The Enterprise Service Bus: Integrating Core Back-End Applications With Web Services
The Enterprise Service Bus: Integrating Core Back-End Applications With Web Services
The Enterprise Service Bus: Integrating Core Back-End Applications With Web Services
Contents
Point-to-Point Integration 3
EAI and Messaging Middleware 3
Application Integration in today’s Web-Centric Environment 4
The IBM Enterprise Service Bus: One Size does not fit All 5
The Adoption Model and Self-Assessment 8
Bottom Line 10
There is nothing like the issue of application integration to remind us just how
complex the IT business really is, despite its immaturity. Commercial business
systems have existed for little more than 40 years, but in that short time we have
generated wave upon wave of – often incompatible – technology. Large enterprises
are typically built around islands of business function and information (silos, as
they have become known), which are not only technically dissimilar but cannot
even share the data and business logic they contain without considerable
‘integration’ effort.
It’s not difficult to see how this complexity has occurred. Constant pressure to
deliver improved business benefit has encouraged companies to explore new
development tools and techniques; new hardware platforms have brought with
them an ever broader selection of operating systems, databases, and software
packages; and the continuous process of company mergers and acquisitions has
forced IT departments to cope with an increasingly heterogeneous infrastructure.
What has become clear in recent years is that integrating dissimilar technologies
is invariably more cost-efficient than the ‘rip and replace’ approach. Large
companies have invested hundreds or thousands of man-years in developing
core business applications and tuning them to achieve optimal performance in a
mission-critical environment. Mainframe applications in particular (which in many
cases were developed in-house and still offer unparalleled performance and data
management characteristics) need to interact effectively with newer web-facing
applications without losing any of their inherent value to the business.
Before considering the products and concepts that are evolving to cope with the
new requirement for ‘secure flexibility’, let us first review how application integration
has evolved.
Point-to-Point Integration
Application integration really became an issue in the early 1990s, when large
companies first recognized the need to share resources between core centralized
systems and the newer breed of client/server technologies. Early point-to-point
integration solutions were often designed in-house, and were intended to cope
with a relatively limited number of data format translations and network protocol
conversions.
What characterized these early integration projects was that applications were
tightly coupled; once joined together, they were not easily put asunder.
Communication between the two applications involved was generally synchronous
in nature, and the interfaces at either end needed specific knowledge about the
data routing and translation processes involved. Any changes in this process
required considerable programming effort and technical expertise.
The solution to the problem came in two forms. With the arrival of MQSeries (now
WebSphere MQ) and message-oriented middleware in 1993, users had at their
disposal a backbone for application integration – a simple bus structure that offered
asynchronous communication. This meant that applications could be loosely
coupled: each app needed little or no knowledge about the data formats or protocols
supported by the other. It also meant that the delivery of dispatched messages
would be guaranteed by the backbone – once and once only. Now larger numbers
of applications could interact and exchange data – a process that often involved
complex sequences of actions – and the messaging infrastructure would manage
The important thing about these products was that they were non-intrusive. They
did not affect the application code itself, and provided a more generic mechanism
for exchanging data and events between networked systems.
In terms of flexibility and support, these EAI and early messaging products made
life a lot easier for the corporate integration specialist. They externalized the
integration logic from the application itself, and allowed changes to be made
relatively easily. But although many EAI functions were absorbed into the application
servers that came along afterwards, they didn’t offer the real automated ‘on the
fly’ capabilities demanded by today’s on-demand service-oriented e-business
environment. Moreover, in many cases they actually added to the complexity of
the infrastructure: companies often found that different EAI tools were deployed in
different parts of the organization and could not then be successfully integrated
with one another.
Not all Internet-led changes have been beneficial for the large enterprise. One of
the main problems is that many of the tools and technologies underpinning the
brave new world of web services are considerably less mature and functionally
capable than the enterprise products that they are intended to replace or with
which they interoperate.
Web applications often proliferate in the furthest corners of the extended enterprise,
well away from the disciplined data center environment, and the management
tools used to control and integrate them are far from ideal. Simple Network
Management Protocol, for example – the TCP/IP-based toolset for managing web-
based traffic – has taken many years even to approach the level of sophistication
of enterprise management tools such as NetView. Similarly FTP (File Transfer
Protocol), which was designed for nothing more than simple file transfer in the
non-commercial world, is increasingly being used as an integration tool between
secure, business-critical applications. And XML, which has revolutionized the way
that data is deployed within web services, is nevertheless very resource-intensive
and can seriously affect enterprise network performance.
Within the SOA environment, IT departments attempting to bridge the gap between
mature, high-performance enterprise applications and the new generation of flexible
but relatively unmanageable web services need a consistent approach to
integration; one which will allow them to focus on the relative performance, security,
and integrity characteristics of each business process and transaction under their
control.
There are many reasons why data centers need to achieve this level of consistency
between enterprise applications and web services. This is not simply a service
management consideration. In some cases the immediate driver is regulatory
compliance. With legislation such as Sarbanes-Oxley, Basel II, HIPAA, ISRS and
the COBIT Act imposing a high level of data and application transparency on
businesses across the world, technologies that contain little or no function for
ensuring the integrity of data need to be tightly controlled (particularly when the
data being managed moves outside the company boundary).
But the main reason for moving towards an SOA philosophy is to improve business
agility. Once the controls are in place to manage system resources, application
services and data on the fly, businesses are in a much stronger position to respond
quickly and efficiently to changing market conditions, competitive pressures, and
other commercial imperatives. For large, complex IT functions the SOA is likely to
be a long-term goal as it can involve considerable organizational change; but it is
a change that needs to be made. Companies that have already invested heavily
in application integration are particulary keen to make the transition, as they
appreciate the level of flexibility that a SOA can provide (and they know that their
competitors are heading in the same direction).
The IBM Enterprise Service Bus: One Size does not fit All
ESBs need to handle a very diverse range of traffic types, as they play a
fundamental part in achieving the integration consistency discussed earlier. Nimble,
lightweight low-cost web services are increasingly combined with business-critical,
highly available processes and transactions, built on other standards or even fully
customized and proprietary. With multi-hop considerations and bandwidth
HTTP JMS
WebSphere MQ
WebSphere Adapters
All ESBs are unique and, in view of the required diversity, ESB implementations
may not be built around a single product; they may need a combination of products
and services that interoperate to provide optimal routing, conversion and
interpretation services for each specific business task. IBM’s recently announced
ESB strategy distinguishes between ‘advanced ESB’ for combining mature
enterprise-level application messaging with web services traffic; and what we would
term ‘lightweight ESB’ for web services and peripheral network use. As shown in
Figure 1, IBM now has separate offerings to target these two specific needs – the
new WebSphere ESB and the more mature WebSphere Message Broker.
WebSphere ESB (WESB) provides connectivity and integration for web services,
and is designed for situations where flexibility, low-overheads and ease of operation
are high priorities. It offers a general-purpose web services backbone which is
likely to appeal to a broad range of new and existing customers.
IBM does not foresee WESB and WMB being used in isolation. Customers are
very likely to combine the products; the latter would sit at the centre of the enterprise,
typically close to centralized mainframe-based data and resources, and the ESB
would perform more web-service-based functions in distributed locations. The
two products share a common heritage and are both closely integrated with the
other components of the WebSphere family, so smooth interoperability is
guaranteed. Morover, both products can be used in conjunction with the
WebSphere Process Server, which works alongside the messaging tools, providing
higher-level monitoring and analysis of the way that traffic and events relate to
changing business processes.
There is, of course, a broader market for WebSphere ESB in smaller companies
supporting a relatively small number of standard web-oriented data formats. For
such businesses, the new product alone may be an attractive choice. However,
we anticipate most interest among the larger companies that have an immediate
need for ESB support and which will benefit from the distinct characteristics of the
two products.
For users SOA often requires a period of self-analysis, at the business level as
well as the IT level, when the customer examines its current portfolio and the
requirements of the business, and sets out a roadmap that will allow the IT function
to evolve into a service-driven infrastructure, closely aligned with the business.
Integrate Composite
Effectiveness
applications
Efficiency Interoperability
Connect
Tasks Connections
In response to this need, IBM has recently launched the WebSphere Business
Adoption Model (http://www.ibm.com/websphere/soa_assessment), which
incorporates an on-line SOA Self-Assessment. The model describes the stages
through which companies typically move as they start to adopt a service-oriented
approach to developing new business applications:
* Automate: At this level, businesses are putting automated tools and procedures
in place to align business and IT processes. This is the ‘effectiveness’ level,
where application integration should be sophisticated enough to contribute
substantially to corporate revenue growth and cost reduction.
Once the self-analysis is complete, they receive a report which outlines their
level of SOA sophistication and details the options available to them and the
benefits that can accrue.
Bottom Line
The emergence of the SOA and the Enterprise Service Bus are a direct response
to the integration challenges facing the enterprise. While we believe that the ESB
concept is still in the early stages of development, products such as WebSphere
ESB and Message Broker offer a very rich implementation of the concept as it
stands today. Enterprises that need to improve their level of IT responsiveness by
embracing web-based technologies (built around XML, SOAP, J2EE and .NET)
while maintaining and enhancing the value of their existing investments should
consider the ESB route very seriously.
This paper was sponsored by IBM. The author, Mark Lillycrop, is Chief Analyst
at Arcati Research.