Wireless broadband is hard. Sure, we rely entirely on it now, but the ability to move packets around a network took years of R&D and engineering work to make happen. First we did it for cellular, then we did it again for Wi-Fi so we could ensure good connections on our planes and trains. Now the question is whether we’ll need to adapt it for edge computing.
Several presentations at the Edge AI Summit in San Francisco got me thinking about this idea of the mobile edge. It was partly due to the big focus on autonomous vehicles, but it also stemmed from the knowledge that 5G connectivity with a lot of capacity and low latency is coming, and vendors want to prepare for the new applications such speeds might engender.
A dedicated mobile edge would solve the challenge of processing large amounts of data really quickly and share that data among a relatively local network. It would require fast wireless speeds that can transfer a lot of data and also likely some type of nearby computing.
First let’s hit the most obvious application that would need a mobile edge: autonomous cars, especially those that want to communicate with other cars on the road. Rahul Vijay of Uber told the audience that each self-driving Uber car generates about 4 terabytes of data a day. A terabyte is 1,000 gigabytes, or roughly 250 HD movies. Today there are 300 to 400 such cars roaming the world, but obviously Uber hopes there will be more.
Moving 4 terabytes of data across 400 cars generates 1.6 petabytes of data a day, which is an obscene amount of data to transfer over mobile networks. Thus, edge processing is already taking place to produce insights required to drive the car, on the car itself.
In addition to much of that data being processed on the car, Vijay said Uber is also creating storage depots with fat connections that can handle the uploads of multiple cars. In areas where a cluster of self-driving cars may overwhelm the network, he suggested use of a mobile data center packed into what looked like a minivan to help process the data.
The data center on wheels would rove around to meet demand. What was surprising to me was that those 4TB per day are handling the automobile’s driving. Other presentations showed the interior of a self-driving car becoming a content mecca, with passengers watching videos, perhaps playing AR or VR games, and data from sensors in the car’s cabin providing information that could influence ads shown, services offered, and even pricing.
This vision of the car holding a consumer captive was disturbing, but fits within the hopes and dreams of car companies and advertisers alike. Modar Alaoui, CEO of Eyeris, which provides computer vision for inside a car’s cabin, said auto manufacturers are looking for services they can add on to the cost of a vehicle’s per-mile charge. Those services might be as sophisticated as offering Netflix movies or even heated seats as a service on a cold day.
However, services and automation in a car’s cabin don’t strike me as things that consume a huge amount of data and by extension, need a mobile edge. Much of that computing can be done inside the car and reside there. But if you imagine VR gaming between people in a car’s cabin, or even in a small geographic area while mobile, then the need for a mobile edge computing strategy becomes more clear.
In that case, said Yves Balchar from LinkedIn, we’ll probably need new protocols to handle data flows in local areas on mobile networks. Today’s internet is a best effort that leads to missed packets and delays that could be catastrophic in autonomous vehicles and merely irritating if one is trying to play a virtual reality game. The IEEE has such efforts under way in manufacturing, but it’s possible we’ll need them in mobile.
And there are other ways. Companies such as Ericsson are already looking at more finely tuned network slicing in 5G networks. Such slicing will set aside dedicated capacity for specific use cases such as autonomous cars, but also medicine and even VR. That technology already exists, although handoffs between towers while they are still dealing with huge amounts of data will likely need some research.
So it’s clear we’re going to have a lot more happening at the edge, and 5G and other technologies may require the development of a highly connected and compute-intensive mobile edge environment. But I don’t think we’ll need it just yet. I’d be curious to hear from you guys on this.