Building breakthroughs in wireless technology

Working on connectivity technology that’s 10x faster, or 10x cheaper, or both.

By: Jay Parikh, Head of Engineering and Infrastructure at Facebook

Facebook’s mission is to connect the world. It’s a big and ambitious goal that poses significant technical challenges. 
It’s not that we don’t know how to connect everyone. That’s easy. Dig fiber lines and put up towers. The problem is that today’s connectivity model costs too much. If there isn’t a large enough population to amortize the cost of the infrastructure, businesses will never invest in a region. That leaves huge portions of the world without connectivity, particularly in rural areas. And even in densely populated countries, the connectivity comes in pockets, has to be shared, and is hampered by physical infrastructure, a cocktail that leads to the connection being unreliable. 
Current business models won’t address these needs. It simply costs too much to put in the infrastructure. Connectivity is expensive. 
That’s why we launched Connectivity Lab. It’s designed with one goal: Develop new technologies that quickly advance state-of-the-art communications to the point that they become viable solutions for others to deploy. In other words, we aim to create connectivity technology that’s 10x faster, or 10x cheaper, or both. (We’re inspired by Mervin Kelly’s “better, or cheaper, or both.”) Sometimes this research means pursuing solutions in places no one has put a network before, like the sky, or reevaluating the technologies we use on the ground to see if we can use them in a more efficient, smart manner. We’ve got updates in both of those pursuits. 
Last year, for instance, we introduced Aquila, an unmanned aerial vehicle (UAV) designed for areas with medium population densities and limited or no connectivity. The lightweight, carbon-fiber Aquila (which is Latin for “eagle”) has an unusual look, with a tailless design and the wingspan of a 737. It’s designed to fly for months at 60,000 to 90,000 feet, putting it above commercial flights and most weather patterns.


The aircraft uses a laser to transmit data to other airborne Aquila aircraft, creating an extensible network in the sky. The planes beam connectivity down to small towers and dishes on the ground. Ground stations then convert the signal into Wi-Fi or 4G networks, providing connectivity at up to 10 Gb of data per second to some of the most remote areas on the earth. This model makes connecting rural regions an order of magnitude less expensive than current models. 
On the ground, we’ve made important progress on two areas of research. One solves the burgeoning problem of backhaul connectivity not keeping up with the demands of immersive technology. The other is a good 5G candidate technology that uses a fresh look at the best way to squeeze bandwidth out of antennas. We call the first Terragraph.
The increased sharing of rich media (photos and video, and eventually virtual reality experiences) can create maddening network slowdowns, particularly in urban areas where network demand often far exceeds capacity. Terragraph addresses this issue with a high-bandwidth wireless system that delivers gigabit speeds to highly populated areas. 
Terragraph uses small nodes hanging on “street furniture,” and client nodes then distribute Wi-Fi or small cell connectivity to people and businesses. Combined with Wi-Fi access points, it is the lowest-cost solution to achieve 100 percent street-level coverage of gigabit-class Wi-Fi. By using off-the-shelf components and leveraging the cloud for intensive data processing, Terragraph will be cost-efficient and highly scalable. And, given the architecture of the network, Terragraph is able to route and steer around interfaces such as physical objects as well as user internet congestion.

Looking at how this plays out on a map, this solution may seem obvious, like someone should have tried it before. But two breakthroughs have made this possible. First, Terragraph operates in the unlicensed 60 GHz band that is traditionally avoided due to its high water and oxygen absorption. Instead of avoiding it, our team took advantage of this unique absorption attribute to mitigate interference and operate at a very high capacity. 
The second is that we’ve applied our data center knowledge to improve the resilience of the network. We’ve made enhancements to the air interface to improve efficiency. We implemented IPv6-only nodes, an SDN-like cloud compute controller, and a new modular routing protocol for fast route convergence and failure detection. 
We’re testing Terragraph on campus at Facebook and beginning to build large-scale trial Terragraph networks in markets around the world, including one in San Jose, Calif. So far, we have demonstrated 2.1 Gbps total bidirectional throughput per link, and we think this number can be much higher in the future. In communicating with multiple nodes, we have been able to switch between nodes in 8 microseconds, or about 125,000 per second. This will be important to enable high-speed, efficient distribution networks.
The 5G candidate technology we’re working on builds on a current technology called MIMO — multiple input, multiple output. MIMO is a commonly used technology today in Wi-Fi and in 4G/LTE. The idea here is that you use multiple transmit and receive antennas to increase the capacity of a given RF signal.
We call our proof of concept of the next generation of this research ARIES — Antenna Radio Integration for Efficiency in Spectrum. ARIES is a form of Massive MIMO. Simply put, Massive MIMO uses a lot more antennas to squeeze the maximum amount of data out of the RF signal. The extra antennas help focus the transmission and reception of signal energy into ever-tighter spots, creating faster reception and better quality.
ARIES pushes Massive MIMO to the extreme. We’ve built a base station with 96 antennas that can support 24 simultaneous streams on the same radio spectrum — 71+ bps/Hz — and we’re on track to provide 100+ bps/Hz of spectral efficiency, the best in the world. This all gets done through some hefty signal processing; our base station processes over 6 billion operations per second. We’re able to stream dozens of videos simultaneously over the same radio spectrum without compromising performance or picture quality.

ARIES prototype

ARIES has important implications. In densely populated spots, a single tower could replace multiple setups. In rural areas, ARIES will send a signal much farther. Our ARIES efforts are focused on outdoor trials, but we hope to make it available to researchers and academics soon. 
We are working on this technology to be open and interoperable via unlicensed spectrum, just like Wi-Fi itself. We want new types of gigabit-class wireless networks to be deployed based on this technology, and we want new business models to be developed to bring more affordable access to people. 
In the same way that we’ve contributed open data center technology through the Open Compute Project, we plan to iterate on Terragraph and contribute our Connectivity Lab projects to TIP — the Telecom Infra Project. 
With these technologies, we think there are ways to get the 1.6 billion people who don’t have access to the internet connected. We know if we prove that these radical new cost-saving approaches work, we can change the fundamental way people are connected. Everyone can have access to the same online opportunities that a lot of us take for granted. It’s going to be a long journey to get to 10x faster, or 10x cheaper, or both, but I’m excited to say that we’re well on our way to discovering new ways to connect the world.

Show your support

Clapping shows how much you appreciated Jay Parikh’s story.