Spring features integration classes for remoting support using various technologies. The remoting support eases the development of remote-enabled services, implemented by your usual (Spring) POJOs. Currently, Spring supports four remoting technologies:
Remote Method Invocation (RMI). Through the use of the RmiProxyFactoryBean and the RmiServiceExporter Spring supports both traditional RMI (with java.rmi.Remote interfaces and java.rmi.RemoteException) and transparent remoting via RMI invokers (with any Java interface).
Spring's HTTP invoker. Spring provides a special remoting strategy which allows for Java serialization via HTTP, supporting any Java interface (just like the RMI invoker). The corresponding support classes are HttpInvokerProxyFactoryBean and HttpInvokerServiceExporter.
Hessian. By using the HessianProxyFactoryBean and the HessianServiceExporter you can transparently expose your services using the lightweight binary HTTP-based protocol provided by Caucho.
Burlap. Burlap is Caucho's XML-based alternative for Hessian. Spring provides support classes such as BurlapProxyFactoryBean and BurlapServiceExporter.
JAX RPC. Spring provides remoting support for web services via JAX-RPC.
JMS. Remoting using JMS as the underlying protocol is supported via the JmsInvokerServiceExporter and JmsInvokerProxyFactoryBean classes.
While discussing the remoting capabilities of Spring, we'll use the following domain model and corresponding services:
public class Account implements Serializable{ private String name; public String getName(); public void setName(String name) { this.name = name; } }
public interface AccountService { public void insertAccount(Account account); public List getAccounts(String name); }
public interface RemoteAccountService extends Remote { public void insertAccount(Account account) throws RemoteException; public List getAccounts(String name) throws RemoteException; }
// the implementation doing nothing at the moment public class AccountServiceImpl implements AccountService { public void insertAccount(Account acc) { // do something... } public List getAccounts(String name) { // do something... } }
We will start exposing the service to a remote client by using RMI and talk a bit about the drawbacks of using RMI. We'll then continue to show an example using Hessian as the protocol.
Using Spring's support for RMI, you can transparently expose your services through the RMI infrastructure. After having this set up, you basically have a configuration similar to remote EJBs, except for the fact that there is no standard support for security context propagation or remote transaction propagation. Spring does provide hooks for such additional invocation context when using the RMI invoker, so you can for example plug in security frameworks or custom security credentials here.
Using the RmiServiceExporter, we can expose the interface of our AccountService object as RMI object. The interface can be accessed by using RmiProxyFactoryBean, or via plain RMI in case of a traditional RMI service. The RmiServiceExporter explicitly supports the exposing of any non-RMI services via RMI invokers.
Of course, we first have to set up our service in the Spring container:
<bean id="accountService" class="example.AccountServiceImpl">
<!-- any additional properties, maybe a DAO? -->
</bean>
Next we'll have to expose our service using the RmiServiceExporter:
<bean class="org.springframework.remoting.rmi.RmiServiceExporter"> <!-- does not necessarily have to be the same name as the bean to be exported --> <property name="serviceName" value="AccountService"/> <property name="service" ref="accountService"/> <property name="serviceInterface" value="example.AccountService"/> <!-- defaults to 1099 --> <property name="registryPort" value="1199"/> </bean>
As you can see, we're overriding the port for the RMI registry. Often, your application server also maintains an RMI registry and it is wise to not interfere with that one. Furthermore, the service name is used to bind the service under. So right now, the service will be bound at 'rmi://HOST:1199/AccountService'. We'll use the URL later on to link in the service at the client side.
Note | |
---|---|
The servicePort property has been omitted (it defaults to 0). This means that an anonymous port will be used to communicate with the service. |
Our client is a simple object using the AccountService to manage accounts:
public class SimpleObject { private AccountService accountService; public void setAccountService(AccountService accountService) { this.accountService = accountService; } }
To link in the service on the client, we'll create a separate Spring container, containing the simple object and the service linking configuration bits:
<bean class="example.SimpleObject"> <property name="accountService" ref="accountService"/> </bean> <bean id="accountService" class="org.springframework.remoting.rmi.RmiProxyFactoryBean"> <property name="serviceUrl" value="rmi://HOST:1199/AccountService"/> <property name="serviceInterface" value="example.AccountService"/> </bean>
That's all we need to do to support the remote account service on the client. Spring will transparently create an invoker and remotely enable the account service through the RmiServiceExporter. At the client we're linking it in using the RmiProxyFactoryBean.
Hessian offers a binary HTTP-based remoting protocol. It is developed by Caucho and more information about Hessian itself can be found at http://www.caucho.com.
Hessian communicates via HTTP and does so using a custom servlet. Using Spring's DispatcherServlet principles, you can easily wire up such a servlet exposing your services. First we'll have to create a new servlet in your application (this an excerpt from 'web.xml'):
<servlet> <servlet-name>remoting</servlet-name> <servlet-class>org.springframework.web.servlet.DispatcherServlet</servlet-class> <load-on-startup>1</load-on-startup> </servlet> <servlet-mapping> <servlet-name>remoting</servlet-name> <url-pattern>/remoting/*</url-pattern> </servlet-mapping>
You're probably familiar with Spring's DispatcherServlet principles and if so, you know that now you'll have to create a Spring container configuration resource named 'remoting-servlet.xml' (after the name of your servlet) in the 'WEB-INF' directory. The application context will be used in the next section.
In the newly created application context called remoting-servlet.xml, we'll create a HessianServiceExporter exporting your services:
<bean id="accountService" class="example.AccountServiceImpl">
<!-- any additional properties, maybe a DAO? -->
</bean>
<bean name="/AccountService" class="org.springframework.remoting.caucho.HessianServiceExporter">
<property name="service" ref="accountService"/>
<property name="serviceInterface" value="example.AccountService"/>
</bean>
Now we're ready to link in the service at the client. No explicit handler mapping is specified, mapping request URLs onto services, so BeanNameUrlHandlerMapping will be used: hence, the service will be exported at the URL indicated through its bean name: 'http://HOST:8080/remoting/AccountService'.
Using the we can link in the service at the client. The same principles apply as with the RMI example. We'll create a separate bean factory or application context and mention the following beans where the SimpleObject is using the AccountService to manage accounts:
<bean class="example.SimpleObject"> <property name="accountService" ref="accountService"/> </bean> <bean id="accountService" class="org.springframework.remoting.caucho.HessianProxyFactoryBean"> <property name="serviceUrl" value="http://remotehost:8080/AccountService"/> <property name="serviceInterface" value="example.AccountService"/> </bean>
We won't discuss Burlap, the XML-based equivalent of Hessian, in detail here, since it is configured and set up in exactly the same way as the Hessian variant explained above. Just replace the word Hessian with Burlap and you're all set to go.
One of the advantages of Hessian and Burlap is that we can easily apply HTTP basic authentication, because both protocols are HTTP-based. Your normal HTTP server security mechanism can easily be applied through using the web.xml security features, for example. Usually, you don't use per-user security credentials here, but rather shared credentials defined at the Hessian/BurlapProxyFactoryBean level (similar to a JDBC DataSource).
<bean class="org.springframework.web.servlet.handler.BeanNameUrlHandlerMapping"> <property name="interceptors"> <list> <ref bean="authorizationInterceptor"/> </list> </property> </bean> <bean id="authorizationInterceptor" class="org.springframework.web.servlet.handler.UserRoleAuthorizationInterceptor"> <property name="authorizedRoles"> <list> <value>administrator</value> <value>operator</value> </list> </property> </bean>
This an example where we explicitly mention the BeanNameUrlHandlerMapping and set an interceptor allowing only administrators and operators to call the beans mentioned in this application context.
Note | |
---|---|
Of course, this example doesn't show a flexible kind of security infrastructure. For more options as far as security is concerned, have a look at the Acegi Security System for Spring, to be found at http://acegisecurity.sourceforge.net. |
As opposed to Burlap and Hessian, which are both lightweight protocols using their own slim serialization mechanisms, Spring Http invokers use the standard Java serialization mechanism to expose services through HTTP. This has a huge advantage if your arguments and return types are complex types that cannot be serialized using the serialization mechanisms Hessian and Burlap use (refer to the next section for more considerations when choosing a remoting technology).
Under the hood, Spring uses either the standard facilities provided by J2SE to perform HTTP calls or Commons HttpClient. Use the latter if you need more advanced and easy-to-use functionality. Refer to jakarta.apache.org/commons/httpclient for more info.
Setting up the HTTP invoker infrastructure for a service objects much resembles the way you would do using Hessian or Burlap. Just as Hessian support provides the HessianServiceExporter, Spring's HttpInvoker support provides the org.springframework.remoting.httpinvoker.HttpInvokerServiceExporter. To expose the AccountService (mentioned above), the following configuration needs to be in place:
<bean name="/AccountService" class="org.springframework.remoting.httpinvoker.HttpInvokerServiceExporter"> <property name="service" ref="accountService"/> <property name="serviceInterface" value="example.AccountService"/> </bean>
Again, linking in the service from the client much resembles the way you would do it when using Hessian or Burlap. Using a proxy, Spring will be able to translate your calls to HTTP POST requests to the URL pointing to the exported service.
<bean id="httpInvokerProxy" class="org.springframework.remoting.httpinvoker.HttpInvokerProxyFactoryBean"> <property name="serviceUrl" value="http://remotehost:8080/AccountService"/> <property name="serviceInterface" value="example.AccountService"/> </bean>
As mentioned before, you can choose what HTTP client you want to use. By default, the HttpInvokerProxy uses the J2SE HTTP functionality, but you can also use the Commons HttpClient by setting the httpInvokerRequestExecutor property:
<property name="httpInvokerRequestExecutor"> <bean class="org.springframework.remoting.httpinvoker.CommonsHttpInvokerRequestExecutor"/> </property>
Spring has support for:
Exposing services using JAX-RPC
Accessing web services
In addition to the support listed above, you can also expose your web services using XFire, a lightweight SOAP library, currently in development at Codehaus.
Spring has a convenience base class for JAX-RPC servlet endpoint implementations - ServletEndpointSupport. To expose our AccountService we extend Spring's ServletEndpointSupport class and implement our business logic here, usually delegating the call to the business layer.
/**
* JAX-RPC compliant RemoteAccountService implementation that simply delegates
* to the AccountService implementation in the root web application context.
*
* This wrapper class is necessary because JAX-RPC requires working with
* RMI interfaces. If an existing service needs to be exported, a wrapper that
* extends ServletEndpointSupport for simple application context access is
* the simplest JAX-RPC compliant way.
*
* This is the class registered with the server-side JAX-RPC implementation.
* In the case of Axis, this happens in "server-config.wsdd" respectively via
* deployment calls. The Web Service tool manages the life-cycle of instances
* of this class: A Spring application context can just be accessed here.
*/
public class AccountServiceEndpoint extends ServletEndpointSupport implements RemoteAccountService {
private AccountService biz;
protected void onInit() {
this.biz = (AccountService) getWebApplicationContext().getBean("accountService");
}
public void insertAccount(Account acc) throws RemoteException {
biz.insertAccount(acc);
}
public Account[] getAccounts(String name) throws RemoteException {
return biz.getAccounts(name);
}
}
Our AccountServletEndpoint needs to run in the same web application as the Spring context to allow for access to Spring's facilities. In case of Axis, copy the AxisServlet definition into your 'web.xml', and set up the endpoint in 'server-config.wsdd' (or use the deploy tool). See the sample application JPetStore where the OrderService is exposed as a Web Service using Axis.
Spring has two factory beans to create web service proxies, namely LocalJaxRpcServiceFactoryBean and JaxRpcPortProxyFactoryBean. The former can only return a JAX-RPC service class for us to work with. The latter is the full fledged version that can return a proxy that implements our business service interface. In this example we use the latter to create a proxy for the AccountService endpoint we exposed in the previous paragraph. You will see that Spring has great support for web services requiring little coding efforts - most of the setup is done in the Spring configuration file as usual:
<bean id="accountWebService" class="org.springframework.remoting.jaxrpc.JaxRpcPortProxyFactoryBean"> <property name="serviceInterface" value="example.RemoteAccountService"/> <property name="wsdlDocumentUrl" value="http://localhost:8080/account/services/accountService?WSDL"/> <property name="namespaceUri" value="http://localhost:8080/account/services/accountService"/> <property name="serviceName" value="AccountService"/> <property name="portName" value="AccountPort"/> </bean>
Where serviceInterface is our remote business interface the clients will use. wsdlDocumentUrl is the URL for the WSDL file. Spring needs this a startup time to create the JAX-RPC Service. namespaceUri corresponds to the targetNamespace in the .wsdl file. serviceName corresponds to the serivce name in the .wsdl file. portName corresponds to the port name in the .wsdl file.
Accessing the Web Service is now very easy as we have a bean factory for it that will expose it as RemoteAccountService interface. We can wire this up in Spring:
<bean id="client" class="example.AccountClientImpl"> ... <property name="service" ref="accountWebService"/> </bean>
From the client code we can access the web service just as if it was a normal class, except that it throws RemoteException.
public class AccountClientImpl {
private RemoteAccountService service;
public void setService(RemoteAccountService service) {
this.service = service;
}
public void foo() {
try {
service.insertAccount(...);
} catch (RemoteException ex) {
// ouch
}
}
}
We can get rid of the checked RemoteException since Spring supports automatic conversion to its corresponding unchecked RemoteException. This requires that we provide a non-RMI interface also. Our configuration is now:
<bean id="accountWebService" class="org.springframework.remoting.jaxrpc.JaxRpcPortProxyFactoryBean"> <property name="serviceInterface" value="example.AccountService"/> <property name="portInterface" value="example.RemoteAccountService"/> </bean>
Where serviceInterface is changed to our non RMI interface. Our RMI interface is now defined using the property portInterface. Our client code can now avoid handling java.rmi.RemoteException:
public class AccountClientImpl { private AccountService service; public void setService(AccountService service) { this.service = service; } public void foo() { service.insertAccount(...); } }
To transfer complex objects over the wire such as Account we must register bean mappings on the client side.
Note | |
---|---|
On the server side using Axis registering bean mappings is usually done in the 'server-config.wsdd' file. |
We will use Axis to register bean mappings on the client side. To do this we need to register the bean mappings programmatically:
public class AxisPortProxyFactoryBean extends JaxRpcPortProxyFactoryBean { protected void postProcessJaxRpcService(Service service) { TypeMappingRegistry registry = service.getTypeMappingRegistry(); TypeMapping mapping = registry.createTypeMapping(); registerBeanMapping(mapping, Account.class, "Account"); registry.register("http://schemas.xmlsoap.org/soap/encoding/", mapping); } protected void registerBeanMapping(TypeMapping mapping, Class type, String name) { QName qName = new QName("http://localhost:8080/account/services/accountService", name); mapping.register(type, qName, new BeanSerializerFactory(type, qName), new BeanDeserializerFactory(type, qName)); } }
In this section we will register our own javax.rpc.xml.handler.Handler to the web service proxy where we can do custom code before the SOAP message is sent over the wire. The Handler is a callback interface. There is a convenience base class provided in jaxrpc.jar, namely javax.rpc.xml.handler.GenericHandler that we will extend:
public class AccountHandler extends GenericHandler {
public QName[] getHeaders() {
return null;
}
public boolean handleRequest(MessageContext context) {
SOAPMessageContext smc = (SOAPMessageContext) context;
SOAPMessage msg = smc.getMessage();
try {
SOAPEnvelope envelope = msg.getSOAPPart().getEnvelope();
SOAPHeader header = envelope.getHeader();
// ...
} catch (SOAPException e) {
throw new JAXRPCException(e);
}
return true;
}
}
What we need to do now is to register our AccountHandler to JAX-RPC Service so it would invoke handleRequest(..) before the message is sent over the wire. Spring has at this time of writing no declarative support for registering handlers, so we must use the programmatic approach. However Spring has made it very easy for us to do this as we can override the postProcessJaxRpcService(..) method that is designed for this:
public class AccountHandlerJaxRpcPortProxyFactoryBean extends JaxRpcPortProxyFactoryBean { protected void postProcessJaxRpcService(Service service) { QName port = new QName(this.getNamespaceUri(), this.getPortName()); List list = service.getHandlerRegistry().getHandlerChain(port); list.add(new HandlerInfo(AccountHandler.class, null, null)); logger.info("Registered JAX-RPC Handler [" + AccountHandler.class.getName() + "] on port " + port); } }
The last thing we must remember to do is to change the Spring configuration to use our factory bean:
<bean id="accountWebService" class="example.AccountHandlerJaxRpcPortProxyFactoryBean">
<!-- ... -->
</bean>
XFire is a lightweight SOAP library, hosted by Codehaus. Exposing XFire is done using a XFire context that shipping with XFire itself in combination with a RemoteExporter-style bean you have to add to your WebApplicationContext. As with all methods that allow you to expose service, you have to create a DispatcherServlet with a corresponding WebApplicationContext containing the services you will be exposing:
<servlet> <servlet-name>xfire</servlet-name> <servlet-class> org.springframework.web.servlet.DispatcherServlet </servlet-class> </servlet>
You also have to link in the XFire configuration. This is done by adding a context file to the contextConfigLocations context parameter picked up by the ContextLoaderListener (or ContextLoaderServlet for that matter).
<context-param> <param-name>contextConfigLocation</param-name> <param-value> classpath:org/codehaus/xfire/spring/xfire.xml </param-value> </context-param> <listener> <listener-class> org.springframework.web.context.ContextLoaderListener </listener-class> </listener>
After you added a servlet mapping (mapping /* to the XFire servlet declared above) you only have to add one extra bean to expose the service using XFire. Add for example the following configuration in your 'xfire-servlet.xml' file:
<beans>
<bean name="/Echo" class="org.codehaus.xfire.spring.XFireExporter">
<property name="service" ref="echo">
<property name="serviceInterface" value="org.codehaus.xfire.spring.Echo"/>
<property name="serviceBuilder" ref="xfire.serviceBuilder"/>
<!-- the XFire bean is wired up in the xfire.xml file you've linked in earlier -->
<property name="xfire" ref="xfire"/>
</bean>
<bean id="echo" class="org.codehaus.xfire.spring.EchoImpl"/>
</beans>
XFire handles the rest. It introspects your service interface and generates a WSDL from it. Parts of this documentation have been taken from the XFire site. For more detailed information on XFire Spring integration, have a look at the docs.codehaus.org/display/XFIRE/Spring.
It is also possible to expose services transparently using JMS as the underlying communication protocol. The JMS remoting support in the Spring Framework is pretty basic - it sends and receives on the same thread and in the same non-transactional Session, and as such throughput will be very implementation dependent.
The following interface is used on both the server and the client side.
package com.foo; public interface CheckingAccountService { void cancelAccount(Long accountId); }
The following simple implementation of the above interface is used on the server-side.
package com.foo; public class SimpleCheckingAccountService implements CheckingAccountService { public void cancelAccount(Long accountId) { System.out.println("Cancelling account [" + accountId + "]"); } }
This configuration file contains the JMS-infrastructure beans that are shared on both the client and server.
<?xml version="1.0" encoding="UTF-8"?> <beans xmlns="http://www.springframework.org/schema/beans" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.springframework.org/schema/beans http://www.springframework.org/schema/beans/spring-beans-2.0.xsd"> <bean id="connectionFactory" class="org.apache.activemq.ActiveMQConnectionFactory"> <property name="brokerURL" value="tcp://ep-t43:61616"/> </bean> <bean id="queue" class="org.apache.activemq.command.ActiveMQQueue"> <constructor-arg value="mmm"/> </bean> </beans>
On the server, you just need to expose the service object using the JmsInvokerServiceExporter.
<?xml version="1.0" encoding="UTF-8"?> <beans xmlns="http://www.springframework.org/schema/beans" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.springframework.org/schema/beans http://www.springframework.org/schema/beans/spring-beans-2.0.xsd"> <bean id="checkingAccountService" class="org.springframework.jms.remoting.JmsInvokerServiceExporter"> <property name="serviceInterface" value="com.foo.CheckingAccountService"/> <property name="service"> <bean class="com.foo.SimpleCheckingAccountService"/> </property> </bean> <bean class="org.springframework.jms.listener.SimpleMessageListenerContainer"> <property name="connectionFactory" ref="connectionFactory"/> <property name="destination" ref="queue"/> <property name="concurrentConsumers" value="3"/> <property name="messageListener" ref="checkingAccountService"/> </bean> </beans>
package com.foo; import org.springframework.context.support.ClassPathXmlApplicationContext; public class Server { public static void main(String[] args) throws Exception { new ClassPathXmlApplicationContext(new String[]{"com/foo/server.xml", "com/foo/jms.xml"}); } }
The client merely needs to create a client-side proxy that will implement the agreed upon interface (CheckingAccountService). The resulting object created off the back of the following bean definition can be injected into other client side objects, and the proxy will take care of forwarding the call to the server-side object via JMS.
<?xml version="1.0" encoding="UTF-8"?> <beans xmlns="http://www.springframework.org/schema/beans" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.springframework.org/schema/beans http://www.springframework.org/schema/beans/spring-beans-2.0.xsd"> <bean id="checkingAccountService" class="org.springframework.jms.remoting.JmsInvokerProxyFactoryBean"> <property name="serviceInterface" value="com.foo.CheckingAccountService"/> <property name="connectionFactory" ref="connectionFactory"/> <property name="queue" ref="queue"/> </bean> </beans>
package com.foo; import org.springframework.context.ApplicationContext; import org.springframework.context.support.ClassPathXmlApplicationContext; public class Client { public static void main(String[] args) throws Exception { ApplicationContext ctx = new ClassPathXmlApplicationContext( new String[]{"com/foo/client.xml", "com/foo/jms.xml"}); CheckingAccountService service = (CheckingAccountService) ctx.getBean("checkingAccountService"); service.cancelAccount(new Long(10)); } }
You may also wish to investigate the support provided by the Lingo project, which (to quote the homepage blurb) “... is a lightweight POJO based remoting and messaging library based on the Spring Framework's remoting libraries which extends it to support JMS.”
The main reason why auto-detection of implemented interfaces does not occur for remote interfaces is to avoid opening too many doors to remote callers. The target object might implement internal callback interfaces like InitializingBean or DisposableBean which one would not want to expose to callers.
Offering a proxy with all interfaces implemented by the target usually does not matter in the local case. But when exporting a remote service, you should expose a specific service interface, with specific operations intended for remote usage. Besides internal callback interfaces, the target might implement multiple business interfaces, with just one of them intended for remote exposure. For these reasons, we require such a service interface to be specified.
This is a trade-off between configuration convenience and the risk of accidental exposure of internal methods. Always specifying a service interface is not too much effort, and puts you on the safe side regarding controlled exposure of specific methods.
Each and every technology presented here has its drawbacks. You should carefully consider you needs, the services your exposing and the objects you'll be sending over the wire when choosing a technology.
When using RMI, it's not possible to access the objects through the HTTP protocol, unless you're tunneling the RMI traffic. RMI is a fairly heavy-weight protocol in that it support full-object serialization which is important when using a complex data model that needs serialization over the wire. However, RMI-JRMP is tied to Java clients: It is a Java-to-Java remoting solution.
Spring's HTTP invoker is a good choice if you need HTTP-based remoting but also rely on Java serialization. It shares the basic infrastructure with RMI invokers, just using HTTP as transport. Note that HTTP invokers are not only limited to Java-to-Java remoting but also to Spring on both the client and server side. (The latter also applies to Spring's RMI invoker for non-RMI interfaces.)
Hessian and/or Burlap might provide significant value when operating in a heterogeneous environment, because they explicitly allow for non-Java clients. However, non-Java support is still limited. Known issues include the serialization of Hibernate objects in combination with lazily-initialized collections. If you have such a data model, consider using RMI or HTTP invokers instead of Hessian.
JMS can be useful for providing clusters of services and allowing the JMS broker to take care of load balancing, discovery and auto-failover. By default Java serialization is used when using JMS remoting but the JMS provider could use a different mechanism for the wire formatting, such as XStream to allow servers to be implemented in other technologies.
Last but not least, EJB has an advantage over RMI in that it supports standard role-based authentication and authorization and remote transaction propagation. It is possible to get RMI invokers or HTTP invokers to support security context propagation as well, although this is not provided by core Spring: There are just appropriate hooks for plugging in third-party or custom solutions here.