How the Quest to Explore the Next Frontier Has Changed the Frontiers of Computing
May 09, 2017 • Blog Post
IN THIS ARTICLE
- To successfully get to Mars, we need to rethink how we build computing systems
- The first human-led mission to Mars will rely more on computational power than manpower
A computer engineer and theoretical physicist discuss the steps needed to get humans to Mars
Mars is alluring for us all. Moviemakers want to profit from it. Astronomers want to study it. Elon Musk wants to colonize it.
Whatever the purpose, the trick will be landing human beings on Mars and then returning them safely to Earth. In the following Q&A, Kirk Bresniker, chief architect at Hewlett Packard Labs, and Sean Carroll, a theoretical physicist at Caltech, discuss the importance of Memory-Driven Computing to this task.
What is it about space exploration that tends to push advances in computing power and architecture?
Kirk: Some people have asked me if Mars is the next "moonshot". As the son of an aerospace engineer, I tell them it's so different this time. With every Moon mission we were in near real-time communication with the astronauts in flight. Help from Houston allowed for conducting dozens of on-the-ground simulations of the problems aboard Apollo 13, leading to solutions that ultimately got the crew home safely. We won't be so lucky going to Mars, which is 20 light-minutes away from Earth's NASA command centers.
Sean: Right, and because we won't have consistent contact with control on the ground, we need computational power that is adaptive in real time to address unanticipated challenges - and there will be plenty.
Kirk: Imagine what that would have to look like? It's very different from what we put in data centers here on Earth where we can cool large, muscular, overworked machines with a well-placed air conditioner. The prospect of a human mission to Mars has made us rethink how we build systems with the sort of power we'd want in a data center but with the footprint and power signature we'd need for serving aboard a spacecraft. Memory-Driven Computing is our answer.
What makes the mission to Mars a unique computing problem compared to other explorations?
Sean: There's so much we don't know. We haven't mapped Mars in great detail and the terrain and atmosphere are not like Earth. So you're falling, you're zooming down toward Mars, scanning it in real time and deciding exactly where to land. How do you do that safely?
The more computer power you have and the more clever your algorithms are, the more choices and flexibility you have at the moment when you're going to land your spacecraft. There's more square kilometers available to you if you can make choices faster and more efficiently during that intense period when you're hurtling toward the surface of the Red Planet.
Kirk: We should also remember that everywhere our Mars explorers go, as soon as they go behind the orbit of the Moon, they're somewhere where a human observer, a human scientist, a human operator, has never been before. Nor have we ever sent electronics more sophisticated than a flip phone along on a human-led mission to space. With a Mars mission, we'd have no choice but to send the most sophisticated computing system possible. On the ground backup will be too far away.
Sean: Or the on the ground backup will literally be on the ground, roaming the Mars surface. One of the dreams is to have remote probes and drones and rovers capable of making decisions based on the data they collect.
Kirk: Exactly! Edge computing in space, enabled by memory-driven systems that can combine as a distributed mesh, instantly accessing tens of thousands of stored procedures, or autonomously completing tasks that advance the mission, at a level of speed and efficiency that's orders of magnitude better than the processor-driven systems NASA has employed to this point.