Welcome!

SOA & WOA Authors: Adine Deford, Liz McMillan, Pat Romanski, Dana Gardner, Elizabeth White

Related Topics: SOA & WOA

SOA & WOA: Article

ESB Myth Busters: 10 Enterprise Service Bus Myths Debunked

Clarity of Definition for a Growing Phenomenon

Myth #2: Microsoft is building an ESB with their "Indigo" project.
Microsoft's Indigo combines Microsoft's Messaging Queuing, its Component Object Model COM+, .NET, and Web services. What they are building is a message bus with Web services extensions. This is very different from an enterprise bus. A messaging bus exposes the details of lower-level messaging techniques and requires the writing of code to define the relationships between applications and services. The mantra of the ESB is configuration rather than coding, which removes the necessity of hardwiring relationships between interconnected applications. An ESB facilitates and promotes the use of loose coupling between applications, which are exposed through the bus as event-driven services. The good news here is that applications built on Indigo will at least be message based and therefore easier to integrate through an ESB.

That said, there are elements of the BizTalk Server that, correctly combined with Indigo, could start to look like an ESB. However, there is one missing element in the formula, in that BizTalk is still a hub-and-spoke integration broker and is subject to all of the caveats mentioned in the ESB versus EAI discussion. You can't split out the XML transformation engine from the rest of the BizTalk Server and expect to run it as a load-balanced service across multiple machines without incurring the cost and overhead of the entire Biztalk Server (see the previous discussion on EAI vs. ESB).

Myth #3: The adoption of WS-* specifications, such as WS-Reliability and WS-Reliable Messaging, obviate the need for an ESB.
An ESB should be designed to accommodate these evolving specifications as they become mature and achieve commercial viability. Evolving WS-* specifications will help make application endpoints more interoperable through an ESB.

As part of the evolving standards process of Web services specifications, there exists much uncertainty due to the many overlapping efforts underway. As these specifications mature and achieve widespread adoption, they will still require an infrastructure to support them. An ESB can provide a consistent model for building, orchestrating, and managing SOAs, while insulating the IT organization from changes in underlying interoperability standards.

A WS-Reliability implementation requires that there be an industry-proven reliable message persistence and store-and-forward processor to support it. A foundational component of an ESB is an enterprise messaging layer that provides quality of service of message delivery through messaging conventions such as message persistence, store- and-forward delivery, message acknowledgements, and interfaces with external XA-compliant transaction managers. The ESB implementation may also provide transparent routing of messages across sophisticated network topologies, and continuous availability of the messaging infrastructure through a fault-tolerant messaging server architecture. The science of making all of this work together to ensure reliability under high-stress enterprise environments requires many person-years of effort to get right.

That being said, ESBs that are implemented today using a proprietary messaging layer should also adopt one or more of the WS-Rel* as additional protocols or "on-ramps" for getting on and off the bus. However, it is not a one-size-fits-all solution and many combinations of messaging and protocol support are necessary.

More Stories By Dave Chappell

David Chappell is vice president and chief technologist for SOA at Oracle Corporation, and is driving the vision for Oracle’s SOA on App Grid initiative.

Comments (6) View Comments

Share your thoughts on this story.

Add your comment
You must be signed in to add a comment. Sign-in | Register

In accordance with our Comment Policy, we encourage comments that are on topic, relevant and to-the-point. We will remove comments that include profanity, personal attacks, racial slurs, threats of violence, or other inappropriate material that violates our Terms and Conditions, and will block users who make repeated violations. We ask all readers to expect diversity of opinion and to treat one another with dignity and respect.


Most Recent Comments
Charlesy 07/03/13 07:28:00 AM EDT

An old post, but worth a small correction. Comparison with competitor products is always dangerous. You need to be very sure of your territory, and unfortunately, although David's description of BizTalk Server has some validity, it isn't 100% accurate. For example, the transformation services absolutely can be invoked separately to the rest of BTS via lightweight services, and load balanced across different boxes. More recent versions of BTS provide pre-built generic WCF and ASMX transformation web services as a courtesy to developers (reduces the need to build custom transformation services). You can use Windows Server AppFabric tools to create BizTalk maps in non-BizTalk projects - e.g., projects that define lightweight transformation services.

David doesn't clearly spell out what he means by 'cost...of the entire BizTalk Server'. If he means licencing cost, then he hits the mark, somewhat. BTS is certainly licensed in a fashion that encourages distribution and load balancing over a small farm of centralized servers, in contrast to the highly distributed approach advocated by Sonic. You can only invoke BTS transformation services on licensed boxes, so from that perspective, he is correct. However, invoking the transformation services does not require loading additional irrelevant BTS plumbing into memory. There is no heavy-weight performance cost imposed by code bloat, or anything like that!

BizTalk maps are emitted as code components containing an executable XSLT resource. You can distribute maps as freely as you wish and invoke the tranforms via code. Obviously, direct invocation in this case assumes the use of either .NET or Mono, although Java/.NET bridges could be used. If you use BizTalk Server's mapping tools to create maps, you may end up with a dependency on BizTalk-specific scripted components invoked in the XSLT, which ties you to licensed BizTalk Boxes. However, it is pretty easy to avoid this if you wish.

One thing Sonic has which BizTalk really does not is dynamic management of code deployment into the run-time environment. BizTalk Server has an built-in code repository, but this is simply a mechanism for managing and storing compiled artifacts and resources for the purpose of exporting installation packages. You still have to manually install those packages or deploy them via additional script. Frankly, though, this is rarely a significant drawback. The types of solution built using BizTalk Server tend to warrant close attention to managing dynamic deployment across a distributed environment using other frameworks and tools, and BizTalk Server plays well with the relevant frameworks. There is even a community-built deployment framework specifically designed for BizTalk Server.

yt67 03/03/05 07:09:34 AM EST

Myth-busting: always entertaining.

Jason 03/02/05 09:16:29 AM EST

A good read!

Javier Camara 02/10/05 04:19:02 AM EST

(This same feedback also posted to another WSJ article about ESBs)

I agree in that the ESB concept is over-hyped. For me, a SOA makes sense if it is viewed as a constellation of web services interacting among them. For this, something like a UDDI server is required for each service locating each other.

For me, all this (i.e. services + directory) is just enough if only synchronous communications are used. If asynchronous communications are needed, then you need also publish/subscribe and store-and-forward, i.e. roughly what a MOM does. You can call it an ESB if you want, although I think this concept in the market encompasses several roles:
1. Publish/subscribe to messages
2. Store-and-forward messages
3. Route messages
4. Transform messages

An interesting thing to note is to implement points 1. and 2. you do *not* need business logic, while to implement 3. and 4. you do.

As I said, I see roles 1 and 2 required in SOAs with asynchronous interactions.

Roles 3 and 4 are also needed in many cases, mainly for integrating disparate systems. However, my main point against an ESB is that, in order to perform these roles, you do *NOT* need of a new, special concept like the ESB. *Any* service in the constellation of services can perform both routing and transformation. It can range from being a single component like an ESB (which I think is a bad idea), or it can just be a set of services (e.g. a different service performing specific adaptation for a system being integrated).

For me, using a single ESB for 3. and 4. breaks the beauty of the SOA idea. You are supposed to made all your data and business logic of your organization available as services in order to be reused, and suddenly you put on top an ESB in which you put *more* business logic (routing and transformation). So my point is that this should be implemented just by means of regular services, and not by specific, central-piece new components called ESBs.

Now, if for implementing routing and transformation you want to use Tibco, WebSphere or whatever, fine - however, the logic created by these products should be at the same level as the other services in the SOA, and not above.

So I am not saying that orchestrating tools are not useful. They are. Only, they are not *imprescindible*; and at any rate they should be viewed just as more services in the SOA. However, this does not fit the marketing strategy of ESB vendors which show its ESB as an *enabler* of a SOA, instead of just one more *component* of it.

Dave Chappell 02/03/05 09:54:43 PM EST

We (Sonic Software) didn't re-lable our product to support the ESB wave, we actually invented the concept. We then worked with the analyst and journalist community to help create industry awareness of the new concepts that are introduced by ESB, which has resulted in a whole new product category.

I would agree with you that there is a great deal of hype right now due to lack of understanding of what ESB is, which is compounded by the number of traditional middleware and EAI vendors who have clamoured to get ESB in their marketing literature without having a full understanding of what it means to have an ESB. Your comment about middleware with new clothes is well taken. You might get that impression depending on where you learned about what an ESB is. That is exactly what I am trying to point out with myth #1 in this article.

A certain amount of hype is expected when a technology category begins to take hold and gain traction within serious IT projects. This can be disruptive to the industry as a whole. This is also the primary reason why I wrote the OReilly book on the subject of ESB--to act as the definitive guide to help educate and provide clarity on what makes up an ESB. Please don't shoot the concept of ESB down until you have had a chance to understand it.
Dave

Larry 02/03/05 04:15:19 AM EST

Not surprising that the representative of a company who over-hyped ESB in the first place, and relabeled their own product ESB to catch the service wave, should now try to claim that anyone who saw through the hype is guilty of spreading myths.
ESB is just the middleware emporor's new clothes.