Architectural Analysis & Design

Architectural Layers

The Logical Architecture is a large-scale organization of the software classes into packages (or namespaces), subsystems, and layers. Decisions about how these elements are to be deployed across physical computers in a networked environment are part of the Deployment Architecture.

Classes in the system are grouped into layers. A layer is a coarse-grained grouping of classes, packages, or subsystems that has cohesive responsibility for a major aspect of the the system. Layers are organized so that “higher” layers call upon services of “lower” layers, but not generally vice-versa.

Architectural Analysis & Design

Architectural Layers

•User Interface

•Application Logic and Domain Objects

•Technical Services

The user interface with the system may be through a graphical user interface or through web-based services. The user interface may consist of packages such as the swing package and the web package, where the swing package refers to the user generated classes that specialize the GUI framework

Software objects representing domain concepts. This layer may be divided into an Application layer containing facades for maintaining session state and controlling workflow to the Domain layer concpts.

General purpose objects and subsystems that provide supporting technical services such as forming an interface with a database.

Architectural Analysis & Design

Using layers helps address these problems:

•Source code changes are rippling throughout the system – many parts of the system are highly coupled.

•Application is intertwined with user interface, so it cannot be reused with a different interface or distributed to another processing node.

•Potentially general technical services or business logic is intertwined with more application specific logic so it cannot be reused, distributed to another node, ore easily replaced with a different implementation.

•There is high coupling across different areas of concern. It is difficult to divide the work along clear boundaries for different developers.

Architecture Analysis & Design

UI

(AKA Presentation, View)

Application

(AKA Workflow, Process, Mediation, ApplicController)

Domain

(AKA Bussiness, Application Logic)

Business Infrastructure

(AKA Low-level Business Services)

Technical Services

(AKA High-level Technical Services)

Foundation

(AKA Core Services, Base Services, Low-level Technical Services)

dependency More application

specific

Range of applicability

GUI Windows, reports, speech, XML, HTML, JSP, Javascript

Handles presentation layer requests, session state, workflow

Handles application layer requests, implementation of domain rules& services

Low-level business services such as currency conversion

Higher-level technical services such as persistence & security

Low-level technical services, utilities and frameworks such as data structures, threads, math, network I/O

Architecture Analysis & Design

Examples of UML notation for packages and layers

Domain

Technical Services

POS Inventory Tax

Persistence Security Logging

Horizontal partitions

Vertical layers

Package or subsystem

Architecture Analysis & Design

Guideline: Model-View Separation Principle

1. Do not connect or couple non-UI objects directly to UI objects. For example, do not let a Sale object have a reference to a JFrame window object. Windows are related to a particular application and platform, whereas the Domain objects may be reused in a new application or attached to a new interface.

2. Do not put application logic (such as tax collection) in the UI object methods. UI objects should only initialize UI elements, receive UI events (such as mouse clicks), and delegate requests for application logic to non-UI objects.

Architecture Analysis & Design

SystemCashier

makeNewSale( )

enterItem(id, quantity)

description, total

endSale( )

UI

Swing

ProcessSale Frame

Domain

…...

makeNewSale()

enterItem(id,qt)

endSale( )

Register

makeNewSale( )

enterItem(id, quant)

endSale( )

makeNewSale()

enterItem(id,quant)

endSale( )

…

System Operations shown in SSD Implementation of System Operations

Designing a Persistence Framework

This presentation illustrates the design of a persistence framework and the use of Design Patterns in constructing that framework. There are free, open-source persistence frameworks that you can use in your project. You do not have to create one of your own. In the Java domain, there is a widely used framework called Hibernate (www.hibernate.org).

Design of a Persistence Framework

Domain Layer Persistence Framework Relational Database

Name City

Harvard Cambridge

Yale New Haven

Michigan Ann Arbor

Cornell Ithaca

University Table

:University

name = Cornell

city = Ithaca

University object

PersistenceFacade

get(OID, class):Object

put(OID, object)

Retrieve from RDB

get(OID, University)

Store object in RDB

put(OID, cornellObj)

Domain object will dematerialize when application goes out of scope

Accessing Persistence Service with a Facade

Facade Pattern

Intent Provide a unified interface to a set of interfaces in a subsystem. Facade defines a higher-level interface that makes the subsystem easier to use.

Applicability

•You want to layer your subsystem. Use Facade to define an entry point to each subsystem level.

•Introduce a Facade to decouple subsystems from clients and other subsystems, thereby promoting subsystem independence and portability.

•A Facade can produce a simple default view of the subsystem.

Collaborations

Clients communicate with the subsystem by sending requests to Façade, which forwards them to the appropriate subsystem object. Clients that use facade do not have to acess subsystem objects directly.

The Facade Pattern

An Object Identifier Pattern

We need a consistent way to relate objects to records in a database and be able to ensure that repeated materialization of a record does not result in duplicate objects.

The Object Identifier Pattern proposes assigning an object identifier (OID) to each record and each object (or proxy of an object). An OID is an alphanumeric code that is unique to each object.

Every table will have an OID as primary key

The Facade Pattern

University Table

:University

name = Cornell

city = Ithaca

oid = uxc123

OID name city

uxh345 Harvard Cambridge

uxy248 Yale New Haven

uxm117 Michigan Ann Arbor

uxc123 Cornell Ithaca

primary key

Mapping between persistent object and Database

The OID may be contained in proxy object instead

The Facade Pattern

:DBProductAdapter :PersistenceFacade

pd = get(…)

//example use of the facade

OID = new OID(“XYZ123”);

ProductDescription pd = (ProductDescription)

PersistenceFacade.getInsance().get(oid, ProductDescription.class);

PersistenceFacade

getInstance():PersistenceFacade

get(OID, class)

put(OID,Object)

The Database Broker

What class should be responsible for the materialization and dematerialization of the objects from a persistent store?

The Information Expert Pattern suggests that the persistent object class itself have this responsibility – It has some of the data (the data to be saved) required by the responsibility. This is termed Direct Mapping.

Problems with the Direct Mapping approach include:

•Strong coupling of the persistent object class to persistent storage knowledge – (violation of low coupling)

•Complex responsibilities in a new and unrelated area to what the object was previously responsible for – (violation of high cohesion). Technical service concerns are mixing with application logic concerns.

The Database Broker

Second option – use an indirect mapping approach.

Create a DatabaseMapper class that is responsible for the materialization and dematerialization of objects from the database. Each persistent object will have its own Mapper class.

The Facade Design Pattern with Brokers

PersistenceFacade

getInstance( ): PersistenceFacade

get(OID, class) : Object

put(OID, Object)

ProductSpecification RDBMapper

get(OID):Object

put(OID, Object)

<<interface>> DBMapper

get(OID):Object

put(OID, Object

class

ProductSpecification FlatFileMapper

get(OID):Object

put(OID, Object

Manufacturer RDBMapper

get(OID):Object

put(OID, Object

Each mapper gets and puts objects in its own unique way, depending on the kind of data store and format.

Facade Pattern with Brokers

class PersistenceFacade {

public Object get (OID oid, Class persistenceClass ) {

DBMapper mapper = (DBMapper) mappers.get (persistencceClass);

//delegate

return mapper.get(oid);

}

public put (OID oid,, Object obj) {

Class persistenceClass = obj.getClass( );

DBMapper mapper = (DBMapper) mappers.get (persistencceClass);

mapper.put(oid, obj);

}

}

Designing the DatabaseMapper Classes

The Template Method pattern should be used to:•Implement the invariant parts of an algorithm once and leave it to the subclasses to implement the behavior that can vary.

•When common behavior among subclasses should be factored and localized in a common class to avoid code duplication.

•To control subclass extensions.

AbstractClass

TemplateMethod( )

PrimitiveOperation1()

PrimitiveOperation2( )

ConcreteClass

PrimitiveOperation1()

PrimitiveOperation2()

{ …..

PrimitiveOperation1( );

……..

PrimitiveOperation2(); …}

Template Pattern

The Hollywood Principle:

“Don’t call us, we’ll call you!”

Example: The swing GUI framework

GUIComponent

update( )

repaint( )

framework class

Template method

hook method

programmer’s classMyButton

repaint( )hook method overridden to supply class specific detail

//unvarying part of algorithm

public void update {

clearBackground( );

//call the hook method

repaint( );

}

Template Method

Consider the POS Terminal Example

<<interface>> DBMapper

get(OID):Object

put(OID):Object

Abstract PersistenceMapper

+get(OID):Object {leaf}

#getObjectFromStorage( ):Object

template method

hook method {abstract}

Template MethodOverriding the hook method

ProductDescription RDBMapper

# getObjectFromStorage(OID):Object

AbstractPersistenceMapper

+ get(OID):Object {concrete}

# getObjectFromStorage(OID):Object{abstract}

DBMapper//template method

public final Object get(OID oid) {

obj = cachedObjects.get(oid);

if (obj == null) {

//hook method

obj = getObjectFromStorage(oid);

cachedObject.put(oid, obj);

}

return obj;

}

//hook method override

protected Object getObjectFromStorage(OID oid) {

String key = oid.toString( );

dbRec = SQL execution result of

“Select* from PROD_DESC where key =“

+key

ProductDescription = new ProductDescription();

pd.setPrice(dbRec.getColumn(“PRICE”); etc

Persistence Framework

NextGen Persistence

Persistence

PersistenceFacadeclass

Abstract RDBMapper

<<interface>> DBMapper

Abstract PersistenceMapper

1

ProductDescription RDBMapper

ProductDescription FileWithXMLMapper

ProductDescription InMemoryTestDataMapper

SaleRDBMapper

Template Method

Implementation of Template Method in POSTerminal

public class PersistenceManager {

public final Object get(OID oid) {

//template method

obj:= chachedObjects.get(OID oid);

if (obj == null) {

//hook method

obj = getObjectFromStorage(oid);

cachedObject.put(oid, obj);

}

return obj;

}

protected abstract Object getObjectFromStorage(OID oid) ;

}

public class ProductDescriptionRDBMapper extends PersistenceManager { protected Object getObjectFromStorage(OID oid) { String key = oid.toString( ); dbRec = SQL execution result of

“Select * from PROD_DESC where key = “ + key ProductDescription pd = new ProductDescription (); pd.setOID(oid); pd.setPrice ( dbRec.getColumn(“PRICE”) ); pd.setItemID ( dbRec.getColumn(“ITEM_ID”) ); pd.setDescription( dbRec.getColumn(“DESC”) ); return pd; } …….}

Transactional States and the State Pattern

Statechart for the PersistentObject

New

[new (not from DB)]

OldClean OldDirty

OldDelete

Deleted

[from DB]

save

rollback / reloadcommit / update

delete

rollback / reload

delete

commit / insert

commit / delete

State Pattern

Context/problem

An object’s behavior is dependent upon its state, and its methods contain case logic reflecting conditional state-dependent actions. Is there an alternative to conditional logic?

Solution

Create state classes for each state, implementing a common interface. Delegate state-dependent operations from the context object to the appropriate state object. Ensure the context object always points to the state object reflecting its current context.

State Pattern

Collaborations (Example)

TCPConnection

Open( )

Close( )

Acknowledgement( )

TCPState

Open( )

Close( )

Acknowledgement( )

TCPEstablished

Open( )

Close( )

Acknowledgement( )

TCPListenOpen( )

Close( )

Acknowledgement( )

TCPClosedOpen( )

Close( )

Acknowledgement( )

state open( )

State PatternApplying the State Pattern to the Persistence Framework

state commit( this );

PersistentObject

commit( )

delete( )

Rollback( )

save( )

setState(PObjectState)

oid: OID

state: PObjectState

PObjectState

commit (PersistentObject obj);

delete (PersistentObject obj);

rollback (PersistentObject obj);

save (PersistentObject obj);

OldDirty State

commit( …)

delete(…)

rollback(…)

OldClean State

delete(…)

save (…)

New State

commit( …)

* 1

{//commit

PersistenceFacade.getInstance() .update(obj) obj.setState(OldCleanState. getInstance()}

Domain Persistence

PersistentObject

oid: OID

timeStamp: DateTime

commit( )

delete( )

rollback( )

save( )

ProductDescription

Sale

Whenever a Domain object class extends a Technical Services class it should be done with hesitation and for a good reason. You will be mixing application logic and the technical concern of persistence.

Persistent Objects

Architectural DesignArchitectural Views

•Logical

•Process

•Deployment

•Data

•Security

•Implementation

•Development

•Use Case

Conceptual organization of the Architecture in terms of subsystems, packages, frameworks, classes, and interfaces. Summarizes the functionality of the major software elements.

Responsibilities, collaborations, and the allocation of responsibilty to processes and threads.

Physical deployment of processes and components to nodes and the network configuration between nodes.

Overview of the data flows, persistent data schema, and the mapping from objects to persistent data.

Overview of the security schemes and points within the architecture that security is applied.

The actual source code and executables. A summary of noteworthy deliverables.

Summarizes information developers need to know about the setup of the development environment – directory structure, version control, etc.

Summary of the most architecturally significant use cases and their non-functional requirements. Summary of those use cases that illustrate significant architectural coverage or exercise many arch. elements.