Today's IT challenges are defined as much by what's already in place as by what may be missing. It would be a great relief to start anew every so often and build computing infrastructure from the ground up, but that's rarely possible in an ongoing business environment. Just about every midsize and large-scale company still runs on legacy software. Often decades old, these systems may be technically unfashionable, but they nonetheless continue to provide necessary functionality.
How did we get to this point? The current state of legacy systems is the predictable outcome of software's evolution. Once every decade or so, the enterprise software market tilts dramatically, and dominant players and products lose traction. As they seemingly fall off the edge of the market, new players and products rise to prominence.
What's needed are midtier products that leverage the value of existing systems, and several new approaches are getting closer to this goal. But underpinning the technology is a new, smarter way to deconstruct legacy systems into elemental components and recompose them into an unimaginably vast set of new application services.
Legacy systems are here to stay, yet most have become increasingly expensive to manage and maintain—as much as 70% of the annual IT budget in some cases. Costs go beyond just running these systems, to keeping pace with upgrades, managing the complexity of increasingly diverse sets of applications, and greater pressure on CIOs to meet new expectations of accountability.
The result is a familiar—but unacceptable—problem to any CIO: growing project backlogs and expanding system complexity as code changes cascade through myriad applications and data structures.
Several new approaches will get you beyond this point. As the search begins for best practices for 2004, forward-looking CIOs are finding ways to improve their systems' leverage and capabilities without falling into the trap of legacy integration. These best practices are based on three intersecting activities in the middle tier—between the ubiquitous Web interface and legacy infrastructure. They are:
Componentization
Virtualization
Orchestration
Componentization will be a boon for software vendors and will finally validate the well-deserved role of C-level IT executives as chief value creators. Virtualization, or what's increasingly referred to as a service-oriented architecture (SOA), eliminates redundant systems by consolidating common functions. And service orchestration develops adaptive integration points across business processes.
Each practice will be explained in this article, but first, you may wonder what happened to the most recent solution, enterprise application integration.
EAI was supposed to address all these issues, but it has fallen short of expectations despite billion of dollars invested over the past five years in software and hard coding. In fact, EAI has itself become a legacy issue. While the strategic importance of integration has grown significantly, few CIOs can brag about the flexibility of their EAI infrastructures.
Both point-to-point and intermediated broker-based integration—the two common designs of EAI technology—have presented the same problems as the legacy applications they were intended to integrate. In the face of EAI overhead and a new IT backlog, it's become clear that integration in itself is not a solution; indeed, if misapplied, it's a new problem.
Maximizing the value of new capabilities and existing applications requires more than simply the veneer of a Web interface covering an otherwise unmanageable morass of spaghetti integration at the back end. It requires a comprehensive business platform between the user interface and back-end systems capable of managing and delivering a diverse set of personalized, on-demand application services. What's most critical is to identify where your infrastructure needs to be and enable it with the best of the open standards and architectural shifts that have emerged in the past months.
Based on a predictable pattern repeated since the earliest days of the industrial revolution, we're on the cusp of a huge IT explosion toward componentization. So far, enterprise software has followed an evolutionary path consistent with that of all technologies that came before it, from bicycles to computers to chemicals. The phases of technology evolution are:
a) A new technology is first introduced as packaged applications.
b) Additional capabilities are stacked on to differentiate competing products.
c) Once the technology is firmly established, innovation comes from decomposing it into elemental components connected through interchangeability standards.
