Is SOA decentralization the "kiss of death"?

Some people never tire of pointing out that SOA is nothing new -- merely old wine in new bottles. Fair enough. Even Michael Liebow, vice president of Web Services and SOA at IBM Global Services, would agree. As he puts it in a recent interview in Network World, "It's safe to say for the last 10 to 15 years there has been a vision of this future that has been hard to do, hard to realize. It's not something that has just been cooked up by somebody in some back room. There has been an effort that has been going on to horizontally integrate companies to provide for varying amounts of re-use, more flexible IT architectures to support business requirements."

So what is new? As Liebow explains, "Now we have a set of basic standards that allow for the discovery, description, communication, cataloging and securing of messages that allow applications to talk to one another...There is native support for these standards in products going off the shelf."

He thinks SOA is now moving to the next level. "What we are seeing now is a significant step up by CIOs to get ahead of this, to standardize, to provide the governance as to how, where and when you do SOA in the organization," he says. "The business unit adoption is not really SOA, it's something I would call SOI, or service-oriented integration. It's exposing some applications to being connected; it has great business value but really low real technical value. It's a way to sub-optimize; you will start to get reduced returns in the future because they create more havoc in the enterprise architecture. While it's a good short-term - you get a nice little bang - you don't get the benefit long term. CIOs need to jump ahead of this because they will end up with just hot spots in a chaotic environment with no hope of being able to control it. In a decentralized environment, it's the kiss of death."

Liebow describes IBM and Microsoft working together to create an SOA management layer that stands above particular products. "From a management standpoint, what we are talking about is messages," he explains. "One customer has 300 business services in production being leveraged by 70 applications, and he is running at about three to five million messages per day. So forget about spam; forget about e-mail. Once you start to get this stuff in production - the volume of activity that will need to be managed, the kind of bandwidth and the processing you'll need - it can be overwhelming. It's very important that it be designed correctly. You can design this stuff and it will work, but if you design it poorly, if you have bloated messages running around, at that scale it will come back to haunt you."  

Obviously, IBM and Microsoft benefit if companies move beyond decentralized deployments and acts of localized experimentation to contemplate large scale efforts. The questions are: Will CIOs embrace corporate-wide SOA plans and will they have the power to implement them? Of course, some people think decentralized experimentation is a great way to drive innovation and initiative. So here's the final question one might ask:  Is decentralization really the kiss of death?