Transcript The Enterprise Java Beans Part Two

The Enterprise Java Beans
Part Two
Biduri Kumala
Christine Angelina
Debora Y.N
Desiana Nurul Maftuhah
Farania Gama A.R
Hipasdo Abrianto
Topics
Covering:
• Chapter 12: Persistence Best Practices
• Chapter 13: EJB Design Strategies
• Chapter 14: Clustering
Chapter 12
Persistence Best Practices
Why? and What to Learn?


Most modern business applications require us
to persist data—create, update, delete,etc
Thus, in this chapter we will cover:



When to use and when not to use entity beans
How to choose between CMP and BMP
A collection of persistence best practices, such as:
versioning EJB components, dealing with a legacy
data design
Note: It’s assumed that we use relational database
1* When to Use Entity
Beans?

Actually we have 2 approaches to persist information:



Through entity beans (CMP or BMP)
Session beans plus JDBC
Points to compare both of them:








Control
Parameter Passing Analogy
Procedural vs. Object-Oriented
Caching
Enforcement of Schema Independence
Ease of Use
Migration
Rapid Application Development
2* Choosing Between CMP & BMP







Code Reduction and Rapid Application Development
Performance
Bugs
Control
Application Server and Database Independence
Relationship
Learning Curve and Cost
2b* Choosing Granularity for Entity
Beans


Granularity  how big/small our entity beans
are
Choices:


Small grained entity beans
Make some of our entity beans be Java classes (in
CMP: ‘dependent value classes’)
...Dependent Value Classes

Example:
package examples;
public class LineItem implements java.io.Serializable{
private String product;
private int quantity;
public void setProduct(String product) {this.product=product;}
public String getProduct {return product;}
public void setQuantity(int quantity) {this.quantity=quantity;}
public int getQuantity {return quantity;}
}


Don’t forget the rules!
(+) quick to develop
(-) lose many of the entity bean benefits
3* Persistence Tips and Tricks
• Beware the Object-Relational Impedance
Mismatch
• Hard-Coded vs. Soft-Coded SQL
• When to Use Stored Procedures
• Normalizing and Denormalizing
• Use Our EJB Object Model to Drive Our
Data Model
3* Persistence Tips and Tricks
• Follow A Good Data Design Process
• Use Surrogate Keys
• Understand the Impacts of Database
Updates
• Versioning EJB Components
• Living with a Legacy Database Design
• Handling Large Result Sets
Chapter 13
EJB Design Strategies
Stateful vs Stateless
Statelessness leads to increased scalability
 Stateful backers: Have to rearchitect entire
system to accommodate statelessness.
Two Virtues of Statelessness:
 Ability to easily pool and reuse components at
little or no overhead.
 Any components can service the client’s need.

Statelessness (Client-spesific Data)


You need to push client-spesific data into
stateless bean for each method invocation.
How?
Pass the data as parameters into the bean method.
 The bean to store data persistently on behalf of a
client.
 Store the client-spesific data in a directory structure
using JNDI.

Advantages of Messaging






Database performance: perform relational database work.
Quick response: allows client to process other item when
they would otherwise be blocking for the method to
return.
Request prioritization: asynchronous server can prioritize
messages in different order then they arrive into the
system.
Rapidly assembling disparate systems: provides rapid
development for systems that have to communicate with
one another.
Loosely coupled system: enables loose coupling between
application.
etc
When not to use Messaging?







When you’re not sure if the operation will succeed.
When you need a return result.
When you need an operation to be part of a larger
transaction.
When you need to propagate the client’s security
identity to the server.
When you are concerned about request performance.
When you want a strongly-typed, OO system.
When you want a tighter, more straightforward system.
How to guarantee a response time
with capacity Planning
Limit the amount of work any given EJB server instance can
process at any one time.
 by limiting the thread pool of your EJB server.
 by limiting the bean instance pool.
How to achieve singletons with EJB?
Singletons  a single instantiation of a class with one global point
of access.
To create a singleton  use a static keyword when defining a class.
But, one restriction of EJB  you CAN’T use static fields in your
bean.
So, how?  JNDI
Wrap Entity Bean with Session Bean
The business process components map very well
to session bean, and the data components map
very well to entity bean.



Session beans perform bulk create, read, update, delete
(CRUD) operations on behalf of remote clients.
Makes entity bean into an implementation detail of
session bean.
The entity beans are never seen to the external client.
Tips and Tricks to make your entity
beans high-performing







Entity bean should not be called directly from remote
clients.
Use your container’s caching option as much as
possible.
Be sure your transactions run on the server as short as
possible.
Having your container batch JDBC updates all at once
at the end of transaction.
Use container managed persistence, if possible.
Lazy-load some of your fields rather than loading it all.
etc
Choosing between local and remote
interface
Local interface  allow you to pass
nonserializable parameters around which is
handy.
For optimal performance: build system using all
local interfaces, then have one or more session
bean wrappers with remote interfaces, exposing
the system to remote client.

How to debug EJB Issues
True debugging is a problem with EJB.
Try these several options:
 IDE Debugging environments.
 Check your database’s logfile.
 Use a JDBC driver wrapper that logs all SQL
statements.
 Use the keepgenerated option
 Decompile the offending classes.
Partitioning Your Resource
Bean-spesific resources:
resources your bean uses that are tied to spesific
data instance in an underlying storage.
Bean-independent resources:
resources that can be used over and over again,
no matter what underlying data your instance
represents.
Assembling Components
The problem with assembling heterogeneous
components is getting them all to work together:
 Need to write some workflow component that
maps to each vendor’s proprietary API and
object model.
 Data mapping.
 Every participant component must agree on a
standard representation, or schema, for
exchanged data.
Bean Reusability
• Reuse as given: The application
assembler uses the acquired bean as it is
to build an application.
• Reuse by customization: The application
assembler configures the acquired bean
by modifying the bean properties to fit the
spesific needs of the application.
• Reuse by extension (subclass): The
application assembler creates custom
application-spesific beans by subclassing
the prebuilt acquired beans.
Legacy Integration with EJB
Most large-scale EJB/J2EE deployments
involve one or more existing systems.
You have two basic choices for dealing with
this legacy system:
• Rewrite that existing system using EJBs.
• Bridge into that existing system.
Chapter 14
Clustering
.: CLUSTERING :.
• Definition :
 -n . Close group or bunch of similar people
or things or occurring together (Oxford D.)
… loosely coupled group of servers that
provide simple view of the services that they
offer individually…(Mastering EJBs).
What’s The Topic ??




Approaches and characteristic of large-scale
system with J2EE application servers.
Approaches that can be employed to
instrument clustered EJBs.
Issues related to design clustered EJB systems.
Issues that impact EJB performance in
clustered system.
Scale System Requirement
•
•
•
Large Scale v.s. Small Scale
 What’s the different ??
Fact ??
Characteristic of Large- Scale System :
1. Reliability
2. Availability
3. Serviceability
Basic Terminology







Cluster
Node
Load Balancing
Highly available sys.
Fail-over (Fo)
Request-Level Fo
Transparent /
Automatic Session Fo.






Single access point
Transaction/second
Request/second
Arrivals per second
Throughput
Invocations/second
Partition Cluster

3 - Tier architecture
 runs the Web

4 – Tier architecture
 separate the web
server component and
Server component
application server
(servlets and JSPs)
component into separate
and application server
processes.
component (EJBs) in
same important
process. to know what difference between them.
There’s
Some may not consider the browser to be a-separate tier,
since no logic runs in……
How EJBs Can be Clustared

Many Location, that vendors can provide
clustering logic (like load balancing or Fo
logic):
 JNDI driver
 Container
 Home Stub
 Remote Stub
- Idempotence

-
Is one method can be called repeatedly with the
same arguments and archieve the same result.
 In a distributed system, it doesn’t impact the state
of the system. Because it can be called called
repeatedly without worry of altering the system.
 The idempotent, leaves thevalue in the persistant
store the same no matter how many invocation of
method () occur..
Continued….

A failed request could’ve occurred may cause, :
1. After the request has been initiated, but before the
method invocation in the serverhas begun to
execute.
2. After the method invocation on the server to
executed, but before the method has completed.
3. After the method invocation on the server was
completed, but before the response has been
successfully transmitted to the remoted client.
Stateless Session Bean Clustering

We can cluster each type of EJB componenent,
firstly begin with stateless session be ans :
 Load Balancing: all method invocations on the
remote home stub and remote stub can be load
balance. EJB server give the flexibelity here..
 Fail Over: automatic Fo on remote home stub can
always occur,if the called method is idempotent .
- Recall the stateless session bean home stubs have
only one method which is an empty () method.
Statefull Session Bean Clustering

EJB server vendor support replication of state
.When a statefull session bean is creation,the must
must be copied to another machine.The bean is
routine synchronized with it’s backup to ensure
that both location are current.
 Replication of statefull data typically can occurs :
1. At the end of every method
2. After the commit of a transaction
Continued…

Most EJB servers perform statefull Fo in one of
two ways, such :
Load Balancing: remote house stubs can freely
load balance create (..) request to different server
in the cluster.
Fail Over : !! That Fo can be occur with statefull
session if the state is replicated across a cluster or
the methods are idempotent.
Entity Bean Clustering

Load Balancing: coding EJB correct, will
wrap entity beans with a session bean
façade most access to entity EJBs should
occur over local interfaces by in-process
session beans rather than remote clients.
 Fail – Over : always access entity beans
using local interfaces, Fo make little sense.
Continued…

Caching: container provides many different types
of caching algorithm, which have the principles
reduce the frequency of ejbLoad () and ejbStore().
 Read-Only Caches: containt a bunch of read-only
entity beans, useful because most enterprises data
is read-only.Implement a validation strategy that
determines when the data in read-only instance is
no longer valid and shold be reloaded from the
persistent store.

Common algorithm include :
1. Timeout
2. Programmatic
3. System-wide notification
 Distributed Shared Data Object Caches.
advanced EJB server feature.This immediately
have obvius problem :cache consistency.Detect
collisions at the database level, through database
triggers.
Continued …
Read- Mostly Caches : have read – only entity
beans that also updated every now and then
without having the burden of a true distributed
shared object cache. The idea
1. When perform the operation
2. When perform a write operation
 The read-mostly characteristic :
1. Each cache uses the different JNDI name.
2. Pattern only requires the use of read only cache.

Massage-Driven Bean Clustering
Different another, because their don’t have home
and remote interfaces, any stubs or skeletons
remotely.
 Like a consumer, really about JMS clustering and
depend upon the clusterable features of the JMS
server and destination.
 Fo of message – driven beans is integrated into the
very nature of the beans themselves.occurs when
being processed is acknowledged as unsuccessful
to JMS server.

Other EJB issues
•
•
First Contact : ehen want to use EJB component, client must always first
connect to the JNDI tree.
Type of Clustered naming server :
1. Centralized
2. Shared, Replicated.
Initial Access Logic
•
•
when application server provides a centralized server, the logic that clients use
to get access to the cluster is simple.
1. Hard-code DNS code
2. IP address of centralized naming server
Arcitect for J2EE vendors that support a shared, replicated naming server by :
1.DNS round robing.
2. Software proxies
3. Hardware proxies