Academic journal article McNair Papers

5. Construction

Academic journal article McNair Papers

5. Construction

Article excerpt

If DOD managed the Grid's creation as it does other defense programs, the path from concept to implementation would be a series of steps: designating a program office, establishing requirements, selecting contractors to do the work, writing and testing code, evaluating the final product, then fielding it.

Creating the Grid calls for different steps if only because of the Grid's size and complexity, and because integration is not construction. DOD needs consensus on a broad vision for the Grid (see chapter 4). It must establish core services, standards for interconnection and interoperability, and methods that let nodes integrate themselves quickly under both normal and wartime conditions, and then hang new nodes and services on the structure so created. In the process of creation, the Grid could become the world's largest testbed for new information technologies.

To address issues involved in its creation, this chapter points out some difficulties of top-down integration: a process that encompasses a requirements definition, a hierarchical decomposition of tasks, and the specification of hardware and software to meet these tasks under central control. Next are examined some prospects for bottom-up integration: where systems, subsystems, and components are given much more leeway in meeting broad goals through their own resources and help negotiated with their peers. Then the role of architecture is discussed, and guidelines for planning, experimentation and R&D suggested.

Some Difficulties of Top-Down Integration

As a top-down project, the creation of the Grid could be the single largest software project ever attempted. Great size makes any project difficult to manage and may lead to results too fragile to field. Because the Grid must absorb and accommodate existing capabilities, the conventional model of systems integration may not fit.

Great Size

A complete top-down Grid could include every process, application, service, database definition, and interface specification entailed in a complete [C.sup.4]ISR system. It may, at the outside, require a billion lines of code costing $5 to $10 billion a year for a decade or two. (1) Software accounts for a growing share of a defense system's costs, and integration is a growing share of the cost of software. A large program, such as the F-22 fighter, may include $5 to $10 billion worth of systems integration (some of that is to ensure that hardware can fit in tight spaces and work together reliably). The next-generation Air traffic control system had, by mid-1997, absorbed $7.6 billion, mostly for software. (2) Although increasing the connectivity of DOD equipment may not be expensive, (3) a top-down tightly coupled Grid may be. Is something so large doable?

Finding money may be a problem. Only the services can raise so much money and only if working together. Dividing the Grid into modules, tasking each to a service, and reintegrating their results could take whatever time was required to build the modules, plus years of politics to the front for apportionment and years of politics to the back for reintegration. Even then, the services might not want to spend billions on a top-down design not of their own making. No service would relinquish its oversight over the Grid's development once it sees how much design decisions will influence its own acquisitions. Many are already building their own version of the Grid: the Army Force XXI, the Navy CEC, (4) the DISA GCCS, various ACTDs under DARPA, and the emerging Total Asset Visibility program for logistics, all of which offer the prospect of partial integration at the risk of making complete integration more difficult (with domain-level stovepipes replacing service stovepipes).

Meeting objectives on time and under budget is always a challenge for such a large software project. The difficulty and thus cost of integrating a system grows more quickly than its size. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.