I am energized by changes to the status quo and challenges to conventional wisdom. This is the most fascinating era I've encountered in 25 years of working in IT. I do see the dark side of the ever-quickening pace of change: frustration. Companies are frustrated by vendors that change their approach almost daily; how can IT plan ahead when that's the case? Vendors are frustrated by the market forces that make such adaptation necessary, and by the lag time between technology changes and the market's understanding of the changes' motives and results. In other words, even when a Sun Microsystems Inc., Microsoft Corp., or Commerce One Inc. can make its internal engineering machinery run faster -- no small feat -- it takes a maddeningly long time before customers grasp the benefits of the effort. So long, in fact, that vendors are forced to kick off a new round of reengineering before they're done teaching the current model to customers.
A brief discussion I had with Commerce One recently brought the issue into sharper focus. The company says it saw the handwriting on the wall a couple of years ago and started re-engineering its composite enterprise apps to be more open and cooperative. But they say that the past two to four months have brought unpredictable, tumultuous change to their space. Quite so; that's been true for all of IT. Six months ago, the priority was on adding Web services interfaces to existing applications, and on selling tools and connectors to make that work easier. Six months ago, we were in the midst of a shift from monolithic computing to a loosely coupled model that emphasized cramming maximum performance into minimum space. Six months ago, it was fashionable to break huge enterprise software stacks into discrete components that could be deployed independently.
The move to graft Web services interfaces onto existing applications grew out of businesses' powerful aversion to "rip and replace" re-engineering. Even a cursory look at Microsoft's .Net framework reveals the incredible attention paid to ensuring smooth interoperability between .Net and applications based on Microsoft's legacy paradigms such as COM and Visual Basic. Even so, the marketing tactic used most effectively against .Net is that its benefits are only obvious when applications are rewritten to use it.
The impact has been profound; .Net has achieved nothing near the success that the technology's cleverness deserves. As a result, Microsoft will have a hard time recovering enormous R&D costs. But just when Redmond was resigning itself to the new market reality, a new "new reality" has emerged. Some customers are telling major software and integration players that rip and replace might not be so bad if two things are true: The customer doesn't have to do the work, and the result is a more dynamic, adaptable solution. If a vendor can prove that replacement nets a quick return and lowers future engineering costs, it might win a sale. That's been Microsoft's message all along, but it has been forced to downplay that strategy because it sounded too much as if the Evil Empire was looking to take over every company's server room. Microsoft backed off its aggressive stance just when some big customers were ready to embrace it.
The pendulum swing between centralized and decentralized hardware usually takes years to complete, giving vendors plenty of time to adjust. But this time, the end of the swing was so abrupt that it caught everybody off guard except IBM Corp. Sun Microsystems Inc., Intel Corp., and Hewlett-Packard Co. were looking forward to another full year of brisk sales for low-profile rack mount servers. Advanced Micro Devices Inc. retooled its product schedule to ship its 64-bit server CPU before its desktop and mobile versions, even though the volume of Opteron systems will initially be poor.
Then over the past couple of months, customer requirements started to shift faster than the pendulum's natural pace. Sun's Greg Papadopoulos summed it up in January as the transition from the performance era to the throughput era. The raw computing power of an individual server node is now far less important than the speed and latency of connections between nodes. That's great news for makers of optical networking equipment but lousy news for CPU makers. The megahertz race that has defined the industry for so long has come to an unpredictably abrupt end. Think of what it does to Intel that its low-power mobile Pentium III may be more prized as a server processor than the Xeon.
Faced with the prospect of little light at the end of the budget tunnel, some IT organizations are seeing the value of outsourcing ever-bigger chunks of their operations. Sun embarked on a campaign to bring vendor lock-in to an end, using this rhetoric to target IBM and Microsoft. Sun admits that it, too, was guilty of selling customers into single-brand solutions. The reformed Sun took a bow in January, then strangely took a backward leap in February when Scott McNealy expressed two seemingly discordant ideas: First, that Sun's hardware and software solutions will interoperate with anything, even Windows, and second, that Sun will walk into a company and take over the entire IT operation. Can these approaches coexist? Whether they can is moot -- they'll have to.
Customers no longer speak with one voice, and rare moments of harmony don't last long anymore. One after another, companies are hitting huge strategic decision points that force them to do what they swore they'd never do. As strange as it is to watch and chronicle a vendor's struggle to adjust to rapidly changing conditions, the smartest thing it can do is stay flexible and avoid marrying itself to one approach or message. Until things stabilize (if they ever do), whatever the customer asks for, the answer has to be "yes."