Cloud players have known for a long time that lower latency can boost revenue. Amazon famously revealed, ten years ago, that every 100 ms of latency costs them 1% in sales. When we consider latencies in the hundreds of milliseconds or even seconds, revenue is impacted by human behavior: every second spent waiting, about 10% of customers will move on to another web page. Clearly, other cloud providers agreed, and now we have a long-term trend toward regional colocation data centers.
In the mobile market, edge computing has the potential to create new revenue as well, but moving the data center out to the mobile radio tower doesn’t save hundreds of milliseconds…it can save about 20 to 40 ms. That means that it won’t have much impact on traditional online revenue, which is based on human web apps.
On the other hand, extreme low-latency mobile use cases can drive new revenue in IoT use cases, with automation of industrial equipment, vehicles, and robotics, as well as new online gaming and AR/VR use cases. In these cases, latency below about 50 ms starts to matter.
Mobile Experts has recently completed an investigation of revenue created in these latency-sensitive applications. We broke down the revenue according to the latency requirement, to illustrate how much revenue comes from sub-5 ms responses, up through 50 ms responses. Note that we considered total latency, not just the radio latency for the mobile network. At these low levels, the computing resources, message sizes, and radio technology all come into play.
Mobile operators are still considering how they will invest in edge computing and integrate it with their networks. Should the operators develop their own cloud platforms? Should they partner with AWS, Azure, or Google? Where should the data centers be located for the best profitability?
In our study, we found that most of the new revenue opportunity during the next 10 years is related to latency levels in the 20 ms range, not the extreme 5 ms case. And looking at the investment required by the mobile operators to put computing power on every radio tower, it’s simply too expensive to deploy edge computing at the “far edge” of the mobile network.
Our conclusion is that the ideal location for mobile edge computing will be in regional data centers, covering a small city or a portion of a major city, but not at every tower. This is a step beyond the current colocation strategy of the cloud providers, where they’ve set up regional data centers that cover wide areas.
The web scale guys will almost certainly control most of the applications. But they can’t address the low-latency applications without a network that brings its core out to each city. So the web guys are stuck with regional colocation, waiting for the networks to improve.
The mobile operators have considered developing their own cloud, but that strategy would be suicidal. They can’t get the developers to support them adequately. Mobile operators do control access to customers, as well as spectrum and local real estate assets. They’re in a good position, but they will be stuck with a low-latency network and no customers if they try to fly solo.
Some operators (notably AT&T) are in position to use their local real estate assets as a cache for interesting content, low-latency gaming, and AR. If their content is compelling enough, they could benefit from their own edge cloud. However, most operators don’t have exclusive content or any other compelling reason to create their own cloud, and they will be better served by hosting AWS or Azure in their edge data centers.
The mobile operators can resolve the conflict by hosting the cloud guys in their local data centers, essentially becoming colocation hosts themselves. Many are starting to develop this idea, or working with their partners to set up local colocation spaces.
How big are the opportunities? Glad you asked. Mobile Experts recently forecasted a rapid rise in revenue for three main apps: Gaming, automotive HD maps, and industrial analytics. Industrial augmented reality and robotics also hold big potential, but may drift away from the mobile operators if they act slowly and private LTE/5G systems take hold instead. Our breakdown by level of latency is very revealing, and shows that the best strategy for operators is to invest in Central Office and other local micro data center sites as a platform to host multiple cloud players. This way, the operators can keep their investment low, and put the burden of heavy computing investment on the big web players.
Here’s the point: The cloud ecosystem and the mobile ecosystem can both benefit from a new revenue stream, and enterprise IoT offers an opportunity to add to the advertising-based economy of today’s internet. Everyone will need to work together to make it happen.