When a complex system is finally deployed, the last thing that its financiers want to hear is that its costs have barely begun to take flight. The time to think about your next-generation system, though, is certainly no later than the day that the current state of the art goes live—and there are good reasons to think about technology succession even more than one generation in advance.
Big, expensive systems can seem like remote abstractions to people whose biggest purchase is a home or a car. I take these things more personally: My photo gallery includes, for example, a picture of me standing at Milepost 0 of the Trans-Alaska Pipeline System, whose tributary pipelines—sadly neglected by BP—have degenerated to a point that the entire field had to be shut down in early August for major maintenance.
Although its been more than 20 years since I visited TAPS for a technical inspection back in a previous career as an Exxon senior engineer, the current situation feels like watching a former home fall apart from abuse and inattention. System succession on the Alaska North Slope seems to have been inadequately considered, with cost containment as the inadequate excuse.
Also in my gallery is a photo of my oldest son on the bridge of a space shuttle orbiter during its preparation for flight: a souvenir of his internship at Kennedy Space Center. Even so, the shuttle is a 25-year-old implementation of a 35-year-old design. Like those Alaskan pipelines, its served its purpose longer than ever intended—and it shows.
At least when Alaskan oil pipelines leak or when shuttle tanks shed dangerous blocks of foam, they show their age in attention-getting ways. Obsolescence in the enterprise data center is much less obvious, but it does show itself in rising difficulty of finding personnel with yesterdays skills—or in a tilt in the wrong direction from creation of new value toward mere custodianship of current assets.
In an IT department thats being run as a node of business strategy, the question is never, "What can be done?" The question is, "Whats worth doing?" And the answer to that question is always changing as costs fall and capabilities rise.
The goal, though, should not be technology leadership as an end in itself but, rather, a consistent edge—one that only needs to be measured in months to be significant—over competitors. That means a lifestyle of conducting campaigns rather than building monuments.
You dont want to wait for the IT equivalent of spilled oil or falling foam to signal that its time for something better. You want to have nontechnical management see IT staff as always having an up-to-date plan for what will be useful to do next, at what cost and on what timetable.
No one outside the IT department is excited about Microsofts adoption of an intermediate representation for incrementally compiled code or about the Java platforms adoption of annotations. Youll find more interest, though, in the prospects of lowering the cost of application integration and transforming proven business logic into mix-and-match services, or in elevating developer productivity without becoming too tightly coupled to any one vendor-specific platform that narrows future options. Keep the focus on whats in it for the business, not on why its attractive to the technical staff.
The time to start thinking about Version N+1 of your IT stack is no later than the day that you get Version N up and running. Ideally, Version N+1 is already taking shape in your mind even while Version N is still being defined.
Speaking of taking things personally, I found myself thinking along these lines during a careful descent at the beginning of this month from Californias Mount Whitney. Many of the trails were more like ditches filled with running water or jagged rocks than nice smooth paths. If you look only one step ahead on a trail like this, youll often find yourself needing several awkward half-steps—or even a backward step—to get to a point where you can proceed easily.
You have to be looking and thinking two or three steps ahead to choose a next step that leads to the smoothest overall progress. Technology migration shares this characteristic. Step carefully, but step lively.
Technology Editor Peter Coffee can be reached at firstname.lastname@example.org.