Five Unconventional Ways of Thinking about Edge Computing & Cloud Streaming

We are undoubtedly witnesses to the birth of a new ecosystem composed of edge server infrastructure, cloud streaming services, and low latency networks. Understanding its early evolution is not easy, as we seek to find trends in the primordial soup of fiercely competitive companies, billions of potential users, and a swell of technologies.

This week, in the course of publishing my third white paper of the year on edge computing, I reflected on some of the peculiar ideas which floated to the surface in the course of research. Not all of these observations are entirely true, but they are mental prods to think harder, and to think a bit differently.

1. Streaming Undoes the World Wide Web: Cloud streaming is anti-HTML. Instead of using markup, code, and bits of text and visual content to compose a user experience, cloud streaming simply renders output to a virtual stream. Instead of a sophisticated web browser, the client PC simply needs to decode video to render the application experience. HTML is nice in its ability to flexibly display on any client; in cloud streaming, servers encode to the client's target aspect ratio and resolution. We still have to get user input into the local device and back into the cloud - taps, swipes, or in the case of VR, head pose. But its a lot simpler than the massive stack of technology which power's today's World Wide Web.
2. Mobile to Mobile Streaming is the Real Enchilada: While the early streaming services focused on PC games, streaming mobile apps to smartphones is more likely to find the largest audience. Facebook launched cloud streaming for games this week - on a mobile (e.g. ARM) infrastructure. Cloud streaming mobile applications to smartphones actually makes sense from user and service provider perspectives. Mobile applications are already designed with touch screens and mobile interfaces in mind - so streaming them from the cloud does not need to translate from a PC or Console experience to a smartphone. PC and console games can serve significantly fewer users per server, so both capital and operating (power) costs are higher.
3. Streaming Breaks The (App) World: Mobile streaming is an attempt to circumvent Apple's ability to take service fees from content and service providers, but its a long game. Apple doesn't want app streaming via mobile browsers to take off because "this breaks the world" as Apple (and to a lesser extent Google) has composed it. How long can Apple hold on? Aggressive competitors like EPIC Games, Facebook, and Amazon have decided to start the bombardment now by launching services and negotiating and potentially litigating. It's a bigger battle, but streaming to browsers and avoiding the app store and going direct to browser is an important part of the assault, breaking down old walled gardens and creating new service models.
4. Local Play is Required for Takeoff: Cloud streaming might not take off without local play, local run solutions. Most cloud services have an offline mode. Google Docs has become increasingly usable and seamless in offline mode, reconnecting and syncing when the network connectivity is broken then re-established. Gamers are accustomed to offline play, and find it confusing to buy a game at full price which they can only play if online. Why can't cloud gaming and soon app streaming services act the same way? And will users tolerate services that cannot function with periods of reduced bandwidth or broken connectivity? Streaming services which assume they can look away from what happens when users are offline are likely to miss a critical stage in the user journey.
5. Edge Computing Must Reconcile Irreconcilable Forces: Edge computing is like trying to place magnets of the same polarity together. A major user objection to cloud streaming is lag or latency in fast action games. Reducing latency can be accomplished by placing really powerful servers very close to end users. However, it seems that the closer a server is located to the end user - at the furthest edge of the network - the higher the server, location, power and operating costs. Making services viable will require a compromise - one which places servers in locations which are close enough to reduce latency but not so far out at the network edge as to drive costs beyond economic viability or end-customer affordability.

If you would like to learn more, you can read the entire paper "Interactive Applications Powered by Video at the Cloud Edge." It presents a TCO and environmental impact model for streaming services that employ video to deliver experiences from the cloud edge. It looks at state of the art solutions for video encoding comparing CPU, GPU, and high density ASICs from NETINT. I also invite you to connect and reach out here on LinkedIn.