public class StockQuoteTimerListener implements TimerListener {
	String context; 
	String url;
	  public StockQuoteTimerListener(String context, String url){
	     this.context = context;
	     This.url = url;
	  }
	  public void timerExpired(Timer timer) {
	     System.out.println("Timer fired. Context ="+ 
		 ((StockQuoteTimerListener)timer.getTimerListener()).getContext());

	}
	   public String getContext() {
	    return context;
	   }
	}

Timers are not persistent in WAS. If you need persistent services, the WAS scheduler or EJB timer service should be used. (In version 5, the Enterprise Edition provided alarms instead of timers. Timers are a replacement for alarms but the alarms are still supported.)

As I mentioned, the J2EE context can flow from one J2EE component to another as long as they are called on the same thread. The asynchronous bean framework makes it possible to transfer all or some of the context to separate threads. The administrator can also configure which parts of the context are going to be transferred when a WorkManager is created. The following contexts can optionally be configured to propagate:

• Security (J2EE security context)
• Work area (another WAS PME explained later in this article)
• Internationalization context (another WAS PME explained later in this article)
• Application profiling definitions (another WAS PME explained later in this article)

Due to the asynchronous nature of this programming model, transactional contexts are not passed to the asynchronous beans. The asynchronous beans get their own local transaction, or they can start global transactions if XA resources are involved.

Asynchronous beans should stay away from caching connections. They can cache connection factories and then get a connection, use it, and release it at the end of the asynchronous execution. That way connection pooling is used correctly.

The WorkManager is the heart of the asynchronous bean framework. A WorkManager is essentially a thread pool. It runs Work instances asynchronously and controls the transfer of J2EE contexts to those threads. The WorkManager is the anchor point that allows developers to create asynchronous beans. Applications will look up the WorkManagers using JNDI. Once an application has obtained a WorkManager, it can use it, for example, to submit Work instances. The WorkManager returns a WorkItem to the application, which can use it to monitor the progress of the work. Applications can define resource references to work managers.

For timers, timer managers are configured on the server much the same way as work managers (see Figure 1).

The asynchronous bean framework is the basis for two new JSRs which have been submitted jointly by IBM and BEA. These are JSR 236 (Timers for Application Servers) and JSR 237 (Work Managers for Application Servers).

For further information on asynchronous beans, see the following link: http://publib.boulder.ibm.com/infocenter/ws60help/index.jsp?topic=/ com.ibm.websphere.nd.doc/info/ae/ae/welc6tech_asb.html.

Scheduler Service
Prior to J2EE 1.4, the J2EE specification defined no solution for scheduling. Applications that required scheduling to run batch jobs or scheduled events needed to do so outside the application server. In version 5, WAS introduced the scheduler service in the Enterprise Edition. Applications could now write batch jobs and scheduled events that ran inside an application server. Batch applications could now benefit from container managed transactions and J2EE security. A developer just had to provide a stateless session bean class and use the provided home and remote interface. A scheduler would then be configured administratively. Tasks could be scheduled using the scheduler API either in Java code or wsadmin. The scheduler service continues to function in WAS Application Server version 6. Figure 2 shows the administrative console screen for creating a scheduler.

Applications can then treat the scheduler as a resource and create tasks as shown:

	
//lookup Scheduler as a Resource
context = new InitialContext();

Object schedulerObj =
context.lookup("java:comp/
env/scheduler/MyScheduler");

scheduler = (Scheduler) schedulerObj;

//Lookup Task Handler as an EJB
Object taskHandlerObj = context.lookup("java:comp/
env/ejb/StockTickerPublisher");
taskHandlerHome =
(TaskHandlerHome) PortableRemoteObject.narrow(
taskHandlerObj,
TaskHandlerHome.class);
//Create Task
taskInfo = scheduler.createBeanTaskInfo();
//Associate TaskHandler with Task
taskInfo.setTaskHandler(taskHandlerHome);
taskInfo.setStartTimeInterval(startInterval
+ "minutes");
taskInfo.setRepeatInterval(repeatInterval + "minutes");
taskInfo.setNumberOfRepeats(repeats);
taskInfo.setName(taskName);
//Create Task in Scheduler
TaskStatus status = scheduler.create(taskInfo);

WAS provides the home and remote interface for the task handler object. Developers only need to provide a bean class with a special process method:

//Example task handler
public void process(TaskStatus taskStatus)
{
	try
	{
		publisherBO.publishStockTicker();
	}
	catch (StockException e)
	{
		throw new EJBException(e.getLocalizedMessage(),e);
	}
}

In J2EE 1.4, the EJB timer service is introduced to the specification. The EJB timer service allows any EJB component that implements the TimerObject interface to be called by some scheduling component. Any EJB type can be used (session bean, entity bean, or message driven bean) as a timed object. The timer service also defines a standard interface called TimerService for scheduling events. In version 6, the EJB timed objects is a special type of WAS TaskHandler. So the EJB timed components build upon the same scheduling infrastructure provided in version 6. By default, an internal scheduler instance is used to manage those tasks, and they are persisted to a Cloudscape database associated with the server process.

The EJB timer service can be configured on the desired server (see Figure 3).

You can choose to use the default pre-configured scheduler or your own (see Figure 4).

Any EJB type can implement the TimedObject interface and implement the ejbTimeout() method:

import javax.ejb.Timer;
import javax.ejb.TimedObject;
import javax.ejb.TimerService;

public class MyBean implements SessionBean, TimedObject {

  // This method is called by the EJB container upon Timer expiration.
  public void ejbTimeout(Timer theTimer) {

    // Any code typically placed in an EJB method can be placed here.

     String whyWasICalled = (String) Timer.getInfo():
     System.out.println("I was called because of"+ whyWasICalled);
  } // end of method ejbTimeout

Keep in mind that you cannot use the scheduler API to schedule tasks for EJB-based timers or use the TimerService API to configure WAS TaskHandlers.

For further information on the scheduler service, see the following URL: http://publib.boulder.ibm.com/infocenter/ws60help/index.jsp?topic=/ com.ibm.websphere.express.doc/info/exp/ae/welc6tech_sch.html.

Transactional Enhancements and EJB Entity Bean Extension

Last Participant Support
One of the major motivations behind J2EE application servers is the ability to provide enterprise applications with a transaction manager that can coordinate transactions across multiple resources. J2EE application server can coordinate transactions across multiple resources as long as they support the XA specification. In a perfect world, all back end resources would have support for XA. However, since this is not a perfect world, this is not the case. There are some back end systems that do not support XA. For example, distributed platforms often communicate with legacy systems that do not support XA, such as IBM? CICS? systems.

Ordinarily, all participants in a two-phase commit process must be two-phase commit-capable. With the last participant support (LPS), you can use a single one-phase commit (1PC)-capable resource with any number of two-phase commit (2PC)-capable resources in the same global transaction. The global transaction commit processing still takes place in two phases. In phase one, all the two-phase commit resources are prepared using the two-phase commit protocol. During phase two, the one-phase commit resource is called to commit first if all the two-phase resources successfully prepared. This way, if the 1PC resource fails, WAS can simply roll back the other resources. On the other hand, if the 1PC commit resource succeeds, WAS can commit the remaining resources. Figure 5 illustrates the process.

Keep in mind that LPS is not a complete substitute for using a resource that fully supports 2PC. It should only be used when you want to commit a one-phase capable resource (that has no 2PC support) with other 2PC resources and you want to get as close to 2PC as possible. LPS introduces a hazard of a mixed commit result because 1PC resources can't completely participate in the 2PC protocol. For example, if the WAS doesn't get a response from the 1PC resource, it has no way of knowing the status. Thus, WAS can't know whether the 1PC resource committed or not and thus does not know how to proceed. In this case, manual intervention will be required to resolve the state of the data in the 1PC resource. No special programming is needed to use LPS; it only needs to be enabled using WebSphere Application Server administration.

For more information on last participant support, see http://publib.boulder.ibm.com/infocenter/ws60help/index.jsp?topic=/ com.ibm.websphere.express.doc/info/exp/lao/ref/rla_rlinks.html.

Activity Sessions
With LPS, you can only have at most one nonphase capable resource in your XA transaction. However, if you have multiple one-phase capable resources and you need some level of coordination between them, WAS provides the activity session service. Activity sessions extend the JTA transaction model by providing an alternative unit of work to deal with 1PC. Activity sessions provide several features:

1.  Client-side demarcation of and coordination of multiple one-phase units of work: The activity session service captures the commit operations of each local and global transaction and upholds them. At the end of the activity session, the container will go back and commit (or roll back) each individual transaction. Keep in mind that the activity session, like LPS, is no substitute for two-phase commitment control, when it comes to data integrity. An activity session may result in a mixed outcome, if some of the single phase resources successfully get committed before another resource fails to commit. In that case, the activity session service will allow the programmer to retrieve the list of resources that were committed and those whose state is uncertain.
2.  Distributed context distribution using the OMG activity service: In order to support a declarative model, activity sessions define propagation of activity sessions from one component to another. The activity session context propagates according to the standards defined by the OMG activity service. See JSR 95 description at www.jcp.org/en/jsr/detail?id=95.
3.  Extensions to local transaction support: WebSphere Application Server supports the notion of local transaction containment (LTC), which allowed the container to deal with unresolved resource manager local transactions and the resources associated with them. An LTC is created by default when an XA transaction is absent. Activity sessions can be used to extend the life of an LTC.
4.  Extended entity bean life cycle support: The entity EJB can then be configured to be activated at the activity session boundaries, rather than at the transaction boundaries. The container will then keep the entity EJB active for the duration of the activity session - which corresponds to the duration of the unit of work - but no global transaction is needed to achieve this result. We often refer to this scenario as the "Commit option C plus" in that it offers a nice compromise between option C and option A by allowing the client to closely control the activation boundaries without resorting to XA transactions all the time.
5.  Association of activity session with HttpSession to scope contexts to the life cycle of HTTP clients: You can scope an activity session to an HTTP session to help scope the life of activity sessions in Web applications. This is a powerful concept that allows you to mimic user level transactions which are made up of multiple resource transactions. Combined with the extended dntity bean life cycle support, you can interact with entity bean data without them reloading for the life of an activity session. This technique makes it possible to implement some long-running transaction semantics.

There are two approaches to implementing activity sessions in your applications: the declarative (or container-managed) approach and the programmatic (or bean managed) approach. The programming model for activity sessions and the terminology which relates to the activity session service closely resemble the standard definitions that apply to transaction management, making it simple for a J2EE programmer to become familiar with the activity session semantics. Activity session scopes can be defined either through declarations that apply to EJB methods invocations (in a very similar way as for declarative demarcation of transactions) or servlets - or programmatically, using the UserActivitySession APIs - which are semantically very similar to the standard UserTransaction APIs.

Figure 6 shows an example of a scenario. Here you have two different entity beans each mapped to a different 1PC resource. The EJB methods are demarcated much like you would regular transactions only using ActivityService demarcations.

Activity sessions are not just bound to local resources. They can even be used to group multiple 2PC.

Activity sessions are being worked on for future versions of J2EE as JSR 95: J2EE TM Activity Service for Extended Transactions (URL).

For more information on the activity service, see: http://publib.boulder.ibm.com/infocenter/ws60help/index.jsp?topic=/ com.ibm.websphere.express.doc/info/exp/ae/welc6tech_as.html.

Other Entity Bean Enhancements
Websphere Application Server v6 adds some enhancements to entity beans:

1.  Dynamic query support: The current EJB 2.1 specification only allows you to define EJB QL statically in the EJB deployment descriptor. However, there are times when an application needs to construct a query dynamically. The dynamic query service provides an API you can use to pass EJB QL at runtime. The dynamic query service is automatically deployed on WAS as a stateless session bean with a local and remote interface. Dynamic queries is being worked on as part of EJB 3.0.

For more information on dynamic query, see: http://publib.boulder.ibm.com/infocenter/ws60help/index.jsp?topic=/ com.ibm.websphere.express.doc/info/exp/ae/welc6tech_que.html.
2.  Extensions to EJB QL: WAS version 6 provides several extensions to EJB QL to meet real application requirements. Some examples include: querying business bean methods, dependent values, and multiple element selects.
3.  Application profiles: WAS uses the notion of access intents on entity beans to define things such as isolation level and lazy loading of collection-based data. In WASv5, an entity bean method could only be assigned one access intent. If an entity bean needed to be accessed by different session beans that had different access intent requirements, you needed to define multiple methods that access the same data or sometimes even create new entity beans mapped to the same data with different access intents. Application profiles allow me to create multiple access intents on the same entity bean method and mark the calling session beans with a particular task. Based on the container task of the session bean, the entity bean will use the correct access intent for the particular use case. This allows entity beans to be more reusable across various data access requirements.

For more information on application profiles, see: http://publib.boulder.ibm.com/infocenter/ws60help/index.jsp?topic=/ com.ibm.websphere.express.doc/info/exp/ae/welc6tech_appprof.html.

Other PMEs

WorkArea Service
The J2EE specification defines a component style of development. Components share data by passing information through well defined interfaces. Certain information, however, gets passed automatically. For example, transactional components such as Enterprise JavaBean components pass a transactional context to each other. Another example is a servlet that can propagate a security context to a stateless session bean. Developers do not have to pass transactional or security handles from one component to another, even when calling remote session beans. J2EE application servers pass this information along transparently. However, most other data traditionally needs to be passed through interfaces. Certain data, however, could benefit from automatic propagation. Usually any data that is needed by many components (for example, profile data about a user) is a good candidate for automatic propagation. Rather than populating every interface with the same information, it would be nice if you could propagate this frequently used information using some J2EE context.

WAS version 6 provides this function through the WorkArea service. By using a simple API, components can pass context information from component to component without passing it directly through programmatic interface. The WorkArea service will propagate to local components using thread local storage and it will pass data to remote components through IIOP. Work areas can be nested as well to allow flexibility of layering your code correctly.

A client component can create a UserWork area:

	
//Lookup UserWorkArea
UserWorkArea workArea =  (UserWorkArea) ctx.lookup("java:comp/websphere/UserWorkArea");

//create workArea
workArea.begin("Profile");
workArea.set("Name","Luke Skywalker");
workArea.set("Age",new Integer(28));

//call component
myEJB.callService(params);

//complete workArea
workArea.complete();

The target component can then access the UserWorkarea to obtain the information.

void callService() throws SomeException
{
InitialContext ctx = new InitialContext();
	//lookup WorkArea
UserWorkArea workArea =
(UserWorkArea) ctx.lookup("java:comp/websphere/UserWorkArea");

	//Access Data
System.out.println("Work Area");
	System.out.println("Name : " + workArea.get("Name"));
	System.out.println("Age : " + workArea.get("Age"));
}

IBM is leading JSR 149 for adding WorkArea support to the J2EE specification.

You can find more information about the WorkArea service at: http://publib.boulder.ibm.com/infocenter/ws60help/index.jsp?topic=/ com.ibm.websphere.base.doc/info/aes/ae/welc6tech_wa.html.

Internationalization Service
Many applications require Internationalization(i18n) support. In most cases, developers can rely on using resource bundles to display data in different languages or regions. However, some i18n requirements are more than just displaying data; applications may need to alter business logic depending on the locale of the user. For example, based on what country you logged in on, an application may need to provide different tax rules. In order to be able to do this, an application would need to get some information about the locale or time zone of the user to properly handle the logic correctly. One could use the WorkArea service to pass i18n, however, having i18n data passed in a special way can benefit greatly. Consider another case where a company needs to develop an internationalized application that is meant to be deployed in various countries. Depending in which country the application is deployed, applications require certain portions of the application to run logic differently. Rather than maintaining different code bases, the i18n service can also be used to determine the locale of the server. This allows applications to deal with both their local i18n context and the caller's i18n context separately. The i18n service allows applications to access both the caller's i18n information as well as the local servers transparently.

The i18n service in WAS 6 allows applications to access the i18n context of the caller:

UserInternationalization i18nService = (UserInternationalization) ctx.lookup("java:comp/websphere/UserInternationalization");

Internationalization callerI18n = i18nService.getCallerInternationalization();

The i18n service also allows applications to access the i18n context of the running environment:

UserInternationalization i18nService =
(UserInternationalization) ctx.lookup(
"java:comp/websphere/UserInternationalization");

InvocationInternationalization invI18n =
i18nService.getInvocationInternationalization();

IBM is working on standardizing the i18n service through JSR 150.

More information can be found at: http://publib.boulder.ibm.com/infocenter/ws60help/index.jsp?topic=/ com.ibm.websphere.base.doc/info/aes/ae/welc6tech_in.html.

Object Pools
J2EE application servers provide pooling for Enterprise JavaBean components as well as resources such as JDBC connections. However, there are times when applications need to pool plain Java objects. Many times developers spend time developing their own object pools or invest in some third party library. WAS provides an object pool that can pool any arbitrary set of objects. The object pool can be configured through the WAS administrative console or through any supported WAS administrative client.

More information about object pools can be found at: http://publib.boulder.ibm.com/infocenter/ws60help/index.jsp?topic=/ com.ibm.websphere.base.doc/info/aes/ae/welc6tech_objp.html.

Startup Beans
Most applications require some sort of initialization. For example, an application may need to cache some read only data for fast access. This data can be initialized at start up. In the Web container, this can be done by marking a servlet for start up in the web.xml file. However, in the EJB container, start up logic can be challenging. The J2EE specification does not define a way to handle start up logic in the EJB container. WAS provides startup beans that can be used to accomplish start up logic in the EJB container.

More information can be found about startup beans at: http://publib.boulder.ibm.com/infocenter/ws60help/index.jsp?topic=/ com.ibm.websphere.base.doc/info/aes/ae/welc6tech_sub.html.

Dynamic Cache
The dynamic cache service is a whole caching infrastructure. It is a large topic and requires its own paper. My book (referenced at the end of this article) has a whole chapter on this topic. In short, the dynamic cache service improves performance by caching the output of servlets, commands, Web services, and JSP files. Dynamic caching features include replication of cache entries, cache disk offload, Edge-Side Include caching, Web services, and external caching.

More information on dynamic cache service can be found at: http://publib.boulder.ibm.com/infocenter/ws60help/index.jsp?topic=/ com.ibm.websphere.base.doc/info/aes/ae/welc6tech_dyn.html.

Conclusion

Acknowledgments

References