I'll bite, but I won't call them DSBs. :) This all feels a lot like Software Fortresses to me. In his book by that name, Roger Sessions defined it as "a conglomerate of software systems serving a common purpose and typically owned by a cohesive group of individuals. These software systems work together in a tight trust relationship to provide consistent and meaningful functionality to a hostile outside world." While the term "hostile" may be a bit strong, it is essentially about establishing ownership domains.

He goes on: "The walls of the fortress has are designed to prevent any communications from coming into the fortress except through approved channels. The approved channels are the drawbridges."

So, if you subscribe to this construct, you could easily have several fortresses within an enterprise, such as along lines of business. These fortresses thus become domains. If each domain has the authority to do whatever they want inside their walls (which, by the purist view of fortresses would be the case), then you could have several ESBs that exist. An ESB clearly could be the drawbridge for a fortress, and for a fortress that has many services inside, you could even have internal ESB usage as well. Certainly, a common pattern seen in enterprises is to have one (logical) gateway (whether it's an ESB, XML appliance, whatever) for external consumers, and another (logical) gateway for internal consumers.

While I'll admit that there are cases where multiple internal fortresses may make sense (for example, where compliance law mandates a separation), I tend to prefer more centralized or minimally, federated approaches. This doesn't prevent an enterprise from having multiple domains, however, they are not fortresses in the purist sense. There would still only be a single, logical service bus/ fabric/network, but implemented in a federated fashion. What's interesting, though, is that the infrastructure to do this federation is not implemented by ESB technology, it's implemented by TCP/IP and DNS. If you have multiple domains within your enterprise, each one of those is going to have its own IPs, such as services.hr.foobar.com and services.sales.foobar.com. It's the hostname that will get it to the right "drawbridge." If you're doing registry/repository lookups to get that endpoint hostname, you may have registry/repository federation (just as we have with DNS), but not ESB federation.

Pointing to the telco comment, my thoughts when I saw that quote were simply that whoever that companies was needs some governance around their technology spending. It doesn't make sense for a company to have 5 different ESBs, and I assume that they meant 5 ESBs from 5 different vendors. Depending on their operational structure, they could have 5 installations of one ESB based upon operational management domains or deployment architecture. Unless they recently purchased 5 smaller companies, each of which brought their own ESB with them, someone in IT should have been stepping up and saying, "we're all looking for the same capabilities, let's buy one rather than five."

-tb

On Apr 5, 2007, at 7:59 AM, Gervas Douglas wrote:

No, it is not necessarily the favourite acronym of a dyslexic Aston
Martin owner.  Here is an extract from an e-mail Miko kindly sent me:

"I'm seeing increased sophistication around lifecycle and federation
issues in Governance and some interesting insights about whether the
Enterprise Service Bus is truly a ubiquitous pattern or whether we are
seeing the "Domain Service Bus" (DSB anyone?) winning the day. I
visited one telco that said, "we like ESBs so much that we have five
different ones!" This of course begs the question — which one is
actually the " Enterprise " Service Bus. The answer? None of them!!!
Five ESB products and no ESB. This is much more common than you would
think. I'm happy to chat about this observation if you're curious."

I guess there is a natural hierarchy of confusion with these buses.
Would someone like to take up Miko´s invitation to chat about DSBs in
this forum?

Gervas




Yahoo! Groups Links


[EMAIL PROTECTED]



Reply via email to