Fleshing Out SOA

As we boarded the plane for our return from IBM Impact SOA conference last week, we couldn’t help but think about the emerging verticalization of SOA. That is, outside of place like Wall Street, which tens to roll its own systems, that there is relatively little appetite for raw technology frameworks and that customers will want domain-relevant templates from which they could layer in their added IP.

We noted last week that verticalization of SOA would set up a new feeding frenzy pitting traditional application providers, middle tier application composers (such as Business Process Management), and systems integrators against one another.

A follow up discussion for Dana Gardner’s Briefings Direct podcast last Friday confirmed our thoughts – although panelists were divided as to ultimate impact. Panelist Jim Kobelius termed this as “hugely disruptive” to the application vendor ecosystem, Joe McKendrick characterized IBM’s aggressive stance in developing vertical SOA as “taking a lesson from Thomas Watson” in positioning itself as business, rather than computing consultants.

Meanwhile David Linthicum countered that this was “not a seismic shift,” in that the apps world is evolving toward composite apps, and that consulting organizations like IBM’s Global Business Services have long developed frameworks that they could repurpose for different engagements. Nonetheless, GBS excluded, Linthicum added that the Accentures of the world have largely been blindsided and are playing catchup ball when it comes to developing vertical SOA frameworks.

The question eventually boiled down to who would own the innovation that comes when you develop a service framework that covers the business processes in your sector, because that’s where you develop the processes that actually run your business. Todd Biske ventured that it would be dangerous for any organization to outsource its process innovation, either to a consulting firm that would develop the extensions, or to a software vendor that would productize it.

Of course, we’ve been down this path before. Enterprises faced that issue as they frantically reengineered their business processes to fit SAP during the run-up to Y2K. Yet as organizations BPR’ed themselves to death, they still wound up with millions of dollars in customization bills. On this go round we still expect plenty of customization, but hopefully the flexibility of SOA will mean that you won’t have to spend millions to do so.

That’s where Gardner brought up an interesting alternative. Maybe the route is to find a mechanism for open sourcing the IP covering the vertical extensions to SOA, that result in assets such as component business or process models. Our take is that the only way that will happen is if vertical industry groups, such as RosettaNet, pick up the slack, and that something like the Apache license (where you don’t have to donate your customizations back to the community) be used for distributing content.

We’re Glad IBM’s Not in the Applications Business

2006 was the year of SOA governance and 2007 is the point when SOA gets owned by the business. That was supposed to be the take-home from IBM’s Impact SOA conference this year.

All right, what do you expect? It’s a marketing message, and of course, there’s still plenty of SOA governance left to do. But as time and technology moves on, IBM says it’s time to get the business driving SOA.

Let’s clarify that. A business analyst is not going to ask for a services-oriented architecture, they are going to ask for a solution to a B2B integration, a partner management, or sales force automation bottleneck. But if you listen to IBM, the solutions may in fact be based on SOA. But please don’t call them applications – hold that thought for a moment.

IBM’s betting on that the business will drive SOA is based on several assumptions. First, that services can more effectively align the things that software does with what the business actually needs. The flexibility of loosely-coupled services that can be orchestrated or blended into a composite app means that IT should be better able to keep pace with the business’s changing needs.

Second, and more importantly, IBM is betting that the balance of power in deciding which projects get done, and how they will be done, is shifting over to the business. They pointed to the emergence of a pivotal new role, the business architect, with deep domain expertise plus tech savvy, who would help steer technology decisions, plus the fact that more lines of business are formally or informally beefing up their IT expertise.

Yet, at this point IBM’s message gets a bit confusing. They say that the line of business is driving more of the technology decision, yet they also maintain that the most successful SOA initiatives occur when you have a strong IT group whose enterprise architecture role is more than advisory. That leaves us with a bit of a riddle, and for technology providers like IBM, a need to sort out to whom they actually sell SOA.

Let’s get beyond the who buys SOA to what they will be buying. IBM says the key to successful SOA is to verticalize it. That is, don’t make it a raw technology buy where the IT organization must construct the services from scratch, but one that has frameworks that spell out the core services, templates for building or assembling those services … or more.

So IBM is gradually rolling out vertical service content, starting with frameworks that specify generic business processes, component business models spelling them out, so-called “content packs” coming from its Webify acquisition (the product is now called WebSphere Business Services Fabric), plus more targeted content that would be developed by its Global Business services (GBS) consulting group (the operation that IBM acquired from PwC).

Of course, those of us in the room hearing this had a field day bombarding the poor GBS guy on why this isn’t a new breed of software application. Two things to remember: first, customers embraced packaged enterprise apps back in the 80s and 90s because they didn’t want to write yet another homegrown accounting system. And second, that IBM declared a decade ago that it was no longer in the apps business – it would instead help customers implement or run their apps.

Now, in GBS’s defense, none of this is terribly new. Integrators have always sought to build frameworks – and promote the fact that they have them – to convey the message that customers can get ready-made solutions configured to their needs much faster than if every effort were from scratch.

GBS says it’s a consulting firm, but when you look at the history of software vendors, many of them started out as consulting firms who productized a solution that they had developed for a particular client. In GBS’s case, the business model would be turning that over to Webify.

Ratcheting up the argument, there’s a great difference between CORBA components (or C++ libraries) that make the service-based frameworks that GBS and IBM are developing look more like apps. It comes down to the fact that, if implemented as web services, they can be far more readily cobbled into whatever you want them to be, and in all likelihood, are likely to be much higher level software assets than what the older, less connectable component technologies provided. So when you look at that claims processing service that GBS is proposing (atop Webify), at what point does this become a claims application?

There’s more to this issue than plain semantics.

Maybe IBM is targeting virgin territory not covered by the SAPs or Oracles of the world – there are certainly many sectors that never bought into ERP. But even in ERP territory, there’s plenty of waterfront real estate still available. And that’s where things get interesting.

Take the example of a BPO (business process outsourcing) provider that performs claims administration for multiple insurance companies. It uses SAP for accounting, but relies on an IBM SOA framework for claims processing. Then the company wants to expose a payment service to its insurance clients that consumes functionality from the IBM and SAP sides. With that functionality residing in the middle tier, it sets up an interesting showdown that in the end will dictate whether IBM sells more WebSphere or SAP sells NetWeaver. Keeping in mind the fact that most of the growth in the ERP space is occurring at the edges, where you extend the ERP system, this is where the next battle for dominance of the enterprise market will be fought.

A Tale of Two Industries

It’s been an interesting week for sure, as we’ve split our time between two ends of the software industry. First, we hung out with investors and CEOs of software firms at Sand Hill’s Software 2007 conference. Then we camped out in a fleabag hotel, forgot to shave, and hung out with developers at JavaOne. This week, we got to have it both ways.

We saw a tale of two industries that was joined by a common theme. Customers are feeling freer to mess around – and even buy – software again, and investors are coming back in. “The VC community is finding higher value companies, so there shouldn’t be any more bubble effect,” said MR Rangaswami, in remarks opening Software 2007. And this year’s JavaOne, which was otherwise quite uneventful, had its fullest turnout in years. Companies have given their developers travel budgets again.

At the macro economic level, it’s a tale of cyclical spending. The bubble was followed by the bust. But whatever goes down must go up –- at some point. IT spending is a lagging indicator -– business must be well into an up cycle before they feel hunger pangs for modernizing their systems to cope with growth. And so, as there are questions on whether the 4 – 5 year old recovery from the bust is now flattening out, coming to a soft landing, or hitting a momentary pause, IT budgets are this point still on the upswing as a reaction to last year’s growth. (We’ve seen the same phenomenon in our consulting business, with vendors having accumulated a backlog of marketing projects.)

On the customer side, it’s safer to invest, or at least to kick the tires, and because of the new pricing structures brought on by open source and Software-as-a-Service, buying, subscribing, or simply evaluating new software no longer requires major capital commitments. That’s prompted legacy players — like IBM, Microsoft, and Oracle to offer free kick-the-tires “Express” editions of their products. And you’ve got vendors that incorporate freebie open source pieces so they can better focus on adding spot innovation, rather than having to reinvent the wheel.

In some cases, this makes markets downright boring. We ended the week at JavaOne, where aside from Sun’s useless announcement of yet another rich Internet client framework (we’ll get to that in a moment), we saw little if any news. Or, as our friend Bill Roth of BEA put it, “JavaOne never ceases to amaze me. Year after year, I expect the show to be a flop. This year is no exception. I am wrong again this year.” Roth was referring to the show’s turnout, which was probably the largest since the days of Sun being the dot in dot com.

Sometimes no news is good news, if it means that vendors and customers are concentrating on implementing product, rather than blindly spewing out new features for new features sake. Or if you listen to Sun, there’s no news because “we’re delivering what we promised last year.”

But, as we hinted above, Sun just can’t avoid painting itself into a corner. We’re speaking of JavaFX, its just-announced answer to Adobe Flex and Microsoft Silverlight for a market where, for now, commonly used Ajax just seems good enough. Instead, Sun reminded us that at heart their engineers are still driving their marketing decisions, believing that neither Adobe nor Microsoft’s Web 2.0 solutions are good enough. And they also reminded us that they still don’t know how or when to partner.

OK, it wouldn’t make sense for Sun to embrace Microsoft’s Silverlight, which is rooted in the .NET framework. But come on, there’s nothing threatening or antithetical to Sun -– or Java –- that’s posed by Adobe’s Flex framework. We hate to get primitive, but let’s face it: Taking an enemy-of-thy-enemy approach would at least allow Sun to ride, rather than fight a wave for a change. That’s because Flex is based on Flash, which is ubiquitous, rather than Silverlight, which is not. Nope, Sun seems doomed to repeat past history, digging itself into yet another NetBeans hole.

Sun isn’t the only player that’s frustrating us by repeating past behaviors. We saw Steve Ballmer back at the Sand Hill conference mention on one hand how Microsoft is leveraging its huge Office advantage with renewed effort to get third parties to hook into it. We saw some cool demos where Dassault Systemes, an engineering software supplier, plug visual 3-D CAD models into Word and Excel documents to more effectively communicate mundane processes, such as pinpointing which compliments in an aircraft fuel pump must get fixed as part of a work order.

But we don’t know if Microsoft is curing its annoying habit of adding useless innovation into its products, as we remarked in a post a few weeks back. When asked after his speech whether the phasing in of Microsoft’s Live hosted software services might prompt it to plug innovation in shorter, more iterative doses that might better respond to what customers really want, Ballmer said that Microsoft would not necessarily change its software development models.

“If anybody thinks future is only innovations are done in short time are worth doing is wrong minded,” he said. On the face of it, we can’t disagree, but we’d hope that developers in Redmond might get out of their bubble a bit more often.

Well, actually, maybe we’re being a bit harsh here, as old habits die hard. You can’t go cold turkey from a software development model. We had a discussion with a director of “disruptive innovation” from a major household e-commerce brand, who discussed the necessity of mixing up software development methodologies to suit the requirements of the particular project. But in his case, his organization is dealing with only a 10-year legacy of development, rather than the 20 – 30 years common among companies like Microsoft -– or financial services companies on Wall Street for that matter.

The SaaS Conundrum

Excuse the travelogue this week, but we’re headed for a schizoid few days. We’re starting the week at Sand Hill Group’s Software 2007 conference, where you hang out with the people who buy, sell, and run software businesses. We’ll end the week at JavaOne, where you find the geeks who develop the software on which many of these businesses run.

So we’re expecting that we’re going to get a couple different slices of reality.

One area is open source. It’s gotten rather scant play at Software 2007, where the prevailing mood among the enterprise apps folks is that open source is fine for commodity infrastructure pieces like Linux or Lucene, an open source Java-based text search engine managed by Apache, but not when it comes to higher value business logic. Put another way, most enterprise vendors couldn’t see any way to make money off open source. Significantly, when we tried attending “The Poster children of Open Source,” those poster kids – who included Marten Miklos of MySQL and Larry Augustine or VA Software, were nowhere to be seen.

Somehow, we’re expecting to hear different answers when we travel up the 101 to Moscone Center to JavaOne later this week. But of course, there you’re dealing with tools, which have become low cost commodities anyway.

But back to Software 2007, the dominant topic appeared to be Software-as-a-Service (SaaS). It’s not the first time that the topic has been broached, and, by little surprise, when SAP announced that its next generation ERP for small-midsize businesses would be SMB based, it became impossible to avoid the topic.

At first glance, SaaS looks awfully attractive to software vendors, who can iterate new versions more rapidly and uniformly throughout th installed base, and for customers who obviously don’t have to worry about infrastructure.

Rubbing it in, Salesforce’s Mark Benioff showed a slide of the company’s new business incubation center for App Exchange providers, which happens to be housed in Siebel’s former San Mateo headquarters.

For software vendors, the hang-up over SaaS isn’t necessarily the technology, but the pricing model. Like open source offerings, the prevalent pricing model for SaaS is monthly or annual subscriptions sans the upfront fee. According to software pricing consultant Jim Geisman, a lot of vendors appear to be kicking the tires, but there’s a dichotomy between startups and established players. It’s like the airlines, where startups like JetBlue have lower cost structures than incumbents like United or American, who remained burdened with costlier legacy structures.

What was more interesting was the lukewarm response from CIOs. A panel, comprising CIOs from FedEx, Unilever, Motorola, and Disney dealt with SaaS as one of their hot buttons. Excluding Disney, which is a large Salesforce customer, the party line was, essentially, it will work in scattered areas. “It’s not a panacea,” noted Neil Cameron, CIO of Unilever, referring to data integration issues, but adding that SaaS made sense in selected “entry level” activities, such as deploying CRM in infrastructure-poor regions like Africa.

But most CIOs were concerned over the fact that SaaS could add yet another data silo to enterprise architecture.

SaaS clearly isn’t disappearing anytime soon. While SaaS could provide some quick hit solutions, vendors should start tempering their hype.

For instance, One SaaS startup told us that you could dispatch with data integration issues through data federation tools. And of course, you could integrate processes between islands of SaaS software in your organizations through Service-Oriented Architectures (SOA). But, just because SaaS providers represent a break (or a return to what we used to call time-sharing), there’s one thing they shouldn’t forget:: The software industry has never been at a loss for solutions, but saying that you have one doesn’t always prove it’s workable.