Like his father, Herman Narula is in the construction business.
His father is the multi-millionaire Harpinder Singh Narula, a construction mogul who builds things like the eight-lane highway that stretches from Delhi to Indira Gandhi airport to the Indian call center hub in Gurgaon. And with his own company, Improbable, the younger Narula is behind a similarly ambitious project. It’s just that he works in the digital world, not the physical.
After graduating from Cambridge, Narula and a fellow student named Rob Whitehead set about building technology that could help engender a new breed of online game, games so vast and complex that they would continue to run—and continue to evolve—even when no one is watching. Basically, he envisions virtual worlds that run across tens of thousands of machines in a wholly unified way, expanding to new machines as they evolve, and with Improbable, he’s providing a way for any game developer to build and operate this kind of alternate universe (see video below). “We’re like an operating system on which you can build these worlds,” the 27-year-old says.
Others have built such virtual worlds in the past, including Eve Online and, most notably, the alternate universe known as Second Life. But Narula and company aim to simplify the process, to give game designers a tool that makes building games for thousands of machines as easy as building for one. Some designers are already using Improbable’s technology to build new games, including Dean Hall, maker of popular indie game DayZ . “The dream of worlds like this isn’t new, but the approach is new,” he says of Improbable. “It’s a quite profound change.” But others, including Narula, believe it can do more.
Narula says Improbable can help simulate everything from traffic patterns to economies to contagious diseases, and Vijay Pande, a professor of chemistry, structural biology, and computer science at Stanford University, agrees. He’s considering the tool as a way of doing biological research, simulating systems of cells.
Pande is a kind of scientist-in-residence at the big-name Silicon Valley venture capital firm Andreessen Horowitz, and last week, the firm pumped $20 million into the company, with Andreessen partner Chris Dixon taking a seat on the Improbable board. Like Pande, Dixon sees Improbable not just as a gaming thing, but as something bigger. “It’s like a super-charged Amazon Web Services,” he says, referring to the seminal cloud computing tools that provide instant access to machine power over the net.
Improbable is just one of many tools that can help build software that runs across tens, hundreds, even thousands of machines—i.e. software for the modern age. Most apps now run in this way, from Google and Facebook and Twitter to Uber and Dropbox. But building such services is an immensely complicated undertaking, and all sorts of tools now aim to simplify things, from Amazon Web Services to Hadoop, an open source tool for analyzing data across a vast array of machines, to Docker, a way of readily distributing most an application across a network of computers.
Narula and company don’t say much about how their technology works. But its team includes engineers who have helped build sweeping applications at places like Google, and Narula does say that the tool uses Docker and a slimmed down computer operating system called CoreOS. In short, these tools provide a means of more efficiency running distributed software, and apparently, Improbable tailors this kind of thing to simulated worlds. Dean Hall describes it as something that lets him plug into an enormous number of machines without having to think about how all those machines will work.
Mark Ferlatte, who spent nine years overseeing the technology that ran Second Life, says the much-dicusseed virtual world operated a lot like the games that Narula envisions. “It sounds very similar to what we did with Second Life,” says Ferlatte, who now runs a consulting firm called TetherPad, which specializes in online infrastructure. “The simulation was running all the time, and scripts would execute and respond to things and do things even when no one was around.” And if Improbable is based on Docker and CoreOS, he adds, it seems a lot like standard technology. But he also says there’s ample room to streamline and improve the creation of virtual worlds like Second Life.
Second Life ran across many distributed machines, he explains, but it was fashioned in such a way that some machines could be overloaded with traffic. Building with a newer breed of technology, Narula says, Improbable fixes this problem, making it easier to run software in a truly distributed manner.
The question is just how easy this really is. “The claims that makes me raise an eyebrow is the ones that say a game designer can designer without thinking about infrastructure at all,” says Ferlatte. “To make a pun their name, that’s improbable.” And though Narula and others pitch it as a way of running biological simulations and modeling economies, Ferlatte believes this is a very different prospect. “You get into precision concerns, which don’t matter as much with games,” he says.
But for Narula, games and simulations aren’t that far apart. In each case, he says, you need a way of running a virtual world that acts like a whole but runs across many machines. Each piece of the code must operate on its own, but also in tandem with any other part of the the whole. “You’re creating a place where there are things. These things interact on various ways. And the only real property that defines this is that each thing isn’t always talking to every other thing,” he says. “This is true of a crowd, of a city, of traffic, of economic activity, of disease propagation.”
Yes, it’s a high-falutin pitch. And Narula acknowledges the enormous ambition of his project. That’s why the called the company Improbable, he says. But at the same time, this is indeed where the software world is moving. And that’s why people like Chris Dixon are so high on the technology. “On the one hand, you have millions of computers available, through things like Amazon Web Services. But on the other hand, it’s really hard to write software that runs across so many machines,” Dixon says. “Improbable can provide a bridge.”
