Tuesday, January 31, 2006

Tibco RV vs Tibco EMS

A lot of people are familiar with the Tibco RV (Rendezvous) product. This product is around 20 years old and powers many mission critical systems across different domains. Tibco has another product called Tibco EMS (Enterprise messaging Server) which is based on the JMS hub-and-spoke model and is quite popular in EAI. So what's the difference btw the two products? - When to use what ?

First we need to understand the differences in the architecture between the two products. Tibco RV is based on the multicast publish/subscribe model, where as Tibco EMS is based on the hub-and-spoke model.

Multicast is the delivery of information to a group of destinations simultaneously using the most efficient strategy to deliver the messages over each link of the network only once and only create copies when the links to the destinations split. By comparison with multicast, conventional point-to-single-point delivery is called unicast. When unicast is used to deliver to several recipients, a copy of the data is sent from the sender to each recipient, resulting in inefficient and badly scalable duplication at the sending side.

Googled out a white paper that contained some good architecture details about Tibco RV. The link is here.
Excerpt from the article:
TIB/RV is based on a distributed architecture . An installation of TIB/RV resides on each host on the network.Hence it eliminates the bottlenecks and single points offailures could be handled. It allows programs to sendmessages in a reliable, certified and transactional manner,depending on the requirements. Messaging can be delivered in point-to-point or publish/subscribe, synchronously orasynchronously, locally delivered or sent via WAN or theInternet. Rendezvous messages are self-describing and platform independent.
RVD (Rendezvous Daemon) is a background process that sits in between the RVprogram and the network. It is responsible for the deliveryand the acquisition of messages, either in point-to-point or publish/subscribe message domain. It is the most important component of the whole TIB/RV.Theoretically, there is an installation of RVD on every host on the network. However, it is possible to connect to a remote daemon, which sacrifices performance and transparencies. The RV Sender program passes the message and destination topic to RVD. RVD then broadcasts this message using User Data Packet(UDP) to the entire network. All subscribing computers with RVDs on the network will receive this message. RVD will filter the messages which non-subscribers will not be notified of the message. Therefore only subscriber programs to the particular topic will get the messages.


So when to use Tibco RV and when to use Tibco EMS (or any other hub-and-spoke model). An article at http://eaiblueprint.com/3.0/?p=17 describes the trade-offs between the two models:

Excerpt:
In my opinion, the multicast-based publish/subscribe messaging is an excellent solution for near-real-time message dissemination when 1 to 'very-many' delivery capabilities matter. RV originated on the trading floors as a vehicle for disseminating financial information such as stock prices.

However, in most EAI cases the opposite requirements are true:
‘Cardinality’ of message delivery is 1-1, 1-2; 1 to-very-many is a rare case. With exception of ‘consolidated application’ integration model (near real time request reply with timeout heuristics), reliability of message delivery takes priority over performance. Effectively, in most EAI implementations of RV, RV ends up simulating a queueing system using its ‘Certified Delivery’ mechanism. While this works, it is a flawed solution (see the discussion of Certified Messaging below).

Reliable delivery is a ‘native’ function of hub-and-spoke solutions. The multicast solution must be augmented with local persistence mechanism and re-try mechanism.
While RV offers reliable delivery (queueing) referred to as Certified Messaging, this solution is flawed in that: Inherently, reliability of CM is not comparable to the hub-and-spoke topologies as the data is stored in local file systems using non-transactional disk operations, as opposed to centralized database in hub-and-spoke topology. Corruption of RV ‘ledger files’ is not a rare case that leads to loss of data.
Temporal de-coupling is not the case. While a message can be queue for a later delivery (in case the target system is unavailable), for a message to be actually delivered both system must be up at the same time.

RV / CM lacks a basic facility of any queueing system: a queue browser that is frequently required for production support (for example, in order to remove an offending message). In contrast, MQ Series offers an out-of the box queue browser and a host of third party solutions. Transactional messaging is difficult to implement in multicast environment. TIBCO offers a transactional augmentation of RV (RVTX) that guarantees a 1-and-onlu-once delivery; however, this solution essentially converts RV into a hub-and-spoke system. Consequently, very few RV implementations are transactional.

Having said that, the shortcomings of hub and spoke include:
  • Single point of failure, when the hub is down, everything is down. MQ Series addresses this problem by providing high-reliability clustering.
  • Non-native publish/subscriber (publish/subscribe is emulated programmatically ) that results in reduced performance, especially in 1-to-very many delivery
  • Overhead of hub management (a need for administering hub objects: queues, channels, etc).
  • Inferior performance, especially in 1 to-many publish/subscribe and request/reply cases.

These shortcomings — in my opinion — do not outweigh benefits and for that reason MQ Series / hub-and-spoke solutions constitute a better choice for most EAI problem. Multicast-based publish subscribe is better left to its niches (high volume, high performance, accepter unreliability, 1 to very many).

Monday, January 30, 2006

ATMI vs Corba OTS

Recently while working on BEA Tuxedo server, I came across the term 'ATMI' - Application to Transaction Monitor Interface. I beleive Tux was quite a popular product for handling large number of transactions in a distributed environment - right from the early days when most business critical applications were written in Cobol.
ATMI is a procedural, non object oriented API that you use to write stateless services to manage transactions. This API is developed by the OpenGroup and forms the core service API of Tuxedo. The Application-Transaction Monitor Interface (ATMI) provides the interface between the COBOL application and the transaction processing system. This interface is known as ATMI and these pages specify its COBOL language binding. It provides routines to open and close resources, manage transactions, manage record types, and invoke request/response and conversational service calls.

But I believe today's TP monitors are based on Corba OTS (Object Transaction service) semantics. Corba OTS is used by any OO application as the underlying semantics for communications with resouce managers in a distributed transaction. Even JTS is the Java langauge mapping of Corba OTS. A JTS transaction manager implements the Java mappings of the CORBA OTS 1.1 specification, specifically, the Java packages org.omg.CosTransactions and org.omg.CosTSPortability. A JTS transaction manager can propagate transaction contexts to another JTS transaction manager via IIOP.

Friday, January 27, 2006

Having both HTTP and HTTPS in a web-application

Web applications often need to have both http and https access; i.e. some URIs in the web application need to have HTTPS access and some can do with HTTP.

Hard-coding the URL address with https whereever required is not an elegant solution as this has to be reflected in all the links in all pages. An alternative soln will be to do a Response.redirect from the page with https in the URL. More detailed info is available at :
http://www.javaworld.com/javaworld/jw-02-2002/jw-0215-ssl.html

Also Struts has a extension through which we can specify which URLs need HTTPS access. For more info check out:
http://struts.apache.org/struts-taglib/ssl.html
http://sslext.sourceforge.net/

Java Service Wrapper

Came across this tool that can be used to run Java apps as a Windows Service or Unix deamon.
Includes fault correction software to automatically restart crashed or frozen JVMs. Critical when app is needed 24x7. Built for flexibility.

The following links provide more info:
http://sourceforge.net/projects/wrapper/ http://wrapper.tanukisoftware.org/doc/english/introduction.html

How to gaurantee "Once-and-only-once" delivery in JMS

In one application scenario, it was required to process the message only once. The message contained a donation amount that needed to be put in the database. Using a Queue and AUTO_ACK semantics, we were able to achieve this. But there was a problem - What would happen if the JMS provider fails before the acknowledgement is received from client. It would send the message again when it comes up. How to avoid this ?

The consumer may receive the message again, because when delivery is guaranteed, it's better to risk delivering a message twice than to risk losing it entirely. A redelivered message will have the JMSRedelivered flag set. A client application can check this flag by calling the getJMSRedelivered() method on the Message object. Only the most recent message received is subject to this ambiguity.

What are XA transactions? What is a XA datasource?

Came across this wonderful explanation by Mike regarding XA transactions here

Excerpt:
An XA transaction, in the most general terms, is a "global transaction" that may span multiple resources. A non-XA transaction always involves just one resource. An XA transaction involves a coordinating transaction manager, with one or more databases (or other resources, like JMS) all involved in a single global transaction. Non-XA transactions have no transaction coordinator, and a single resource is doing all its transaction work itself (this is sometimes called local transactions).

XA transactions come from the X/Open group specification on distributed, global transactions. JTA includes the X/Open XA spec, in modified form. Most stuff in the world is non-XA - a Servlet or EJB or plain old JDBC in a Java application talking to a single database. XA gets involved when you want to work with multiple resources - 2 or more databases, a database and a JMS connection, all of those plus maybe a JCA resource - all in a single transaction. In this scenario, you'll have an app server like Websphere or Weblogic or JBoss acting as the Transaction Manager, and your various resources (Oracle, Sybase, IBM MQ JMS, SAP, whatever) acting as transaction resources. Your code can then update/delete/publish/whatever across the many resources. When you say "commit", the results are commited across all of the resources. When you say "rollback", _everything_ is rolled back across all resources.

The Transaction Manager coordinates all of this through a protocol called Two Phase Commit (2PC). This protocol also has to be supported by the individual resources. In terms of datasources, an XA datasource is a data source that can participate in an XA global transaction. A non-XA datasource generally can't participate in a global transaction (sort of - some people implement what's called a "last participant" optimization that can let you do this for exactly one non-XA item).

Wednesday, January 25, 2006

Message Driven Beans Vs Stand-alone JMS clients

In many design scenarios, asynchronous messaging is required- for integration purposes, for handling large volume of requests etc.
In the J2EE scenario, we often have to decide whether to write a simple JMS client or go for a MDB in a EJB container. Though, a simple JMS client is neat and easy to write, it has some disadvantages:
JMS has no built-in mechanism to handle more than one incoming request at a time. To support concurrent requests, you will need to extend the JMS client to spawn multiple threads, or launch multiple JVM instances, each running the application.On the downside, we have issues of security, transaction handling, and scalability.

The main advantage that MDBs have over simple JMS message consumers is that the EJB container can instantiate multiple MDB instances to handle multiple messages simultaneously.
A message-driven bean (MDB) acts as a JMS message listener. MDBs are different from session beans and entity beans because they have no remote, remote home, local, or local home interfaces. They are similar to other bean types in that they have a bean implementation, they are defined in the ejb-jar.xml file, and they can take advantage of EJB features such as transactions, security, and lifecycle management.

As a full-fledged JMS client, MDBs can both send and receive messages asynchronously via a MOM server. As an enterprise bean, MDBs are managed by the container and declaratively configured by an EJB deployment descriptor.
For instance, a JMS client could send a message to an MDB (which is constantly online awaiting incoming messages), which in turn could access a session bean or a handful of entity beans. In this way, MDBs can be used as a sort of an asynchronous wrapper, providing access to business processes that could previously be accessed only via a synchronous RMI/IIOP call.
Because they are specifically designed as message consumers and yet are still managed by the EJB container, MDBs offer a tremendous advantage in terms of scalability. Because message beans are stateless and managed by the container, they can both send and receive messages concurrently (the container simply grabs another bean out of the pool). This, combined with the inherent scalability of EJB application servers, produces a very robust and scalable enterprise messaging solution.

They cannot be invoked in any manner other than via a JMS message. This means that they are ideally suited as message consumers, but not necessarily as message producers. Message-driven beans can certainly send messages, but only after first being invoked by an incoming request. Also, MDBs are currently designed to map to only a single JMS destination. They listen for messages on that destination only.

When a message-driven bean's transaction attribute is set to Required, then the message delivery from the JMS destination to the message-driven bean is part of the subsequent transactional work undertaken by the bean. By having the message-driven bean be part of a transaction, you ensure that message delivery takes place. If the subsequent transactional work that the message-driven bean starts does not go to completion, then, when the container rolls back that transactional work it also puts the message back in its destination so that it can be picked up later by another message-driven bean instance.

One limitation of message-driven beans compared to standard JMS listeners is that you can associate a given message-driven bean deployment with only one Queue or Topic
If your application requires a single JMS consumer to service messages from multiple Queues or Topics, you must use a standard JMS consumer, or deploy multiple message-driven bean classes.

Tuesday, January 17, 2006

On Screen Scraping

Recently I came across the term "Screen Scraping" when interfacing with a legacy application. I was not really sure what it mean in the context of EAI.. A google search came up with a good explanation at http://en.wikipedia.org/wiki/Screen_scraping

As a concrete example of a classic screen scraper, consider a hypothetical legacy system dating from the 1960s -- the dawn of computerized data processing. Computer to user interfaces from that era were often simply text-based dumb terminals which were not much more than virtual teleprinters. (Such systems are still in use today, for various reasons.) The desire to interface such a system to more modern systems is common. An elegant solution will often require things no longer available, such as source code, system documentation, APIs, and/or programmers with experience in a 45 year old computer system. In such cases, the only feasible solution may be to write a screen scraper which "pretends" to be a user at a terminal. The screen scraper might connect to the legacy system via Telnet, emulator the keystrokes needed to navigate the old user interface, process the resulting display output, extract the desired data, and pass it on to the modern system.