Opinion: Defense software requires infrastructure too

Aircraft carriers make awesome TV — enormous floating metal platforms, with jets accelerating loudly off their decks into the sky, flaring molten orange trails from their afterburners. But software “platforms” that keep a carrier battle group functioning, linking sensors across air, land, sea and space to protect our military personnel? Not so sexy.

Software and artificial intelligence (AI) algorithms may not have been designed for network news, but software is an indispensable capability for our military. Our service members require cutting-edge software to enable all the hardware platforms — ships, tanks, and planes — they use to carry out missions to defend our freedoms around the world.

And while all would agree that a well-maintained physical infrastructure is essential in the civilian world to establish new communities, grow our economy and build better lives, we also need infrastructure for the advanced software our military requires. We are facing adversaries who are actually quite good at software and technology, and we must be better and faster than them, or we will cede our strategic advantage.

Infrastructure for software embraces four important elements:

• Unlimited computing power and unlimited data storage;
• Open APIs (application programming interfaces) that allow different software applications to talk to one another;
• Continuous deployment and delivery of software; and
• Ability to operate at the edge of communications networks.

Unlimited computing power and unlimited data storage

This is the prerequisite for everything — the equivalent of securing reliable energy supplies in the world economy. AI and machine learning will allow us to do things on a scale and at a speed that humans cannot do, but they require access to enormous data sets, and a vast amount of computing power to rapidly run tests on that data. Ensuring that unlimited storage and computing power are available will allow our smartest engineers, working for new companies in the cloud, to find new ways of using software to keep our systems ahead of our adversaries.

Open APIs

We should require any company that is providing any capability to the Department of Defense to have open APIs so that anyone can connect, anyone can write to the system, and anyone can receive data from the system and authenticate to the system — not just the original contractor on the program of record. This allows new companies, new disrupters, new unforeseen technologies, and new capabilities to be added to already existing systems at any moment.

That is essentially how the internet works today in the civilian world, allowing new capabilities to be plugged in, or new apps to be developed and added to smartphones. Contrast that model with the F-35 Joint Strike Fighter, where plugging in new capabilities is really hard for any outside contractor. Open APIs will allow us to guard against vendor-lock in the long term.

Continuous deployment and updating of software

Things change quickly in the software world. We must have infrastructure that allows us to update it every day — or every hour — if necessary, and release new capabilities that maybe didn’t exist a week prior, just like smartphones automatically update overnight. In the military of today, it simply isn’t good enough to wait for periodic servicing: If it is going to take 12 to 18 months to update software on a fighter jet, that’s not going to be effective on the battlefield.

Operating at the edge

We will need to have systems that can operate at the very edge, far from home in denied environments that don’t have access to continual communications — high in the mountains of central Asia, in the middle of the Pacific, or in a zone where our adversaries are blocking all signals. That means building resilient systems that can disconnect and reconnect, synchronize and go back in as and when possible.

Prioritize outcomes, not architectures

Finally, it is important to remember that we will never be able to fully guess the requirements of the future. In the civilian world, when we zone areas for development and build new access roads, we don’t know what buildings are going to be constructed, what new businesses are going to be established there, or who is going to choose to relocate there.

Similarly in the military when we start applying AI algorithms to massive data sets, we won’t know exactly how those results will go together with missions and objectives that evolve over time. So our requirements must be more flexible. It is going to take a little courage because we are going to have to break down barriers between combatant commands and the services so they all can share data. It is going to be disruptive, not the norm. But we have to do it. And we need to know we have the software infrastructure to make it all possible.

Chris Lynch is the chief executive officer at Rebellion Defense. Previously, he was founding director at Defense Digital Service. His team launched high-impact programs, including JEDI Cloud and Hack the Pentagon, and provided technical expertise on the Department of Defense’s most critical technology challenges, including the F-35 Joint Strike Fighter and Next Generation GPS OCX. Chris is a serial entrepreneur from Seattle, where he founded venture-backed startups and led engineering teams at enterprise companies. He is a fellow at Harvard Kennedy School’s Belfer Center and lecturer at Stanford’s computer science department, where he encourages nerds to serve in government.