How can you spend $5Bn for any single project?
Today's servers have 12 computer cores in a dual-CPU system costing $3000. To spend $5Bn to get a program completed even with customized software smells strongly of how we did business in the 70's, not how its done today.The underlying causes are a) over-specifying by the buying agency (remember the $300 hammer); b) an unwillingness (perhaps a strong positive reluctance) to move to low-cost COTS technology; c) provider pressure to maximize solution complexity.
Making headway on this problem requires a directional dictat from the National CIO's office. One or more of these dinosaur programs should be refunded as a COTS-only program, biting the bullet on the software implications, which may be minimal if a rewrite is already planned. The implication is COTS acquisition costs, support costs and overall risks will much outweigh the dubious benefits from a unique, proprietary design.
COTS has worked very well for the commercial market. Oracle, for instance, converted all of their internal data centers, ironically, from Sun minicomputers to Compaq servers, saving a good deal of money and improving flexibility and stability as well as throughput per $$ and per watt. It is time for these lessons to apply to the government and military space, more than the small steps being implemented at this time.


