Cracks in the software monolith

The Oracle 11i e-business suite is now the designated whipping boy, slammed by Forbes, eWEEK, and CNet's News.com for misleading marketing, swarms of bugs, and consulting fees that balloon well beyond the price of the software itself.

Well, guess what? None of these complaints are anything new--and they don't apply to Oracle alone. SAP, PeopleSoft, J.D. Edwards, Baan, and other vendors of end-to-end enterprise software have endured similar criticisms over the years.

These vendors and a handful of others epitomize the "monolithic" approach to enterprise computing: Roll out a complete e-business solution from one software company, the argument goes, and you can be up and running faster without having to tie disparate systems together. And when things go wrong, you know whom to call: either the monolith or its certified consultancy.

Oracle's recent flaying merely illustrates the flaws inherent in such huge, monolithic software packages--and shows that the times are finally catching up with these dinosaurs.

With every revision, salespeople promise you anything and software developers push buggy code out the door, all to ring up those quarterly earnings. Like it or not, you have to retrain staff across the company and customize the software to fit your existing business processes--expect consultants to pitch camp as they make sure everything works properly. Finally, when everything runs smoothly, out comes yet another revision--and the process begins again.

Still, many enterprises prefer the predictability of this treadmill to the alternative--the "integrated" approach: Choose best-of-breed solutions and hire consultants to put them all together for you, piece by piece. Given the choice between dealing with one enterprise software vendor and dozens, IT managers often choose the monolith. The bugs will be fixed, eventually, and while no single software company can possibly create best-of-breed applications for every e-business discipline, at least you won't be saddled with training and support for a tangled matrix of solutions.

Viewed in this practical light, the monolithic argument sounds plausible--but the problem is in upgrading. In large companies, switching to a new monolithic enterprise package all at once may mean training thousands of users and upgrading scores of servers, which can result in disruption bordering on chaos. Making the shift incrementally--and centrally supporting two different version of the same software--can be even more unwieldy.

Much has changed since the monolithic approach ruled in the 1990s, and these changes--both in business and in IT--will put the monolithic approach to enterprise software on the road to extinction:

• The Internet has made integrating diverse systems practical, while weakening the rationale for a single, overarching software solution.

• New XML-based integration standards--most promisingly, those that describe the industrywide Web services initiative--will foster component architectures that result in reusable code rather than wasteful, one-off integration efforts.

• Businesses now need to share data with each other outside the firewall--a process that is inherently less flexible with a monolithic system than with one integrated from the ground up.

• For competitive reasons, companies must now tailor customers' setups to a degree that may not be supported by a monolithic system.

With a one-size-fits-all e-business suite, which must err on the side of completeness, consultants often spend most of their time "customizing" by turning off irrelevant features rather than adding functionality. That means you're paying for the development time that went into those features--whether you use them or not.

It's always easier to stick with the devil you know--and many corporations will stick with monolithic solutions such as Oracle 11i or mySAP. Yet if IT is more important than ever to business success, then technology needs to map to business processes as closely as possible. Companies can do this most effectively by buying, customizing, and integrating best-of-breed breed applications as they need them, rather than waiting for a monolithic software company's next revision.

No enterprise can simply scrap its large-scale back-office systems--there's simply no viable alternative when it comes to core financials. However, both SAP and Oracle have recently been forced to open their APIs, offering the opportunity to integrate best-of-breed solutions into that core. Add them over time, as needed--a custom portal solution from Bowstreet, say, or an application server from BEA Systems--and put off upgrading that monolith. Someday, the beast will shrink to the point that it loses its grip on your IT infrastructure, and you'll be able to upgrade it as you would any other component software, without all the disruption.

Eric Knorr is a an award-winning freelance writer and consultant who frequently works with e-business consultancy Envivid Solutions in San Francisco. He is the founding editor of CNET's Computers.com.