When Google’s executives floated a vision for the Stadia cloud gaming service that could make graphically intensive gaming available on any device, they knew the company wouldn’t have to build all the necessary technology from scratch. Instead, the tech giant planned to leverage its expertise in shaping Internet standards and installing infrastructure to support YouTube video streaming for more than a billion people worldwide.
Today, Stadia’s cloud gaming platform, which is in the final stages of development, relies on offloading computing and graphics processing tasks from a user’s device to warehouse-sized data centers that may be many miles away from the player. When the service launches later this year, customers will be able to almost instantly start gaming on Stadia by launching a simple client program that runs on Chromecasts, Chromebooks, PCs, and smartphones.
“Our vision is to have Stadia available on all devices that stream YouTube—a truly platform-agnostic service,” says Majd Bakar, vice president of engineering for Stadia.
That’s easier said than done because cloud gaming has stricter technical requirements than streaming videos. Video streaming services such as YouTube and Netflix only need to deliver video on demand when a person presses play. Video streaming services don’t have to manage a live, interactive gaming experience that also includes feedback from a player’s controls. With Stadia, a user’s device must perform a small amount of additional processing to handle these inputs.
So how did Google build a platform that could deliver a real-time multiplayer gaming experience on any device? The company harnessed preexisting Internet technologies and infrastructure developed for YouTube to prepare for Stadia’s planned November 2019 launch in the United States, Canada, and a dozen European countries.
For Stadia users, Google recommends a minimum Internet download speed of 10 megabits per second (Mbps) to experience a game with 720p resolution at 60 frames per second. Higher broadband Internet speeds allow for better gaming experiences—a download speed of 35 Mbps will allow for 4K High Dynamic Range (HDR) gameplay at 60 frames per second with 5.1 surround sound. The service requires a minimum upload speed of 1.5 Mbps.
Delivering high-intensity graphics over relatively little bandwidth would not be possible without video codecs that compress digital video files into smaller files. Compressed video files require less bandwidth and less time to transfer between a YouTube or Netflix server and a person’s device. That makes a huge difference when video represents more than half of all downstream Internet traffic—and YouTube alone makes up more than 10 percent.
Like YouTube, Stadia will also lean heavily on video codecs to compress what players see into chunks of data that are easier to deliver. One common video codec called H.264 has become a compression standard for popular video streaming services, Blu-ray discs, and HDTV broadcasts since it debuted in 2003. A decade after the H.264 debut, Google also developed a competitor called VP9 that could deliver HD and 4K video streaming on YouTube using half the bandwidth of other codecs.
Google has created “purpose-built custom hardware accelerators” for Stadia that perform speedy video encoding at scale based on both of those video codecs, Bakar says. He added that the Stadia team has also “spent a significant amount of time tuning both the encoders to achieve an optimal balance between latency, quality, and bitrates for real-time interactive streaming under all network conditions.”
Another technology that has laid the foundation for Stadia’s cloud gaming vision is the transport protocols—including basic protocols such as TCP and UDP—that act as middlemen by passing data between devices connected to the Internet. Stadia will rely on specialized protocols developed by Google to ensure fast and reliable transfers of data involving both gameplay visuals and player inputs.
One of the protocols used by Stadia, called WebRTC, arose from an open-source project supported by Google, Mozilla, and Opera. WebRTC allows software developers to build real-time video and audio communication into web browsers such as Google Chrome and apps such as Google Hangouts and Duo.[shortcode ieee-pullquote quote=""Our vision is to have Stadia available on all devices that stream YouTube—a truly platform-agnostic service."" float="left" expand=1]
Stadia will also rely upon Google’s QUIC (Quick UDP Internet Connections) protocol that has reduced connection times and minimized delays in transmission in comparison with the older TCP protocol. Google’s Chrome browser and various apps already use QUIC for more efficient data transmission. But the protocol’s capability to deliver data with less latency can also make a big difference for online gaming experiences that require players to make split-second decisions.
Another speed boost comes from Google’s development of a congestion control algorithm called BBR (Bottleneck Bandwidth and Round-trip propagation time) that runs inside computers and smartphones connected to the Internet. This algorithm can accurately measure Internet traffic congestion so that it can regulate how fast it sends data and how much data it puts into a network at any given time.
“Stadia’s real-time adaptive streaming technology adjusts to network quality in real time,” Bakar says. “Alongside techniques like BBR, we are able to detect network impairments such as congestion prior to them happening—in the case of adverse network events, we optimize for a fluid and responsive gameplay experience.”
These video codecs, Internet protocols, and congestion control algorithms help Google make the best use of the existing bandwidth available in today’s Internet infrastructure—namely, fiber optic cables that span oceans and continents. But like many tech giants, Google has built up its own private Internet infrastructure so that it can quickly deliver online services such as YouTube to customers worldwide.
If there is a physical heart for Google’s services, it’s 16 huge data centers located in the Americas, Europe, and Asia. Each data center is filled with many rows of racks filled with computer servers and other hardware. The Stadia service will make use of a custom-built AMD GPU (graphics processing unit) that can deliver 10.7 teraflops of performance. That compares very favorably to the graphics cards in traditional video game consoles such as the XBox One X with 6.0 teraflops and the PlayStation 4 Pro’s 4.2 teraflops—and Stadia could even leverage many graphics cards at once to supercharge gaming experiences.
To connect those data centers to customers, Google has built a sprawling fiber optic backbone and network dedicated to delivering Google’s services around the world. Even when it comes to laying fiber across oceans, Google has shared or complete ownership of approximately 8.5 percent of the world’s submarine cables.
Stadia’s peer-to-peer connections will make use of Google’s data centers and fiber optic cables to completely eliminate multiplayer latency when players are connected to the same data center, Bakar explains. That infrastructure should reduce latency even if users connect to separate data centers while playing together through Stadia.
Last but not least, Google has placed more than 7,500 edge nodes, which are Google servers installed in the networks of Internet service and network providers. Those edge nodes represent the Google infrastructure endpoints closest to customers. “Stadia servers are deployed on Google’s Edge locations that are closest to partnered [Internet service providers] to further ensure a seamless and consistent gaming experience,” Bakar says.
Google will need to make the most of its vast infrastructure and expertise in managing Internet traffic to deliver a compelling and smooth cloud gaming experience. Whereas live video streaming may still have 500 milliseconds (half a second) to manage possible network glitches without noticeable interruptions, Stadia could have just 16 milliseconds or less to deliver a smooth experience for real-time interactive gaming.
And because cloud gaming is a real-time activity that requires minimal latency, Stadia cannot rely on common techniques such as offline encoding of video files, caching video content in advance along the network path, or taking advantage of video client buffers to smooth out network glitches.
Still, Google’s success in supplying YouTube and other Google services to billions of customers worldwide has yielded useful lessons for the company’s move into cloud gaming. “As long as these devices have good Internet connectivity and are able to decode high quality video, they can handle Stadia,” Bakar says.
A version of this post appears in the September 2019 print issue as “How YouTube Led To Google’s Cloud-Gaming Service.”