1

Foundations of Hybrid and Embedded Software Systems

UC Berkeley: ChessVanderbilt University: ISISUniversity of Memphis: MSI

NSF

Model-Based Design

DSML CompositionModel Synthesis andModel Transformation

Janos Sztipanovits and Gabor KarsaiNovember 14,2002

2Sztipanovits/Karsai

NSF

Middleware Services and OS

ComponentLibraries

MiddlewareOS

FrameworkLibraries

PlatformQoS ORB OSSchedulingFault

Tolerance

CORBAIDL

QoSAPI

POSIXAPI

ReplicationAPI

LibraryComponents

LibraryComponents

LibraryComponents

LibraryComponents

Model-Based Design andComposition

Auto-Generated Code Model-basedGenerators

GeneratorTechnology

Domain-Independent Models(models of computations: SDF, FSM, CSP, …)

User Interface

DecisionSupportControl

SensorSystem

Domain-Specific ModelsDS ModelLibraries/Patterns

DI ModelLibraries/Patterns

Domain-SpecificModeling

Technology

ToolsApplication

2

3Sztipanovits/Karsai

NSF Platforms(There are many…)

Hardware CPU, MEM, I/O

Operating System

QoS Middleware

Application

Hardware CPU, MEM, I/O

Operating System

QoS Middleware

Application

ControlAlgorithm

QoS parameters

Measuredvars.

Controlvars.

Application

OS

Strong isolationbetween SW andHW by Active Control

TTP Bus

ServoController

ServoController

CNI

ControlData Processor

BG BG

TTPController

CNI

ControlData Processor

BG BG

TTPController

ServoController

ServoController

CNI

ControlData Processor

BG BG

TTPController

CNI

ControlData Processor

BG BG

TTPController

Time-Triggered Architecture (distributed, hard real-time, safe)

QoS Middleware (such as CORBA)

Integration framework, composition mechanisms, components

4Sztipanovits/Karsai

NSFChallenges in Model-Based Design

ImplementationSpace

Mapping

ApplicationSpace

Gen./Synth.

DSML

ApplicationModels

DSML

Implement.Models

Pi-1

Pi

Pi+1

Composition of

• Domain Specific Modeling Languages (DSML)

• Model Synthesis

• Model Transformation

Design Model-BasedDesign

Model-BasedDesign of Embedded Systems

3

5Sztipanovits/Karsai

NSFSpecification of Domain Specific Modeling Languages (DSML)

ConcreteSyntax

C

AbstractSyntax

A

SemanticDomain

SSemantics

parses to

ConceptsRelationsWell formed-nessrules

Mathematical abstraction for specifying the meaning of models

Notation forrepresenting models

L = < C, A, S, MS, MC>

MC

MS

6Sztipanovits/Karsai

NSF Concrete Syntax and Abstract Syntax

Notation forrepresenting models:E.g.: Block Diagram

ConcreteSyntax

C

AbstractSyntax

A

SemanticDomain

SSemantics

parses to

MC

MS

Mathematical abstractionfor specifying the meaning of models

But What About S?

Signal Flow Language (SF)Concepts, RelationsWell formed-ness rules:

Self.InputPorts()→forAll(ip ip.src()→forAll(x1,x2 x1=x2))

UML-CD/OCL

4

7Sztipanovits/Karsai

NSFSemantics via Meta-Modeling

ConcreteSyntax

AbstractSyntax

SemanticDomain

Semantics

parses to

Meta-modeling language withwell-defined semantics

ConcreteSyntax

AbstractSyntax

SemanticDomain

Semantics

parses to

Represented byMeta-model

Semantics

DSML

Meta-modelStructuralSemanticsDSML:

StateFlowDOMAIN-MODEL

META-MODEL

Meta-Model of StateFlow using uml/OCLas meta modeling language.

8Sztipanovits/Karsai

NSF Semantics via Translation

ConcreteSyntax

AbstractSyntax

SemanticDomain

Semantics

parses to

ConcreteSyntax

AbstractSyntax

Semantics

parses to

SemanticDomain

translator

DSML

Semantics

Modeling language withwell-defined semantics

Synchronous Dataflow (SDF)

translatorBehavioralSemantics

ASF

Hierarchical Signal Flow (HSF)

Lee, Sangiovanni-Vincentelli

5

9Sztipanovits/Karsai

NSF

Consistency!

DSML Composition

Semantic domain impact!

ComposedBehavioralSemantics:HYBRIDSYSTEM

ConcreteSyntax

AbstractSyntax

SemanticDomain

Semantics

parses to

AL1

L1||L2ConcreteSyntax

SemanticDomain

Semantics

parses to

StructuralSemantics

AbstractSyntax

translation

HSF+ SDF+Cont.Dyn

AL2

ConcreteSyntax

AbstractSyntax

SemanticDomain

Semantics

parses to

ConcreteSyntax

SemanticDomain

Semantics

parses to

AbstractSyntax

translation

FSM CSP

10Sztipanovits/Karsai

NSF Simple Support for Compositional Meta-Modeling

`

Child inherits all associations except containment associations where parent functions as container.

Interface Inheritance

Child inherits all of the parent’s attributes and those containment associations where parent functions as container.

Implementation Inheritance

Complete equivalence of two classes Equivalence

Informal semanticsSymbol Operator

Composition Operators

Metamodel composition with GME

6

11Sztipanovits/Karsai

NSFResearch Agenda on DomainSpecific Modeling LanguagesPrecise, compositional meta-modeling

Multiple aspect modeling in the compositional meta-modeling framework

Practical issues:– Examples, meta-model libraries– Meta-programmable tools– Link to UML-2

12Sztipanovits/Karsai

NSF Model Synthesis and Transformations

• Model Synthesis

• Model Transformation

Model-BasedDesign of Embedded Systems

MatlabCode-Gen.

if (inactiveInterval != -1) {int thisInterval =

(int)(System.currentTimeMillis() -lastAccessed) / 1000;

if (thisInterval > inactiveInterval) {invalidate();

ServerSessionManager ssm =

ServerSessionManager.getManager();

ssm.removeSession(this);}

}

}

private long lastAccessedTime = creationTime;

/*** Return the last time the client sent a request

associated with this

* session, as the number of milliseconds since midnight,January 1, 1970

* GMT. Actions that your application takes, such asgetting or setting

* a value associated with the session, do not affect the

access time.*/

public long getLastAccessedTime() {

return (this.lastAccessedTime);

}

this.lastAccessedTime = time;

/*** Update the accessed time information for this session.

This method* should be called by the context when a request comes

in for a particular* session, even if the application does not reference

it.*/

public void access() {

this.lastAccessedTime = this.thisAccessedTime;

this.thisAccessedTime = System.currentTimeMillis();this.isNew=false;

}

lastAccessedTime = 0L;

lastAccessedTime = ((Long) stream.readObject()).longValue();maxInactiveInterval = ((Integer)

stream.readObject()).intValue();

isNew = ((Boolean)stream.readObject()).booleanValue();

MatlabCode-Gen.

Domain-Specific Modeling Languages

GeneratorModel-Based GeneratorTechnology

- Modeling of generators- Generating generators- Provably correct generators - Embeddable generators

ConfigurationSpecification

Analysis ToolCode

Pi-1

Pi

Pi+1

7

13Sztipanovits/Karsai

NSF Meta-generators: Model Transformations in Tool Integration

Approach (Karsai): Meta-models for source and target models plus transformations, then generating the transformer

Transformer

Domain-specific model

Target model

Meta- model for source

Meta- model for target

Meta- model for transform

GENGEN

Roles transformations play in model-based design:

•Refining a design into an implementation

•Code generation

•PIM -> PSM mapping

•Support for model interchange for tool integration

NSF

Meta-ProgrammableModel Builder

(GME)GENGEN

GeneratorSpecification

G: ASF →ASDF

Instance of

ASDFMeta-model-2

lSF∈ LSFHSF Domain model

lSDF ∈ LSDFSDF Executable modelSDF Platform

ASFMeta-model-1

instance of

Transformer

Meta-generators: Model Transformations in Component Integration

8

15Sztipanovits/Karsai

NSF Constraint-Based Model Synthesis

16Sztipanovits/Karsai

NSF Generative modeling: Extensions to Meta-Modeling LanguagesIterative (“for”) and conditional (“if”)constructs

in models can greatly enhance expressiveness while reducing complexity.

Input design: with “for” loopWhat it means:

Transformer

Replicate the middleof pipeline “Order”times!

Order = 3

9

17Sztipanovits/Karsai

NSFOther Uses of Transforms on Models: Design PatternsDesign patterns capture prototypical solutions

to a design problem. As such, they are parametric and reusable.

+operation()+add(in c : Component)+remove(in c : Component)+getChild(in index : int)

Component

+operation()

Leaf

-children

1

*

+operation()+add(in c : Component)+remove(in c : Component)+getChild(in index : int)

Composite

forall g in children g.Operation()

COMPOSITE PATTERN

+run()

Primitive

Compound

+run()+add(in c : PC_Component)+remove(in c : PC_Component)+getChild(in index : int)

PC_Component

+run()

Primitive

-children

1

*

+run()+add(in c : PC_Component)+remove(in c : PC_Component)+getChild(in index : int)

Compound

forall g in children g.run()

COMPOSITE PATTERNapplied to Compound and Primitive

Design pattern (as a transform)

Input designDesign with pattern applied

Transformer

18Sztipanovits/Karsai

NSFResearch Agenda on ModelTransformationsLanguages and tools for meta generators

Model synthesis using explicit design patterns

Model synthesis using constraint-based design-space exploration

Generative modeling extensions to languages

Embeddable generators