Thinking in a Different Way: Computing as a Utility
Lately, we’ve been out promoting our embedded computing hardware to a wide audience of senior Army leadership, and I must say that the response has been great. But, as you probably know, these people are not the ones who actually buy stuff from us (those are primes and integrators). They are, however the folks that command those who most depend on how well our stuff performs—our soldiers. You can see why we want to know what they think and what they want from our products. We’re listening, I assure you.
The Army has a history of standing up grand initiatives. Those who follow Army matters will remember ARH (Armed Reconnaissance Helicopter), FCS (Future Combat System) and GCV (Ground Combat Vehicle), just to name a few. All of these programs launched with great fanfare, only to fall to the budget axe. Today’s Army, and the other international armies we work with, plan differently now, and nowhere is this more important than in the area of embedded computing. We see a much more pragmatic approach to fulfilling current and future needs; an aim for future capabilities that build on the strong evolution of commercial computing hardware and software. At GE, this is what we’ve been doing for our entire history, and now we’re feeling the strong resonance of our philosophy in the user community.
Over the last few months, I’ve been sharing our views on the best route forward in providing the sometimes competing facilities of security (cyber, information assurance), low SWaP-C (size, weight and power, cost) and the high performance computing needed on vehicles. The load on embedded computing for enhanced Situational Awareness—or, as we like to call it, Situational Understanding—includes huge increases in demand for C4I, ISR, PNT and especially video processing.
Speaking of video processing… I recall one system a few years back that used SD (standard definition) resolution cameras requiring 26 CPUs to handle 360-degree views. Today, those cameras are UHD (Ultra High Definition) an increase of over 50 times the processing demand.
Tapping into the profound improvements in computing power per size / weight / electrical power has allowed us to keep pace (mostly) with these evolving computing demands, but under the credo of “smaller, faster and more expeditionary” and the strong desire for commonality (Army initiatives such as “convergence”) we need to do more than depend on Moore’s law: we need to think about embedded computing in a different way.
Many years ago, I was involved in photovoltaic power generation. Indeed, I was one of a group of 12 people who were picked to conceive, design and build the world’s first grid-connected, utility-scale power plant. It was one megawatt—small by today’s standards, but a fantastic achievement none the less and a highlight of my career. (By the way: the site was unmanned and utilized 108 commercial computers to supervise everything. When completed, we threw the switch and walked away.) Looking back on that experience recently, I had what for me was an epiphany. Embedded computing should be like a power system; in other words, a utility.
We all use utilities every day; we have gas, electricity and water delivered to our homes in a complex network that manages those resources dependent on need. While they are finite, demand is spread around a large pool of users so that the demand of one user is offset by the lack of demand by another. Careful management and initial proper sizing makes sure that that this balance is struck. In military, platform-level embedded computing, are we not in the same situation? Certain systems demand computing power in certain situations, but not every system demands maximum power all the time. Yet: this is generally how computing power is provided—as a series of dedicated (“stovepiped”) computers, each capable of full performance even when idling along. It’s wasteful and inefficient and the Army cannot afford either of those things.
Think about the internet. This is a computing utility. Every day, we utilize large, centralized servers to do complex calculations as well as transact finances, deliver emails, recipes, music and even cat videos. We access these servers with Macs, PCs and Linux rigs—smartphones and tablets running a variety of operating systems—and it all works! And, like all utilities, we don’t think a thing about it; it’s just there.
Army ground vehicles would benefit from an embedded computing utility and it could work just like the servers that provide the data services we know so well. Centralized, properly sized, protected and highly reliable embedded computers would parse demand. Any one system could be served by the entire resource and all systems would function smoothly in the ebb and flow of their demands. This integrated computing power would be fully utilized, eliminating the “dead weight” of resources sitting at idle. Oh, and, by the way: tying individual systems together on the “Computing Utility” allows seamless data interconnection as well.
At GE’s Intelligent Platforms business, we have the hardware and the middleware, and we’re showing the Army our ideas for innovative new architectures that can make the embedded computing utility a reality.