The recent acquisition of Gakai by Sony Computer Entertainment could be the sign of an impending bonanza for cloud-based gaming services. As reported in Engadget and elsewhere, the consumer electronics giant paid $380M for the streaming game service in order to deliver new entertainment experiences for SCE’s network. Sony stands to gain big from the operational efficiencies that Gakai’s infrastructure and team bring to the mix, and Gakai’s team got a well-deserved payday for building a great platform. But is this deal the beginning of a trend in cloud gaming, or the end of the line?
Cloud Gaming: the Promise and the Reality
Other leading companies in cloud gaming haven’t fared as well as Gakai. OnLive has struggled mightily, with the company recently undergoing a massive reorganization culminating in founder Steve Perlman’s exit from the company. OTOY is surviving, even thriving, but as a rendering and animation service for the enterprise, not a consumer games delivery platform.
At the same time, the industry is making new investments, such as AMD’s stake in cloud gaming platform company CiiNOW. As existing players retrench and new ones enter the space, let’s take a look at how the reality of cloud gaming matches up with the hype.
The promise of cloud gaming is to deliver high production value 3D games without a download. It seems like a great idea: run old-school 3D applications that were originally compiled for the desktop on high-end server machines in the cloud, then stream the applications to the end user, delivering input back to the server machine for processing. This would appear to solve two problems at once:
1) Relieve developers of the pain of porting complex, already-working applications to new and alien platforms and;
2) Deliver new game experiences on demand to consumers without requiring large downloads or purchasing high-end game hardware.
It sounds good on paper. In practice, there are several problems. First, this approach requires a reliable server infrastructure to deliver the video streams and user inputs back and forth in real time – as opposed to the cheap and disposable wild-west setup that characterizes much of the back end powering today’s Internet. Such an infrastructure can get expensive to build out and operate. Second, the compute load is put entirely on the server side, running on powerful hardware capable of executing high-end games; more expense. This likely explains the massive capital investments made in this space thus far. Finally, all those costs ultimately have to get passed on to the consumer, the developer or both, creating friction on either end. Caught between these twin pincers of high up-front investment and slow customer uptake, something had to give: Gakai sold to someone who could make the formula work; OTOY went vertical, focusing on immediate revenue generation; OnLive went into a nosedive.
Maybe it’s just a sophomore slump. After all, this is not the first time an emerging technology has gone through growing pains. Gakai apparently was able to fit their operating costs into a working model; maybe others can figure it out, too. I wish these companies well and I hope they can make a go of it. But I am not very optimistic because in my opinion, the premise of cloud gaming is fundamentally flawed.
The Assumptions Behind Cloud Gaming
There are several assumptions underlying cloud gaming. The first is that developers don’t want to port their games to a new platform. On the face of it, that’s true. Nobody wants to do extra work if they don’t have to, so the idea that a game could instantly be made available on a new platform and reach new users at little or no additional cost is clearly seductive. But the issue here is hidden cost. All that computation and rendering power formerly running on a consumer’s PC is now pushed up to the cloud, and the resulting expense must be passed on to the developer, the consumer or both. So while there may be little upfront investment, the developer is potentially paying for it on the back end and needs to factor that into the cost/benefit calculation.
A second assumption is that downloading software is a barrier to consumer adoption. Obviously this is true in some genres of gaming, especially social and casual. But it is most decidedly untrue with core gaming audiences, who routinely and ritually download gigabytes of games, patches, mods and expansion packs. The advent of mobile devices and app stores has also muddied this picture: in the mobile world, no-download is not an option. There are no “casual game portals” per se, only apps and various channels for downloading and installing them. This model has changed the consumer mindset so much in the last few years that even the desktop/web world has embraced it in the form of desktop- and browser-based app stores from Apple, Google and others. In short, the idea that end users aren’t comfortable downloading and installing software might have been valid back when OnLive was originally courting investors; these days it’s not nearly as important.
The final assumption behind cloud gaming is that the average consumer’s hardware is not capable of powering a quality game experience. This is also likely a holdover from an earlier time. Nearly every device in a consumer’s possession is quite capable of powering a high production value immersive experience. For desktop and web, the no-download in-browser technologies now available are capable of powering a quality game experience. Between WebGL, Adobe Stage3D (in Flash 11) and Google Native Client, there are several options open to developers. As for devices, even the lowliest phones have accelerated 3D graphics and a native code stack that can run game engine logic at high speed. Not small enough? Apparently even a $25 Raspberry Pi can run Quake 3.
So, if developers are willing to port, audiences are willing to download, and most consumers have gaming-capable machines, what advantages, if any, does cloud gaming confer? None that I can think of.
A Place in the Cloud
In the near-decade since cloud gaming was first conceived, Apple, Google and Amazon have reshaped the media technology landscape. Most consumers today have gaming-ready devices, and the modern cloud has lowered barriers to entry and leveled the playing field for back-end service providers. So does this mean there is no longer a need for cloud gaming services? Not at all. The idea of dividing up the processing work between client and server is a good one; it’s just a question of where you make the cut. Up to now, cloud gaming has done it in the wrong place. It’s time for the industry to rethink cloud gaming to match the new reality.
Part of the confusion around cloud gaming is a lack of understanding about how games actually work. It’s tempting to think of video games as similar to streaming video: they are entertainment products with high production value content, and, well, the word “video” is right there in the name. But games aren’t streaming media; games are applications, interactive media, vs. passive media types such as video and audio. With streaming media, user input is minimal and infrequent, and a certain amount of delayed response is tolerated; with games, input comes several times per second and the application must react in real time or risk user wrath.
From a technical standpoint, streaming media can employ dumb client and server machines, specialized at delivering and decoding large amounts of uniform and predictable data. Games, on the other hand, require smart clients and servers, a general-purpose computer on each end of the pipe. The infrastructure for a cloud-based game system needs to reflect these realities. What the industry could really use is a cloud computing fabric capable of supporting powerful client-server applications. Good news, we already have one: it’s called the cloud.
The modern computing cloud provides a cost-effective means for deploying applications by exploiting economies of scale. The reason the cloud can work so cheaply and flexibly is that it is primarily based on commodity hardware and, often, virtualized machines that allow computing resources to be shared among several customers. This setup is great for running operations that range from large-scale consumer web sites to high-end compute jobs; it’s not so great for running applications that need to respond to the user several times a second – in other words, video games.
Instead of executing entire games on a high-powered server, cloud games should be split between client-side rendering and interaction, and server-side game and business logic. This is not a radical idea; online and multiplayer games have been designed this way for years, well before the emergence of cloud computing. It’s just good client-server system design, putting the computation load where it belongs.
The latest incarnation of online games from Zynga and other large developers already take full advantage of the cloud in this way to reduce operating costs. However, they rely on homegrown software stacks built on top of basic cloud services. Even using the cloud, it’s still an expensive endeavor to develop and operate the back end for an online game, implementing game logic, missions, user data, virtual economies, transaction processing, lobbies, chat rooms and multiplayer messaging. For small shops and developers just getting started, this cost can be prohibitive. For more established developers, offloading these chores to an outside service may be more economical. To me, this is where the real cloud gaming opportunity lies.
Game developers want to focus on design, production and game programming. They don’t want to build complex back end logic, large databases or transaction servers, nor do they want to deal with managing the intricacies of AWS or Rackspace hosting and CDNs.
I believe we are going to see good market uptake for cloud-based solutions that tackle these kinds of problems. San Francisco start-up Roar offers a cloud-hosted platform for rapid back end development of social and casual games.
Earlier this year Zynga announced a third party platform that provides a scalable back end and leverages their massive installed base to help developers reach new players. In the coming years I expect to see more platform services like these emerge to take the handle the heavy lifting of game back end development, deployment and hosting. As for the current crop of cloud gaming companies… I’d say it’s time to get with the program.
Tony Parisi is an entrepreneur and career CTO/architect. He has developed international standards and protocols, created noteworthy software products, and started and sold technology companies. Tony’s passion for innovating is exceeded only by his desire to bring coolness and fun to the broadest possible audience.
Tony is perhaps best known for his work as a pioneer of 3D standards for the web. He is the co-creator of VRML and X3D, ISO standards for networked 3D graphics. Tony continues to build community around innovations in 3D as the co-chair of the WebGL Meetup (www.meetup.com/WebGL-Developers-Meetup) and a founder of the Rest3D working group (http://www.rest3d.org/). Tony is also the author of the upcoming O’Reilly Media book, WebGL Up and Running.
Tony is currently a partner in a stealth online gaming startup and has a consulting practice developing social games, virtual worlds and location-based services for San Francisco Bay Area clients.
Visit Tony’s blog at: blog site, http://www.tonyparisi.com/