Gartner middleware expert Massimo Pezzini has published a succinct briefing paper on how enterprises are using web services to encapsulate business processes for deployment within a service oriented architecture. It's available direct from Gartner for $95, or better still, you can get it for free by signing up at CNet's TechRepublic site.
Called Best Practices for Deploying Web Services via Integration, the paper does a great job of explaining how enterprises can bridge the gap between what they currently have and where they want to get to, introducing the concept of "pseudo-SOA" to describe an architecture where pre-existing software components are bundled up into "composite web services".
He writes that most current projects are limited in scope, consisting of only a few published web services, but expects that to change: "Users will therefore be challenged to set up composite web services implementation processes capable of scaling from small, opportunistic projects to large-scale, business-critical endeavors."
He goes on to set out a four-step approach to implementing a composite web services strategy. "The first step is the hardest," he writes. "Application architects, possibly jointly with business analysts, have to figure out which collection of services is required, their granularity and their interface." He recommends starting small, but notes that mature implementations tend to plateau at anywhere between 400 to 1,000 services.
Once the components have been built, they can then be assembled into coarser-grained business services. For this step, enterprises are looking for "... tools that can be used by less technically skilled and more business-oriented developers. Enterprises will increasingly adopt rules engines, programmatic integration servers, business process managers (BPMs), and integration brokers to assemble services requiring complex orchestration of large numbers of basic components."
posted by Phil Wainewright 3:07 AM (GMT) | comments | link
Wednesday, October 09, 2002
XDocs to bypass the browser
Microsoft fleshed out more of its web services strategy with announcements of new servers and desktop products this week, most notably with CEO Steve Ballmer's unveiling of a new technology for Office codenamed XDocs.
Speaking at Gartner's ITExpo event in Orlando today, Ballmer revealed how XDocs will allow Office users to retrieve and manipulate data retrieved from remote databases using a form-based interface. XDocs works with the XML features that are planned for Office 11, the next release of Microsoft's flagship desktop application suite. Its introduction will bring functionality that previously had to be accessed using a web browser directly into the Office environment, thus moving closer towards the company's long-term vision of a Universal Canvas founded on Microsoft technology.
Earlier this week at the company's Exchange conference in Anaheim, California, Microsoft had announced details of Jupiter, a project to unify its ecommerce, content management and BizTalk server products into a single platform incorporating business process automation and management capabilities. Jupiter will also include integration with Office 11 and with the Visual Studio.Net development platform.
Jupiter was a powerful statement of Microsoft's growing emphasis on business process automation, but it was XDocs that proved the most intriguing of the week's announcements. I was reminded of Dan Bricklin's recent analysis of the factors that either help or prevent users from taking advantage of software automation, Why Johnny can't program, in which he concluded that users find a form-based or dialog-driven metaphor one of the most accessible formats.
However, Microsoft will find some stiff opposition to this attempt at weaning users away from the browser back into Office, particularly from companies like Macromedia, whose Flash MX environment and associated technologies already brings many of the promised features of XDocs onto desktops via the browser, without the disadvantage of the 18-month or longer wait until XDocs and Office 11's XML features actually ship.
posted by Phil Wainewright 3:05 PM (GMT) | comments | link
Integration, development and reuse
The demarcation between application development and application integration is disappearing. "Every integration problem requires some development," pointed out Chet Kapoor, VP and general manager of BEA's Integration Group, quoted in a CRN news item last week. "The same thing with development projects, they require some integration. So why the chasm between the two?"
BEA has a fresh outlook on enterprise application integration (EAI), one that it hopes gives it a strong differentiation from companies that CTO Scott Dietzen was recently heard describing as "boutique integration vendors." At a briefing attended by Stencil Group analyst Brent Sleeper, who later published his notes online, Dietzen said that much of the complexity of EAI is caused by the way it's currently done, rather than being an inherent feature. BEA's line is that moving to a standards-based, interoperable model of EAI centered on application servers will get rid of a lot of the unnecessary overhead. Hence BEA's announcement, reported in the CRN story, of a complete application development and integration solution "that links simplified application development and EAI as if they are one requirement, not two separate needs."
BEA's new thinking is a welcome recognition of the move away from vertically integrated software stacks into an emerging horizontally integrated structure. Business automation is not something that anyone should be building from the ground up today. Much of the functionality that's needed for a new requirement is already automated, either as part of a legacy system or within readily available software packages and online services. This existing layer of application infrastructure simply has to be harnessed to fulfil the specific business process. Next-generation application development is therefore a matter of co-ordinating the reuse of existing application functionality. As time goes on, that co-ordination will itself increasingly be achieved through automated, standards-based processes, and the need to write new code, even for integration, will fade away.
BEA's latest partnerships seem to be well aligned to take advantage of this shift in emphasis. Its closer relationship with Siebel is related to that vendor's attempt to morph into an author of business process automation building blocks. Its recent tie-up with HP seems even cannier, as the newly merged computing goliath has adopted a strategy that is eminently suited to the emerging realities of the web services era. Quoted in NetworkWorld last week, HP Services chief Ann Livermore said the company's aim was to "be fabulous at the IT infrastructure designing and building and running it," while leaving business process in the hands of systems integration partners.
All of this still needs systems integrators and business consultants to get their heads round a vastly different approach to application development and integration projects than the one they're used to. Ted Shelton, chief strategy officer of Borland, interviewed by ZDNet, recently summed up what is needed: "I'm hopeful that a new generation of software architects will ultimately comprehend the value and applicability of the looser, message-based architecture and you'll see all sorts of new kinds of things emerge but that takes time. That takes a lot of time."
posted by Phil Wainewright 5:03 AM (GMT) | comments | link
Monday, October 07, 2002
Fear and loathing in web services
According to Gartner, the web services hype cycle passed its peak of inflated expectations a couple months ago, and will plunge headlong into the trough of disillusionment next year, emerging into mainstream acceptance only in 2004. A clear map is now emerging of the rocky road ahead:
There are no big-bucks web services projects in the pipeline. Instead, companies are using the technology to get more out of their existing IT infrastructure. General Motors, whose CTO Tony Scott delivered a keynote at last week's Internet World tradeshow, is a case in point. Here's what he told InternetNews.com: "Despite its $3 billion IT budget, GM is constrained to spending only 20 to 40 percent of it on new development, Scott explained, with the rest going to maintenance and upgrades of current systems ... with the advent of Web services, Scott said the company has turned its attention to linking up the company's IT systems." What's more, GM is currently deploying the technology only in pilots.
Vendors will war on higher-level standards. Setting out Gartner's hype cycle for web services in an article last month for .Net magazine, Gartner research director Mark Driver warned of the likelihood of bitter standards battles and looming fragmentation as vendors fight to recoup their investment in developing more sophisticated web services architectures: "As Web service APIs work their way up from low-level, horizontally focused (technical) foundation technologies to higher-level, domain-specific vertically oriented solutions, vendors will be less likely to share their innovations and more likely to overlap and compete for customer market share." (The article also has a useful diagram of the hype cycle timeline as it applies to web services).
Organizational politics will slow adoption. In a report published today, Jupiter Research predicts that business executives and information technologists will each fear a loss of control if an organization adopts web services technology. According to a SearchWebServices interview with the report's author: ".. an IT staff may feel its job security is threatened as a company's technology portfolio becomes more streamlined and the need for custom development decreases. Conversely ... executives may feel that their own success is tied to business processes that fall exclusively in the realm of IT departments."
Expect to see each of these pain points explored in the tech media in increasing detail as we move into next year stories of systems integrators shutting up web services practices in dismay at the shortage of lucrative contracts, customers blaming standards disarray for their refusal to move ahead with web services projects, and (in perverse contradiction to the other stories) angry CIOs complaining that implementing web services has undermined the size and influence of their empires.
Then, just at the moment when everybody hates web services so much they can't even bear to read about the technology any more, mainstream adoption will begin to build, unnoticed by the media as it moves on in search of other topics.
The hard truth is that web services is never going to be a get-rich-quick opportunity. The only rapid changes in fortune will be experienced by those vendors and solution providers who bulk up too fast in expectation of quick returns, and whose subsequent demise will add extra fuel to the negative mood. More in tune with today's sober economic environment, web services is going to be about sustaining profits rather than overnight wealth. Sensible players will slowly build their market presence and revenues, while customers will implement web services in projects where they can carefully measure incremental return on investment step by step. It may sound dull and dreary, but in the long run it will pay off.
posted by Phil Wainewright 3:45 AM (GMT) | comments | link
Assembling on-demand services to automate business, commerce, and the sharing of knowledge