Session beans are widely used to provide transactional services for local and remote clients. To write a session bean, you need an interface and an implementation class.

@Local
public interface MyBeanInt {
  public String doSomething (String para1, int para2);
}

@Stateless
public class MyBean implements MyBeanInt {

  public String doSomething (String para1, int para2) {
    ... implement the logic ...
  } 
  
}    

When you invoke a session bean method, the method execution is automatically managed by the transaction manager and the security manager in the server. You can specify the transactional or security properties for each method using annotations on the method. A session bean instance can be reused by many clients. Depending on whether the server maintains the bean's internal state between two clients, the session bean can be stateless or stateful. Depending on whether the bean is available to remote clients (i.e., clients outside of the current JVM for the server), the session bean can be local or remote. All these are configurable via standard annotations on the beans.

After you define a session bean, how does the client get access to it? As we discussed, the client does not create or destroy EJB3 components, it merely asks the server for a reference of an existing instance managed by the server. That is done via JNDI. In JBoss AS, the default local JNDI name for a session bean is dependent on the deployment packaging of the bean class.

Of course, you should change local to remote if the bean interface is annotated with @Remote and the bean is accessed from outside of the server it is deployed on. Below is the code snippet to get a reference of the MyBean bean in a web application (e.g., in a servlet or a JSF backing bean) packaged in myapp.ear, and then invoke a managed method.

try {
  InitialContext ctx = new InitialContext();
  MyBeanInt bean = (MyBeanInt) ctx.lookup("myapp/MyBean/local");
} catch (Exception e) {
  e.printStackTrace ();
}

... ...

String result = bean.doSomething("have fun", 1);

... ...

What the client gets from the JNDI is essentially a "stub" or "proxy" of the bean instance. When the client invokes a method, the proxy figures out how to route the request to the server and marshal together the response.

If you do not like the default JNDI names, you can always specify your own JNDI binding for any bean via the @LocalBinding annotation on the bean implementation class. The JNDI binding is always "local" under the java:comp/env/ space. For instance, the following bean class definition results in the bean instances available under JNDI name java:comp/env/MyService/MyOwnName.

@Stateless
@LocalBinding (jndiBinding="MyService/MyOwnName")
public class MyBean implements MyBeanInt {

  public String doSomething (String para1, int para2) {
    ... implement the logic ...
  } 
  
}