments; considerable performance overhead

• Monitoring — employed in production environments; captures real usage profile

Use Cases — Online & Offline Analysis

The obtained monitoring data can, for instance, be used for

• Performance evaluation (e.g., bottleneck detection)• (Self-)adaptation control (e.g., capacity management)• Application-level failure detection and diagnosis• Service-level management• Software reengineering: (architecture) reconstruction, modernization




Motivation







• Performance evaluation (e.g., bottleneck detection)• (Self-)adaptation control (e.g., capacity management)• Application-level failure detection and diagnosis• Service-level management

• Software reengineering: (architecture) reconstruction, modernization




Motivation







• Performance evaluation (e.g., bottleneck detection)• (Self-)adaptation control (e.g., capacity management)• Application-level failure detection and diagnosis• Service-level management• Software reengineering: (architecture) reconstruction, modernization


AgendaAgenda

1 Context & Motivation

2 Model-Driven Instrumentation for Dynamic Analysis

3 AOP for Legacy Languages

4 Monitoring Instrumentation

5 Conclusions


AgendaModel-Driven Instrumentation for Dynamic Analysis





5 Conclusions


OverviewModel-Driven Instrumentation for Dynamic Analysis

Imple

men

tation

Arc

hite c

ture

Dom

ain

AST

Code

DSL

ADL

Shop

search()

Instr.code



Imple

men

tation

Arc

hite c

ture

Dom

ain

AST

Code

DSL

ADL

Shop

search()

Instr.code

Model

- Drive

nSoftw

a re

Dev

elopm

e nt



Imple

men

tation

Arc

hite c

ture

Dom

ain

AST

Code

DSL

ADL

Shop

search()

Instr.code

Model

- Drive

nSoftw

a re

Dev

elopm

e nt

Sta

tic

Anal

ysi s



Imp

lem

enta

tio

nA

rch

ite c

ture

Do

mai

n

AST

Code

DSL

ADL

Shop

search()

Instr.code

Mo

del

- Dri

ven

So

ftw

a re

Dev

elo

pm

e nt

Sta

tic

An

alys

i s

Queries

AnalysisMeasures

InstrumentationDirectives



Imp

lem

enta

tio

nA

rch

ite c

ture

Do

mai

n

AST

Code

DSL

ADL

Shop

search()

Instr.code

Mo

del

- Dri

ven

So

ftw

a re

Dev

elo

pm

e nt

Sta

tic

An

alys

i s

Queries

AnalysisMeasures


<<OpExecProbe>>

<<RespTReq>>

<<AvgRespT>>



Imp

lem

enta

tio

nA

rch

ite c

ture

Do

mai

n

AST

Code

DSL

ADL

Shop

search()

Instr.code

Mo

del

- Dri

ven

So

ftw

a re

Dev

elo

pm

e nt

Sta

tic

An

alys

i s

Queries

AnalysisMeasures


<<OpExecProbe>>

<<RespTReq>>

<<AvgRespT>>

Mo

del

- Dri

ven

Inst

rum

enta

tio

n

'@intercept#Call:CallProbe[]Call bookstore.searchBook



Imp

lem

enta

tio

nA

rch

ite c

ture

Do

mai

n

AST

Code

DSL

ADL

Shop

search()

Instr.code

Mo

del

- Dri

ven

So

ftw

a re

Dev

elo

pm

e nt

Sta

tic

An

alys

i s

Queries

AnalysisMeasures


<<OpExecProbe>>

<<RespTReq>>

<<AvgRespT>>

Mo

del

- Dri

ven

Inst

rum

enta

tio

n

'@intercept#Call:CallProbe[]Call bookstore.searchBook

Dyn

amic

An

alys

is

MonitoringEvents

Measure-ments

:AvgRTval=730

:EvalResval=true

:OpExectin=211tout=955

Results


Overview—DynaMod ExamplesModel-Driven Instrumentation for Dynamic Analysis



Public Sub searchBook()...

crm.getOffersEnd Sub

Code





Code

e[]k

AST





Code

e[]k

AST

CatalogBookstore

searchBook()

DADL (DynaMod ADL)





Code

e[]k

AST

CatalogBookstore

searchBook()

DADL (DynaMod ADL)DMeasures

:OperationInvocationCountMeasure

:AvgOperationRTMeasure





Code

e[]k

AST

CatalogBookstore

searchBook()




DInstrumentation

:OperationExecutionProbe





Code

e[]k

AST

CatalogBookstore

searchBook()




DInstrumentation


+AST





Code

e[]k

AST

CatalogBookstore

searchBook()




DInstrumentation


+AST

'@intercept#Call:OpCallIcptr["Bookstore", "searchBook","CRM", "getOffers"]

Instrumented





Code

e[]k

AST

CatalogBookstore

searchBook()




DInstrumentation


+AST


Instrumented

MonitoringRecord MonitoringEvents

:OperationExecutionRecord

tin=34tout=38

class="CRM"operation="getOffers"





Code

e[]k

AST

CatalogBookstore

searchBook()




DInstrumentation


+AST


Instrumented



tin=34tout=38


DEvent MonitoringEvents

:OperationExecution

tin=34tout=38





Code

e[]k

AST

CatalogBookstore

searchBook()




DInstrumentation


+AST


Instrumented



tin=34tout=38


DEvent MonitoringEvents

:OperationExecution

tin=34tout=38

DMeasurementst

:AvgOperationRT

avgRTMillis=5676

Measurements

count=67643:OperationInvocationCount


AgendaAOP for Legacy Languages





5 Conclusions


AOP Framework—VB6 ExampleAOP for Legacy Languages







'@intercept#Call:OpCallIcptr["Bookstore", "searchBook", "CRM", "getOffers"]






before(IcptOpCall)after(IcptOpCall)

OpCallIcptr







OpCallIcptr

callerClass:StringcallerMethod:StringcalleeClass:StringcalleeMethod:Stringobject:Variant

IcptOpCall







OpCallIcptr


IcptOpCall

Private Sub Class_Initialize()...

End Sub







OpCallIcptr


IcptOpCall


End Sub

'@intercept"Class_Initialize"]

#Execution:OpExecIcptr["Bookstore",







OpCallIcptr


IcptOpCall


End Sub



before(IcptOpExec)after(IcptOpExec)

OpExecIcptr







OpCallIcptr


IcptOpCall


End Sub




OpExecIcptr

class:Stringmethod:Stringobject:String

IcptOpExec







OpCallIcptr


IcptOpCall


End Sub




OpExecIcptr


IcptOpExec

Public Sub searchBook()…Dim iCall0 As IcptOpCallSet iCall0 = New IcptOpCallCall iCall0.init("Bookstore", "searchBook", "CRM", "getOffers")Call opCallIntcptr.before(iCall0)crm.getOffersCall opCallIntcptr.after(iCall0)

End Sub

AVB6C


AVB6C — IDE IntegrationAOP for Legacy Languages


AVB6C — IDE IntegrationAOP for Legacy Languages

Acknowledgment : Our DynaMod student assistants, Eike Schulz and Benjamin Schnoor,are responsible for large parts of the implementation!


AgendaMonitoring Instrumentation





5 Conclusions


Kieker Monitoring & Analysis FrameworkMonitoring Instrumentation

Analysis Plug-In

Monitoring log

e.g., file system, database, message-oriented middleware

e.g., trace information, workload, response times, resource utilization, loop counts

Monitoring Record

M

M

M

M

M

M

e.g., AOP-based method call interception

M

e.g., architecture reconstruction,performance evaluation, online adaptation control, failure diagnosis

Kieker.Monitoring Kieker.Analysis

Kieker Framework — Core Characteristics [vHRH+09]

• Flexible architecture (custom probes, readers, writers, analysis plug-ins)• Integrated & extensible record type model for monitoring & analysis• Logging, reconstruction, analysis/visualization of (distributed) traces• Low overhead (designed for continuous operation in multi-user systems)• Evaluated in industry case studies

http://kieker.sourceforge.net


Kieker Monitoring & Analysis FrameworkMonitoring Instrumentation

Analysis Plug-In

Monitoring log


e.g., trace information, workload, response times, resource utilization, loop counts

Monitoring Record

M

M

M

M

M

M

e.g., AOP-based method call interception

M

e.g., architecture reconstruction,performance evaluation, online adaptation control, failure diagnosis

Kieker.Monitoring

Monitoring Probe

Monitoring Controller

Monitoring Log Writer

Kieker.Analysis

MonitoringRecord

Consumer

AnalysisController

Monitoring Log Reader

Kieker Framework — Core Characteristics [vHRH+09]

• Flexible architecture (custom probes, readers, writers, analysis plug-ins)• Integrated & extensible record type model for monitoring & analysis• Logging, reconstruction, analysis/visualization of (distributed) traces• Low overhead (designed for continuous operation in multi-user systems)• Evaluated in industry case studies

http://kieker.sourceforge.net


Instrumentation of Legacy LanguagesMonitoring Instrumentation

Monitoring log


Monitoring Record

Kieker.Monitoring(Java)

• Goal: Reuse existing Java-based Kieker.Monitoring component

• Requires communication bridge among legacy language and Java• Instantiation of presented AOP framework for monitoring probes• Developed proof-of-concept implementation for VB6



Monitoring log


Monitoring Record


Kieker.Monitoring(DLL)

• Goal: Reuse existing Java-based Kieker.Monitoring component• Requires communication bridge among legacy language and Java

• Instantiation of presented AOP framework for monitoring probes• Developed proof-of-concept implementation for VB6



Monitoring log


Monitoring Record



M

M

M

M

M

M

M

• Goal: Reuse existing Java-based Kieker.Monitoring component• Requires communication bridge among legacy language and Java• Instantiation of presented AOP framework for monitoring probes

• Developed proof-of-concept implementation for VB6



Monitoring log


Monitoring Record



M

M

M

M

M

M

M

Public Sub searchBook()...'@intercept#Call:OpCallIcptr["Bookstore", "searchBook",

"CRM", "getOffers"]crm.getOffers

End Sub

• Goal: Reuse existing Java-based Kieker.Monitoring component• Requires communication bridge among legacy language and Java• Instantiation of presented AOP framework for monitoring probes• Developed proof-of-concept implementation for VB6



Monitoring log


Monitoring Record



M

M

M

M

M

M

M

Public Sub searchBook()...'@intercept#Call:OpCallIcptr["Bookstore", "searchBook",

"CRM", "getOffers"]crm.getOffers

End Sub

Public Sub before(icptOpExec As IcptOpExec)Dim tin As Varianttin = monitoringCtrl.currentTimeNanosicptOpExec.object = tin

End Sub

Public Sub after(icptOpExec As IcptOpExec)Dim r As OperationExecutionRecordDim tin, tout As VariantSet r = New OperationExecutionRecordtin = icptOpExec.objecttout = monitoringCtrl.currentTimeNanosCall r.initFields(icptOpExec.class, icptOpExec.class, icptOpExec.class, tin, tout)Call monitoringCtrl.writeRecord(r)

End Sub


OpExecProbe


IcptOpExec

• Goal: Reuse existing Java-based Kieker.Monitoring component• Requires communication bridge among legacy language and Java• Instantiation of presented AOP framework for monitoring probes• Developed proof-of-concept implementation for VB6


AgendaConclusions





5 Conclusions


ConclusionsConclusions

Summary

• Context: DynaMod project for model-driven modernization

• Model-driven instrumentation for dynamic analysis

• AOP for legacy languages

• Monitoring instrumentation employing AOP framework and Kieker

• Current focus on Visual Basic 6

Future Work

• Refine and extend meta-models for queries, measures, instrumentation etc.

• Development of corresponding tool support — e.g., M2M/M2Code transformations

• Evaluate approach on other abstraction layers — e.g., DSLs, ASTs

• Additional AOP features — e.g., loops, branches

• Additional programming languages — e.g., COBOL, Natural, Structured Text

• Study performance overhead


Literature

Thomas Stahl and Markus Völter.

Model-Driven Software Development – Technology, Engineering, Management.Wiley & Sons, 2006.

André van Hoorn, Sören Frey, Wolfgang Goerigk, Wilhelm Hasselbring, Holger Knoche, Sönke Köster, Harald Krause, Marcus Porembski, Thomas

Stahl, Marcus Steinkamp, and Norman Wittmüss.DynaMod project: Dynamic analysis for model-driven software modernization.In Andreas Fuhr, Wilhelm Hasselbring, Volker Riediger, Magiel Bruntink, and Kostas Kontogiannis, editors, Joint Proceedings of the 1stInternational Workshop on Model-Driven Software Migration (MDSM 2011) and the 5th International Workshop on Software Quality andMaintainability (SQM 2011), volume 708 of CEUR Workshop Proceedings, pages 12–13, March 2011.Invited paper.

André van Hoorn, Holger Knoche, Wolfgang Goerigk, and Wilhelm Hasselbring.

Model-driven instrumentation for dynamic analysis of legacy software systems.In Proceedings of the 13. Workshop Software-Reengineering (WSR ’11), 2011.

André van Hoorn, Matthias Rohr, Wilhelm Hasselbring, Jan Waller, Jens Ehlers, Sören Frey, and Dennis Kieselhorst.

Continuous monitoring of software services: Design and application of the Kieker framework.Technical Report TR-0921, Department of Computer Science, University of Kiel, Germany, November 2009.


Model-Driven Instrumentation for Dynamic Analysis of ... · Software Engineering Group, Univ. Kiel...

Documents

Transcript of Model-Driven Instrumentation for Dynamic Analysis of ... · Software Engineering Group, Univ. Kiel...