Could the COTS movement be heading in the wrong direction?
Movements and organizations have a funny way of losing their original focus as they evolve and grow to maturity.
By John Keller, chief editor
Military & Aerospace Electronics
Movements and organizations have a funny way of losing their original focus as they evolve and grow to maturity. A compromise here and a compromise there, and pretty soon a movement starts to bear little resemblance to what it looked like in the first place. What was once a unified movement starts to fragment. The worst phase of this process begins when the organization collectively becomes more concerned with perpetuating itself rather than promoting its identified cause. This phase typically alienates the movement's original proponents, and often precedes the movement's ultimate demise when enough people decide to chuck it altogether to refocus on their original goals.
Lately I've started to worry about the COTS movement, and how it may be heading down this road. COTS, short for commercial off-the-shelf, represents a movement that now is eight years old. While it once described an exciting new paradigm shift from custom mil-spec components to more powerful and affordable commercially developed components, COTS is simply the way that defense and aerospace electronic systems designers do business today. The concept of COTS is mature, and as such is subject to the decay and fragmentation that have plagued many other mature movements.
It bothers me that COTS is becoming synonymous with "throw-away." The dizzyingly fast rate of obsolescence among COTS components and subsystems has forced electronics designers into a strategy of embracing rapidly obsolescent — or throw-away — technology. They have little choice. Commercial electronics suppliers rarely listen to the needs of military systems designers. Why should they? Military and aerospace systems have become such small parts of the electronics business, compared to more-lucrative personal computer, personal digital assistant, and telecommunications markets, as to be virtually invisible. Military systems designers simply have to use commercial throw-away technology in many areas for lack of a viable alternative.
Don't misunderstand me. Embracing throw-away technology, and setting forth long-term plans on how to use it to best advantage, certainly is not a bad thing. Lots of worthwhile electronic technology is essentially throw-away. All you have to do is look at desktop PCs to understand that. This approach enables users to move relatively quickly from one generation of technology to another, and keeps them close to the leading edge.
Still, I am concerned about the scope of throw-away electronics in military systems. Increasing numbers of system components are falling into the throw-away category. It isn't just chips and boards anymore. We're seeing large powerful computer servers put to work in military systems, and these servers go obsolete in a couple of years, sometimes even faster. The U.S. Navy's Acoustics Rapid COTS Insertion (ARCI) program is one example of this (see feature on page 16).
The ARCI program today uses the Compaq 8500 8U server as the main processor for the world's most advanced sonar signal processing. The reasons for this are solid: it's rapidly upgradeable, affordable, and it does the job. Period. It's difficult, if not foolish, to argue with the logic of the ARCI's system architects — Digital System Resources Inc. (DSR) of Fairfax, Va., and the Lockheed Martin Naval Electronics & Surveillance Systems-Undersea Systems (NE&SS) in Manassas, Va.
Still, throwing away an entire computer server when it becomes obsolete is not the same as throwing away one or two single-board computers. I wonder what the long-term costs and logistics issues of supporting and continually upgrading a server-based sonar signal processing architecture like this will be.
One answer, according to DSR and Lockheed Martin, is to move away from the large 8U server and toward the new generation of 2U servers. That way, upgrades involve smaller components, and processing power continues to increase. In addition, the system footprint per gigaflop also shrinks substantially. Each of these 3.5-inch-high 2U servers is a stand-alone dual-processor computer that connects to other servers in a rack via a high-speed serial interconnect such as Fibre Channel or Gigabit Ethernet.
These 2U servers sound to me a lot like the new generation of so-called "blade servers" from Compaq, Hewlett-Packard, IBM, and others. Blade servers, much like the ARCI 2U servers, are self-contained computers that fit into racks and communicate among themselves over high-speed serial networks. ARCI designers admit that they most likely will purchase 2U servers in bulk, and each lot from only one company, to take advantage of volume discounts. ARCI engineers say flat-out that their architecture will not use blade servers, and I take them at their word.
However, the ARCI architecture clearly illustrates how tempting it is for systems designers to evolve from a VME, CompactPCI, or similar single-board computer architecture to a blade server architecture. After all, blade servers are available today, offer staggering computing power, and are available at volume prices.
This is what worries me most of all. The whole notion of COTS is not so much about using store-bought instead of home-brew technology as it is about employing open-systems architectures where hardware and software from many different vendors compete on a level playing field. The emergence of blade servers as they may apply to military and aerospace electronics run precisely counter to the concept of open-systems architectures.
Then-Defense Secretary William Perry had a lot on his mind when he mandated the use of COTS wherever possible back in 1994. He wanted defense systems designers to quit re-inventing the wheel and purchase off the shelf the kinds of technologies they hitherto had been developing themselves — often at higher prices and lower capability than was available in the commercial marketplace.
But Perry's COTS mandate went much father than that. At the core of the Perry COTS Doctrine was a move to prevent component and subsystem suppliers from holding systems integrators captive to proprietary hardware and software. Perry wanted companies to compete fairly for jobs to provide microprocessors, board products, backplane databuses, enclosures, and support software, rather than to create virtual monopolies by selling proprietary hardware and software that would not function with third-party offerings.
The new generation of blade servers appears headed in this direction. Today there are no industry standards for this technology. It's no secret that servers are one of the last bastions of profitability for the major computer companies. That said, I doubt if any blade server supplier today seeks to marginalize that profitability by adhering to industry standards that would render this technology as much a commodity as VME single-board computers are today.
It is in the blade server supplier's interest to maintain blade servers as proprietary technology. So each time a military and aerospace electronics designer specifies a blade server, he is stepping onto a slippery slope that well could lead him back to the bad old days of closed systems and monopoly suppliers.
Where this will lead is to a fragmented COTS supplier base of open-systems COTS and closed-system COTS, and this is not what the military needs.