J2ee Best Practices

        J2EE Best Practices using a Real-life Example.


  


  


 Architecture:      How to Go… eBay

Carol McDonald

From there…

…to here?

Microsoft IIS

J2EE™ Container

Presentation Tier

eBayISAPI.dll EJB™ Container Business Tier

MSXML

Services

Logging

Web Container

Sun™ Tech Days

Security

[email protected]

Configuration

Technology Evangelist

Integration Tier





Monolithic Proprietary







1

Layered Loosely coupled Modular Standards based

2


 Is       What Needed…


  

™
 J2EE       Core Patterns


  


Collection of best practices and Patterns for for

J2EE™ architecture




Learning the technology is not enough
Industry best practices


Also J2EE™ architecture bad practices and





Proven solutions

refactorings
Based on Sun Java Center Services experience

Proven solutions
How to avoid bad practices?


Reuse design
Robust and scalable architecture
Improve quality
Flexibility and maintainability


Create a common vocabulary

3

4

1


 BluePrints       Java Program TM


  


http://java.sun.com/blueprints
JavaTM Technology Series


Web Tier best practices
EJB Tier best practices

books


Persistence Options

 Designing Enterprise Applications


Container Managed Persistence

with the J2EE Platform, 2nd Edition 


  


        Agenda

Covers J2EE 1.3 platform features TM


J2EE Performance

 Designing Web Services with the


JMS Best Practices (depending on the time)

J2EE Platform (available soon)


Sample applications  Java Pet Store Demo  Java Adventure Builder Demo  Concrete code for the BluePrints

guidelines  Starting points for your applications

5

6


        Application Resources


  


        Standard performance

problems


  





Watch memory, and garbage collection
watch network traffic
Minimize object creation

Standard application resources:


CPU




Memory




I/O







Network




Database





Use Caching and Pooling

Watch for memory leaks inefficient database access, poor SQL code
Use Caching and Pooling


7

8

Watch disk I/O
Avoid excessive logging

2


       


  


  


        Layering


   

     

 

Web Tier or Client

Presentation Layer

Network

Data Transfer Objects i.e. Business Data Values

EJB Session Beans

Business Logic Layer Service Layer

Entity EJBs

9

Domain Layer Persistent Entities

10


        Rationale for Layers


  


  


        Model View Controller


Model = business data and rules, shared by all clients


More flexible, extensible:


Controller defines the application behaviour, interprets user

actions and maps them into actions performed on the model


Presentation Layers


View = renders the contents of the Model


Frequent changes, design to be flexible


Business Services

Client

Thin Browser


Implement “use-cases”

HTTP Request

HTML Page


Public Service interface
Interface should remain valid for changes in presentation

Presentation Layer

Controller

View

and data store


Data Layers

Model Layer


Fewer changes, designed to optimize data access

Persistence Layer

11

Problem Domain Services

Enterprise Resources

12

3


  


        Model View Controller


  


        Service to Worker = MVC Pattern with more detail

Separate: Model data Control Presentation of view 

  

 

  

Command Factory

3. Parse Request 5. Create 1. Request Client

   

     

  










4. Get cmd

Front Controller

6. Execute

2. Common logic




Command Helper

8. Dispatch 10. Response

 

7. Business logic

View Dispatcher


 

View Helper Data Bean

Servlet JSP

9. Get Data

Java™ Objects

13

14


    Use an     MVC Framework


  


  


   


       


Use Struts, JSF, or Sun Java Studio ... application

framework
Frameworks provide a skeleton which you can extend

Form Bean

View 1

1 ActionServlet (Controller)


Request processing

3

2 4


Faster, easier development of

Action Action

Model

5


Response generation

Action


Improve

View 2 Struts-config.xml ........ ........ ........


Design
Modularity

Action


Maintainability

15

16

4


      Struts and Core  J2EE Patterns Front Controller

Command


  


  


        Ebay.com: Presentation Tier

Model Request Handler

Request Request Action Servlet

Action Mapping (xml)

Action

Intercepting Filter

Business Service

Front Controller

Navigation

Client

Command

?

Business Service

and Dispatch

Client Navigator

Dispatcher

Uses Response

XML

View Processor

JSP View

Action Forward

Intercepting Filter

Action Form

View Processor

Transformer Helper

Core J2EE

XSL

Pattern

Response

Strategy

View

Dispatcher

View Helper

17

18


       

Business Delegate


  


  


        Ebay.com: Presentation Tier

Service Interface

Request Handler




Client independent from ejb tier
Ejb details hidden
Mock objects can be used to test client w/o ejb tier

Request Intercepting Filter

Front Controller

Navigation

Command

Service Locator

Dispatcher

XML

View Processor Intercepting Filter

Response

Business Service

and Dispatch

Client Navigator

Business Delegate

View Processor

Transformer Helper

XSL

Core J2EE Pattern Strategy

19

20

5


        Cache for Performance


  


Some things can be cached and shared, use a

singleton:
InitialContext object (JNDI object used for lookups)
Anything retrieved from JNDI, EJB Home interfaces
Repeated lookups can be expensive! Use Service Locator Pattern


Some things are user specific, cached in session:
Cache search results when you are only displaying a few

at a time (static data)
Value List Pattern

21


        ServiceLocator code 01 public class ServiceLocator { 02 private InitialContext ic; 03 private Map cache; 04 private static ServiceLocator me; 05 06 private ServiceLocator() { 07 ic = new InitialContext(); 08 cache =new HashMap(); 09 } 10 public static ServiceLocator getInstance() { 11 if (me == null) { 12 me = new ServiceLocator(); 13 } 14 return me; 15 }

23


  


        ServiceLocator code


  

01 public EJBHome getRemoteHome(String jndiHomeName, Class className){

EJBHome home = null; 03 if (cache.containsKey(jndiHomeName)) { 04 home = (EJBHome) cache.get(jndiHomeName); 05 } else { 06 Object objref = ic.lookup(jndiHomeName); 07 Object obj = PortableRemoteObject.narrow(objref, className); 08 home = (EJBHome)obj; 09 cache.put(jndiHomeName, home); 10 } 11 return home; 12 } 15 02

24

6


  code       Client


  


      Tips for Servlet  Performance




Don't store a lot of data in the HTTP Session
Remove servlet sessions when they are no longer needed:

01 ServiceLocator serviceLocator = 02

ServiceLocator.getInstance();

03 try { 04

ProjectHome projectHome =(ProjectHome) serviceLocator.getHome(“ProjectHome.class”);

05


  

...


In logoff call session.invalidate()

25

26


      Servlet Tips  





  


      Tips for JSPs  

DO use high performance J2EE design patterns




Avoid scriptlets in JSP
Use JSTL (JSP 2.0 Standard Tag Libraries):


MVC, Service Locator





  




DON’T hold resources – explicitly free them




Pass data to JSP in Servlet request object not session
If JSP does not use the HTTP session
Use <%page session="false"%> to prevent HTTP

Sessions from being automatically created

27

28

7


       


  




Do you need declarative transactional support?
Do you need to distribute your business logic?
Do you need JMS, JAX-RPC, RMI?


   

     

  29


  


    Do you     need EJBs?




Do you need to support multiple client types?




Do you need method-level security on your objects?




Do you need Clustering for Scalability?

30


        Architecting Applications


  

ebay Case Study


  


 Use     ebay case   realization

ebay Case Study

View Items




Functional requirements captured via use-cases
Use-cases implemented using MVC & Command design patterns
Implement Business Logic as Services

View Items by category Any User


Verbs in use cases help define services

Any user can view auction items by category


Nouns in use cases help define business data

objects

31

32

8


 Items      Requirements View Design


  

ebay Case Study


  


 Items       View Design

ebay Case Study Front Controller

Centralize Request Hndlg

Transfer Object Assembler

Action Command

Transfer Object

Convert to Biz Command Assemble Data

Business Delegate

Use case Biz logic Retrieve Data by category

33

Session Facade

Value List Handler

Service Locator

Data Access Object

34


  


     Object   Data Transfer Client Object

EJB Network boundary

GetCity() GetState()

Before

DON'T DO too much network traffic

GetZipCode()

Client Object

Item Info Data Transfer Object

EJB


                       

GetItemInfo() Return DTO GetCity()

After

DO Reduce network traffic with Data Transfer Object

GetState() GetZipCode() Network boundary

36

9


  


        Data Access Object Use DAO for read-only tabular access to a large set of data, when query results are large and short-lived BusinessObject

Uses

DataAccessObject

creates/uses

obtains/modifies

Encapsulates

DataSource

Use Data Access Objects to encapsulate data-access logic -provides a simplified interface for performing complex JDBC operations -allows decoupling of data-access optimizations

TransferObject

38


           


  

Data Access Object (DAO)

public class FacadeEJB implements SessionBean { protected EbayDAO dao; public void ejbCreate() { try { dao = EbayDAOFactory.getDAO(); } catch (EbayDAOSysException se) { throw new EJBException(se.getMessage()); }} … public ListChunk getItems(String categoryId, int startIndex, int count, Locale locale) { try { return dao.getItems(categoryId, startIndex, count, locale); } catch (EbayDAOSysException se) { throw new EJBException(se.getMessage()); }} …. }

39

10


  


  List       Value Handler


For querying, filtering, and

Interface

ValueListIterator

+GetSize():int +getCurrentElement():Object +getPreviousElements(count:int):List +getNextElements(count:int):List +resetIndex():void

Stateful Session

Client

displaying large amounts of data:


ValueListHandler

ValueList

ValueListiterator

Execute Search

Value List Handler handles the search, can cache results and provides a traversable result set

DataAccessObject

ValueListHandler

Create

Execute Search

Create Return Value List

Return Sub List

Get Next

Get Sub-list Return Sub-list

<>

Client


  


        Example Caching Results

TransferObject

ValueList

Get Sub-list

Get Previous Return Sub-list Get Current

Get Current

Return Value Object Get Size

DataAccessObject

Get Size Return Size

41

42


        Returning Search Results Options


  

Stateless Session

Client




ValueList Handler as a Stateful Session EJB caches search results, and returns sub list
ValueList Handler as Stateless Session EJB:






  


        Stateless Example

Create

DataAccessObject

ValueListHandler ValueListiterator

Execute Search (index, count)

Execute Search (index, count)

Create Return Sub List

Execute Search (index+count, count)

Select * FROM item WHERE id=? LIMIT [count] OFFSET [startIndex]

Return Sub List

Instead of Value list could use JDBC 2.0 scrollable ResultSet but keeps JDBC data

ValueList

Return sub List

Execute Search (index+count, count)

Create

Return sub List

Select * FROM item WHERE id=? LIMIT [count]OFFSET [startIndex]

source connection open ! (J2SE 1.5 will have disconnected scrollable ResultSet) 43

44

11


  


        Session Facade pattern Session Bean

Entity Bean 1

Entity Bean Plain Java™ Object

2. getData()

Client

2. getData() Entity Bean 2

Data Access Object 1

Session Bean

Data Access Object 2

Session Facade

1. doThis(...)

3. process




Use the Session Facade:

Java™ Object


provide a simple interface to a complex subsystem of

enterprise beans
Enforce a clear and strict separation of business logic from

presentation and data logic 46


       Use the Session Facade


  

pattern


  


 the      Facade Use Session

To Execute complex use cases in a single network call to

To group finer grained calls to local entity beans

reduce network overhead Direct Bean Access

Don’t:

Local vs. Remote Interfaces

EJB

Remote Interfaces

Client

Local Interfaces

Network EJB Object

EJB client

Direct Entity Bean access results in excessive network overhead

EJB client

EJB Object

and multiple transactions

Network

Do: Network Client

47

F a c a d e

Pass by Value: Serialize/deserialize method parameters

EJB

Direct Entity Bean access results in excessive network overhead

Pass by Reference: Better performance

48

12


        Design your local &


  

remote EJBs

Use Session Facade to group calls to local EJBs in a Coarse Grained Interface

     !

     

     

     

Determine which EJBs will be collocated within same VM

  &

" #

  

 !

Relationship Reference

 !$ % 

  

 

49


  


        Tier eBay.com: Business


        Session EJB tips


  

ebay Case Study Dispatch Application Controller

Command

Context Object

Persistence Domain Store

DAO




Data Source

Remove stateful session beans when finished (user logs out)


Business Logic/Data Application Service Facade

Business Object

Transfer Object Assembler

Value List Handler




   

Limit size of objects stored in session beans due to memory cost and I/O for passivation

Transfer Object

Core J2EE Pattern Strategy

51

52

13


  


        Session EJB? Stateless or Stateful





        s Service Tier DON'T

Stateless session beans are pooled and can service multiple clients


  




DO NOT store large amount of data in Stateful SessionEJB
DO NOT access of Entity EJBs remotely


give best performance




Stateful store conversational state=1 per client (are activated and passivated)
For Performance Strive for Statelessness when application has a high number of users




DO NOT implement fine-grained components as remote EJBs


convert Stateful to stateless by adding parameters

to the bean methods that hold the extra state 53

54


  


      Design Patterns  Summary Mediate business processing Locate services

Session Facade Obtain composite value objects

Entity

Encapsulate data

Access data sources

Message Driven Bean


Exchanges data J2EE tiers

Service Locator


Service Locator


Holds results of JNDI lookups

Access business list


Value List Handler

Value List handler


Handles larger result sets from database queries


Business Delegate

Encapsulate data

Transactional Data Business Components


  


Data Transfer Object

Locate services

Encapsulate data

Value Object Assembler Invoke business methods


        Design Patterns Summary

Transfer Object


Simplifies the coupling between the J2EE Web &

EJB tiers

Access data sources


Session Facade


Provide business logic functionality

Encapsulate data


Data Access Object


Isolation layer for interfaces to a systems resources

Data Access Object


Etc.


www.sun.com/developer/blueprints/

55

56

14


  


       


        Data Access Options




     


Great choice for ease and performance




data rows; they should have business logic. To simply access data, use Plain JDBC encapsulated in a DAO


  


 vs.     BMP CMP  

SQL

Non-relational, non-object data access, Legacy systems, CRM, SAP...
Use J2EE Connectors


      Effect of Good   CMP

Bean State

Optimizations

24 20

1) Bean provider manages State and Data consistency 2) Bean provider handles relationships and OR Mapping

16 12

JDBC SQLJ

CMP Bean EJB Container

Persistence Manager

JDBC Driver

SQL

Simple BMP Optimized BMP CMP

8

Bean State

4 0

Database

1) Container manages State and Data consistency 2) Container/PM provides concurrency, relationships and OR Mapping

59


  

28

Database

EJB Container

Tabular access
EJBs should not be simple wrappers on database

58

JDBC Driver

Object-oriented (transactional) data model
JDO

57

JDBC SQLJ


  


EJB Container Managed Persistence




BMP Bean

in J2EE:

Benchmark Score

# of Database Calls

60

15


  


      CMP 2.0 Entity   Beans


  


        Relationships Container Managed

Entity Bean is now an abstract class, container extends it <> java.io.Serializable

Address Local Entity

Order

1

Shipping Address <> javax.ejb.EnterpriseBean

Local Entity

1 Client

Product

OrderBean

1

1

<> javax.ejb.EntityBean

LineItems

ProductOrdered

M CMP Entity Bean abstract class (Contains data handling logic)

You (bean provider) write this code.

OrderHome

M

LineItem Local Entity

CMP Entity Bean subclass (Contains persistence logic)

EJB Tier

Container provides this class.

61

62


       and Accessors for CMP

CMR


  


 2.0       CMP Relationship




public abstract class OrderBean implements EntityBean { private EntityContext context;


  

In CMP 2.0, you declare fields and relationships in deployment descriptor
Container generates all the necessary code

//access methods for cmp fields

<ejb-jar> <enterprise-beans> ... define your enterprise beans ... elements represent container-managed persistent fields ... define EJB relationships ...

public abstract String getOrderID(); //primary key public abstract void setOrderID(String id); . . . //access methods for cmr fields public abstract Collection getLineItems(); public abstract void setLineItems(Collection lineItems);

63

Handling

64

16


        Example CMP Wizard


  


      Advantages of   CMP 2.0 for developer





  

container manages the persistence and the relationships, not you!
No SQL or persistence logic in source code




Freedom from maintaining interactions with the data store
EJB™ Query Language (EJB QL)
Portable code

65

66


        2.0 Advantages of CMP for Container





  


  CMP       Some Optimizations

Optimization is possible because persistent fields are only accessible via get and set methods




Aggressive Loading
Loading data for Entities in a relationship in the

same query


Containers knows




Which field is being accessed
Whether or not field is in a relation
Whether of not there is a transaction
What's in cache
The current clustering environment
Can optimize based on above


67


  

Lazy Loading
Deferring loading of any data until it is accessed




Dirty Writes
Only update data which has been changed

in the database

68

17


  


 Standard       CMP: Vs. Vendor Specific Features


  


        Pessimistic Locking Serialized Access to same Bean C1




Time

Standard features

1000


Declarative specification of:

1000


Persistent attributes, abstract schema, relationships,

queries for finder/select methods(via EJBQL), transactional attributes




C2

State in Database

2000

2000


     



   

    

Vendor specific 3000


Tools, concurrency and consistency semantics,


     



  

   2000

 

3000



   

caching semantics, performance and usability

The entity blocks until previous transaction commits 69

70


  


        Optimistic Locking With “Update-conflict” Detection


  


        Usage Guidelines

Concurrent Access to same Bean

Time

State in Database 1000

2000

1000

2000

C1

C2


     






     



 

1000

 

   !   

3000





Recommended for

consistency when concurrent updates would be frequent

Retry

2000



   3000

!  "!   




Optimistic


Use when

concurrent access is mostly reading
For scalability
Requires exception handling

SunOne deployment descriptor

Concurrent Access to same Bean (same Primary Key) Lock obtained at update time, conflict detection

71

Pessimistic

72

18

  !' " 



       





  


        
   !' "




Sun Java System Application Server 7 supports these consistency levels (in sun-cmp-mappings.xml)


  

Chose optimistic concurrency for CMPs with “read-mostly” access

 Check-modified-at-commit = most optimistic  Check-all-at-commit  Lock-when-modified  Lock-when-loaded = most pessimistic  None

73

74


       Modes Database Isolation


  




Choose the lowest cost database transaction isolation level that avoids corrupting the data.




Transaction levels in order of increasing consistency and cost are:


        Entity Bean Tips


Read Uncommitted
Read Committed
Repeatable Read
Serializable


75


  




Do not use EJB entity beans for batch loading or queries that return large result sets. Use Data Access Objets encapsulating JDBC




Use CMP rather than BMP entity bean when possible




Do not call EJB entity bean get & set methods from client
Wrap with session beans to provide course

grain access and single transaction context

consider READ COMMITTED with optimistic locking 76

19


       


  


  


 App       Tune Server


EJB Bean Cache and Pools

      


JDBC
connection pools

          
 


Prepared statement cache size
JVM Heap size
(go to garbage collection Performance session)


Correct settings depend on application,

test to determine sizes 77

78


 Key       Parameters Tune Container





  

 &   !$ & (  #  !

 (  ( 

  !
 

Stateful session bean and Entity bean cache
Cache = EJB instances with state
Minimizes activation and passivation




   

    

 



   !                

      

     

Stateless session and Entity bean pools
Pool = EJB instances with no assigned state
Minimizes creation and initialization

79


  


        The EJB Container

80

20


  


        Entity Bean


        Entity Bean Caching




NewInstance() setEntityContext()

minimizes creation and initialization

Commit Option A
At the end of the transaction, the instance stays

EjbActivate()

Bean pool size

ready (cached) and the instance state is valid ( ejbLoad not called )

Does not exist unsetEntityContext()




Commit Option B
At the end of the transaction, the instance stays

pooled

ejbFind()

ready (cached) but the instance state is NOT valid ( ejbLoad will be called )

EJB Instance

ejbCreate() ejbPostCreate()

ejbRemove() ejbActivate()




ejbPassivate()

Bean cache size minimizes activation and passivation

EJB Object

its state is valid (passivated)

ejbStore()

EJB Instance




business method

81

82


     The Sun's EJB   Container

!$  



  

on your app server


  

  

commit-option (B|C) (entity beans)




steady-pool-size (not for messagedriven)
max-pool-size (0 = unbounded)

Max-cache-size (0 = unbounded)
cache-idle-timeout-in-seconds (0 = indefinitely)



Best Option: Usually B, if Entity will be accessed again. do profiling with your application


     The Sun's EJB   Container







Commit Option C
At the end of the transaction, neither the instance nor

ready

ejbLoad()


  

"     #       

   $  %     & 

    %   %   %    




pool-idle-timeout-in-seconds (0 = indefinitely)




Adjust a bean’s pool size:

 Increase when observing excessive creation

and deletion of bean instances  Decrease when accumulating a large number

83

84

of instances in pool

21


  


        Stateful Session Bean


     The Sun's EJB   Container

!$  

Bean cache size minimizes activation and passivation

does not exist




Max-cache-size (0 = unbounded)




cache-idle-timeout-in-seconds (0 = indefinitely)




removal-timeout-in-seconds (stateful session)

ejbRemove()

1) newInstance()

removal time out

2) setSessioncontext()

"     #       

    $  %     & 

    %   %   %    




ejbPassivate()

3)ejbCreate()

Method-ready EJB Object

passive

EJB Instance


  

EjbActivate()

business method

85

86


        Bean Stateless Session and Message Driven Bean


  


     The Sun's EJB   Container

  

Bean pool size minimizes creation and initialization


  




steady-pool-size (not for messagedriven)
max-pool-size (0 = unbounded)

does not exist

1) newInstance() 2) setSessioncontext()

EjbRemove()




pool-idle-timeout-in-seconds (0 = indefinitely)




Adjust a bean’s pool size:

3)ejbCreate()

method-ready pool Business method

 Increase when observing excessive creation

EJB Instance

and deletion of bean instances  Decrease when accumulating a large number

87

88

of instances in pool

22


     The Sun's EJB   Container


  


  Tips       JDBC

 (&  &    !$





Select a good JDBC driver




Tune connection pool size
Close resources as soon as you’re done with them (in finally)

Per bean: in ejb module’s sun-ejbjar.xml

 and





  


E.g. Statements, Connections, ResultSets…

Global defaults: in server instance’s server.xml




 <ejb-container> and <mdb-container>

Use JDBC’s PreparedStatement when possible:
stmt = connection.prepareStatement(..);
Tune statement caching




Turn off Auto-Commit
Group updates into a transaction

89

90


        Database Statement




Caching


  


        Database Performance


  




Common performance bottleneck
Typical problems:

'   

 15–25% Performance improvement


primary key to big (use integer)

 Requires knowledge of database specific


Inefficient queries - sending SQL data that asks the database to do more work than necessary

properties, i.e. Oracle <jdbc-connection-pool datasource-classname= "oracle.jdbc.pool.OracleDataSource" ...> <property name="ImplicitCachingEnabled" value="true"/> <property name="MaxStatements" value="200"/>


run the SQL through the EXPLAIN SELECT command
index the columns in your WHERE clauses


Large Data Sets - processing large sets

of data in ResultSets 91

92

23


     Tips  Summary Performance






  


       EJB Tips Summary

Tips for better performance


  


EJB Container Services – use appropriately for:


Tune app server and infrastructure
Container Caching and Pools


Distributed transaction management
Robust security service

Database access


Resource management (threads, sockets, database connections)


Use JDBC for:


Container persistence


Batch loading: session bean or message bean
Large result sets: value list handler


Remote accessibility


Use CMP rather than BMP Entity Beans
Use right isolation level and database


Dispatch and life-cycle management.

transactional control (locking)

93

94


        Manage Expensive









95

Resources


  


      Design Patterns  can Significantly Help Performance

Cache “EJB homes” Cache data sources Minimize use of HTTP sessions Release database connections Remove unused stateful session beans Use local Interfaces [Session Facade]


  




Session Facade
Service Locator
Value List Handler
Data Transfer Object

96

24


       


 !
!  "     


  


        JMS Communication

 !


    


  

  


Loosely coupled: 

Producer doesn't need to know consumers or consumer method arguments


Reliable: 

Message is stored until it can be delivered


Asynchronous 

97

Producer is not blocked, consumer does not have to be available

98


     Use JMS for  





  

Asynchronous Interaction
Concurrent processing
Scalability
Batch processing




Broadcasting messages to multiple recipients
Reliable messaging
Messaging with Transaction Support


Loose Coupling

99

25


  


        Publish and Subscribe


  


        Message-Driven Bean

Concurrent Asynchronous Processing

Example: Stock Price Changes


High Scalability, High Throughput
MDB instances are pooled by the container
Allow for asynchronous concurrent message consumption

from the same destination

Container Topic

Container

JMS Provider

Price Change

Desti nation

Queue

Msg-

MDB Consume Consumer

MDB driven Instances Instances

r

Bean

Msg-Driven Bean Class

Traders/Brokers

101

102


  


        MDB Facade Pattern

Mailer MDB Asynchronous Message Delivery

Order Approval MDB

TRANSACTION

EJB Tier

LineItem Entity

Purchase JMS Topics Order Queue

Card Entity

1:m

Order Bean

Process Order <Entity EJB>

Address Entity

Procure Inventory


Use the MDB Facade:

Inventory Management Bean

Mail


provide a simple asynchronous interface to a complex

subsystem of enterprise beans

<Entity EJB>

Publish/subscribe

Purchase Order Entity MessageDrivenBean

103


  

MDBs
  Container       Managed Transactions

104

26


  


        JMS Persistent Messages


  


        Example Problem Tracking

Guaranteed Delivery - but has a performance cost

 !

)*  

0* !


   ,* !-

.*  /* 

+*  

1* !-

Existing Synchronous Client Access

EJB Tier

  

Problem Solver




A Session EJB provides the business logic for the Customers or help desk to synchronously enter Problem data, or search for similar problems
Entity EJBs provide the business logic for accessing/updating the domain Data objects

2*  

Problem Rep Company

     105

106


  


        Internal View Extension Solution

Remote Help Desk New Problem


        JMS Existing Synchronous Client Access

EJB Tier

Problem

Report Request

Rep Company

Monthly Report

107




Use JMS for loosely coupled applications that need reliable, scalable document oriented message exchange




Including Message-Driven Beans and JMS in your J2EE™ application can greatly increase performance, robustness, and efficiency

Problem Solver

Problem Handler

Report Generator


  

108

27


  


 Message     Facade   XML


       Tips for XML messaging




EJB Architectural Pattern XML msg

External Application

XML Messaging Facade JMS Queues

Asynchronous

Message Driven Bean

Java™ Value Objects

Entity Bean


Within a system just use java data objects





Ref JavaOne

109

Use XML for messages between systems…

Convert the data from XML into a Java™ class as soon as possible
Can use JAXB for this

Example implementation of XML interaction model on top of Java™ interaction model


        JMS Do's

hit




Entity Bean

Synchronous

Inside your Application don’t overuse XML
Parsing and processing can cause a performance

Entity Bean Stateless Session EJB


  

Use XML mainly for transmission

110


  




Watch message size
Only use XML when necessary (lower performance)


Only use reliable messaging when necessary (lower performance)

111

28


  


        FUTURE: JBI and J2EE

Web App. Module (JSP/Servlet API)

EJB Module (EJB API)


       


  #  

JBI BP Module (WS-BPEL and WSDL APIs)

208 “Machine” SPI Normalized Message “Bus”

208 “Binding” SPI

Binding Framework

WS-I

JMS

JCA

EDI VAN

SOAP

MQ

EIS resource

AS1/AS2

113


  

114


        Use Tools


  


        Summary


  


J2EE Patterns and Best Practices are

proven solutions




Use IDE or Xdoclet to keep EJB class, home interfrace, remote interface consistent
Use Junit for Unit testing
Use Ant to automate deployment and testing
All of these are provided with Sun Java Studio

115


Reuse design
Robust and scalable architecture
Improve quality, performance
Flexibility and maintainability


JavaTM BluePrints Program
http://java.sun.com/blueprints
Designing Enterprise Applications

with the J2EE Platform, 2nd Edition

116

29


  


        Resources:


       


  

Carol McDonald Technology Evangelist [email protected]

Sun™ Tech Days

http://java.sun.com/blueprints Java Performance Tuning

117

J2EE Performance Testing

118

30