Composite application assembly and agile development are two themes that have emerged strongly in recent feature stories on Loosely Coupled. The closer we get to writing about actual customer implementations, the more these themes seem to recur. Is it a coincidence? Or our editorial bias? Perhaps it is the latter, but I suspect that even if it is, the bias will prove to have been prescient, and that these two themes will emerge as defining characteristics of successful web services deployments.
The progression is that web services are providing the infrastructure for service-oriented architecture, which enables enterprises to begin to assemble new composite applications by joining together elements of their existing software investments. The loose coupling inherent in the architecture allows them to do this rapidly in small, tactical projects, spurring the adoption of agile development techniques. Driving all of this is a desire to achieve greater business process agility, and there is a dawning recognition that this can only be achieved by unplugging rigid integrations and embracing the concept of continuous change in short, unplug and play.
Here's a sample of what people have been saying about composite applications and short, agile development cycles in recent articles:
Genient prefers to describe its approach as creating a "composite application" made up of the separate processes and functions, which automatically adapts to changes in the individual components. For companies looking to break the expensive and time-consuming technical cycle of application development and integration, adopting a BPM enterprise architecture could be the answer Radical promise of BPM, Mar 24th
According to Intalio's Ghalimi ... put [web services] in the context of a service oriented architecture, and it becomes a foundation to enable more agile process automation: "You're selling continuous process improvement, and XML and SOA becomes just another technical detail." Tactics, not strategy, drive SOA adoption, Apr 15th
Vinod Khosla spelt out his vision for iterative specification at December's Real-Time Enterprise conference: "I suggest planning on systems that are always wrong but can be changed very rapidly, that are adaptive. Get something running in 90 days, users try it, they tell you what's wrong, you can easily change it. Don't make it a massive transformation for the company. Get into a world of rapid modification as an architecture. Do a thousand 90-day projects as opposed to one three-year project. Each one, if you make a mistake, is not a large mistake because you haven't invested a lot into it." SIs adjust to living without integration, Apr 23rd
[Commerce One's] Singh argues that in this new era of composite applications, customers should focus not on what it takes to build an application, but on what it takes to change it. "They should plan on the first system being wrong," he says. "They will have to build a system that can be changed rapidly, so they can't expect it to be perfect from the get-go." He adds that short timescales are essential: "If it sounds like a three-year project, walk away it should be 90 to 120 days." Solving the 'last mile' challenge of B2B, May 6th
One of the most difficult concepts for technologists to get their heads around in a service-oriented world is that the world is no longer oriented around speed. Customers very rarely notice the 0.1 percent of extra performance you've squeezed out of the software by recompiling it, but if in the process you changed some aspect of the user interface without warning them, they'll be screaming blue murder.
As I've remarked in my ASPnews column this week, there's never been a shortage of startups coming along with new ways of speeding web applications and web services across the Internet, but virtually no one has given any thought to how providers are going to ensure they maintain reliable, consistent quality of service. AmberPoint's announcement this week of the company's Service Level Manager is a welcome exception.
One of the best case studies among AmberPoint's customers for this new product is AgentWare, described in eWeek's coverage. This is a company that either lives or dies by its ability to both consume and deliver services over the Web. It is an online travel information service, which collects inventory information about airline seats, hotel rooms and rental cars for syndication to various travel web sites. How on earth they managed to stay on top of service quality before deploying AmberPoint's product, I just cannot imagine.
The important point to realize about this product is that instead of being about feeds and speeds, it provides a means of managing and measuring the customer experience. That makes it of interest to business managers, because it elevates the discussion away from the nature of the technology infrastructure and towards delivering quantifiable business results. For all they care, the technology might as well be powered by a hamster running round on a treadwheel. What matters is that it works. Service Level Manager allows them to set realistic service standards and then stick to them, something that business managers value far more highly than setting world-beating speed records.
posted by Phil Wainewright 12:31 PM (GMT) | comments | link
Wednesday, May 07, 2003
Writing applications in PowerPoint
According to Carroll Pleasant of web services early adopter Eastman Chemical, using development tools from NextAxiom is "the closest thing to a Powerpoint-to-VB compiler that you're ever going to see." The quote comes from a good, detailed case study by Eric Parizo at searchWebServices. Eastman used the NextAxiom tools to complete development of a management scorecard application in just three days, compared to an expected month or more without it. "We wrote zero code to assemble the scorecard application," says Carroll. "We put the pieces together by drawing a picture of the application assembly in a GUI environment, and that's pretty cool."
"Some may still want to prevent managers from defining business processes themselves, saying it's too complex a job and should be left to specialists. That may be true right now, but it won't be by the week after next."
The article has some useful explanations of the concept of 'mobile systems' and how the mathematics of pi-calculus allows software to model constantly changing, loosely coupled systems. This may sound very dry, but it allows the authors to make a radical challenge to the status quo of software development: "Imagine a typical business application today and ask: 'Why does it stay the same? It's digital stuff, right? So why can't it change? Why does a new version of the application have to be developed for every situation? Why can't it adapt to me?' It's now legitimate to ask such questions."
So tell me, why can't I write applications in PowerPoint? You may think I'm joking, but I'm deadly serious. PowerPoint slideshows are the most structured documents in the Office suite, and if you add XML behind them, is it so absurd to imagine business managers using PowerPoint to assemble and modify business processes within composite applications? Maybe not right now, but as crazy as it may have sounded a few short months ago, it's already edging within the bounds of possibility. Give it a year or two, and this apparently ridiculous notion may well have become something we'll all be taking for granted.
posted by Phil Wainewright 12:42 PM (GMT) | comments | link
Tuesday, May 06, 2003
Tumbling towards agility
The horizontalization of enterprise computing proceeds apace. A spate of announcements from HP, EDS and Veritas, following on from earlier moves by IBM and Sun, gives the impression of an industry in a headlong rush to offer computing infrastructure as an on-demand service. It is as though everyone has suddenly realized that the creation of a standards-based, service-oriented application architecture renders differentiation redundant at the underlying systems layer. It doesn't matter what you buy any more, just so long as it's guaranteed to keep running.
Vendors are tying themselves in knots in their eagerness to demonstrate agility. EDS has launched an "agile infrastructure architecture" that provides "an industrial-strength agile infrastructure" designed to "improve clients' agility," according to its press release issued today. Apparently, it has implemented the architecture "faster than competitors" over a period of "more than two years," which strikes me as an illuminating notion of agility.
HP, meanwhile, unveiled its "Darwin Reference Architecture" in a none-too-subtle evocation of the fears many vendors must harbor about their future after the current phase of computing evolution has played out. As Darwin might have said, only the most adaptive will survive. HP's press release succinctly dismisses an entire generation of enterprise application software development to the dustbin of redundancy: "Business applications, such as SAP, PeopleSoft or Siebel, increasingly will be delivered as application services to business processes. For example, rather than being built redundantly into monolithic 'order entry' and CRM applications, a discrete business process like 'verify customer entitlement' may be delivered as an application service in both contexts."
The unseemly rush is justified if you believe a recent prediction from Forrester, quoted yesterday in an InfoWorld article about the current crop of utility computing announcements: "According to Forrester Research, four datacenter innovations web services, server provisioning, storage virtualization, and network route optimization will 'recast' datacenter architecture and allow customers to slice IT spending in half." On that analysis the pickings will be slim, even for the survivors.
posted by Phil Wainewright 10:24 AM (GMT) | comments | link
Assembling on-demand services to automate business, commerce, and the sharing of knowledge