Intel is getting behind the development efforts for the upcoming next-generation 5G networks, but not perhaps in the way that might be imagined. Rather than the wireless network itself, Intel’s focus is on the back-end infrastructure to support this, which it sees as a great opportunity for its processors and other datacentre technology.
At its annual Intel Developer Forum (IDF) event in San Francisco, the firm discussed its vision for a 5G world, where mobile networks will be much more than just a technology for making phone calls and will instead serve a bewildering array of devices – such as sensors connected up via the internet of things (IoT), autonomous vehicles, and other network-enabled infrastructure.
“If you look at 5G, we view it as the age of the machines, and everything that enables a fully mobile, connected, intelligent society,” said Asha Keddy, vice-president in Intel's mobile and communications group and general manager of the company's standards and advanced technology team.
For this reason, 5G is proving a harder nut to crack than earlier mobile standards, because it will have to serve a wide array of use cases, from connecting billions of small low-power sensors to low-latency streaming of high definition video, to users with tablets or smartphones – and almost certainly some applications that have not been thought of yet.
“We don’t know what the future will look like. If you look back to 2G networks, it was all about phones and voice, and then we started having the data revolution. By the time we were rolling out LTE [4G networks], we started requiring things that hadn’t been planned for when LTE was originally designed,” added Keddy.
Initially, 5G networks are likely to be made up of multiple wireless technologies, including new air interfaces operating at millimetre wavelengths with data rates measured in gigabits per second, and with the aim of keeping latency down to below 1ms for an end-to-end round trip delay.
However, while the exact standards are still a work in progress, AT&T senior vice-president for wireless network architecture Tom Keathley said that he expected to see some early pre-standard deployments as soon as next year, with full roll-outs slated for some time following 2020.
Unlike earlier generations of mobile network technology, 5G will be built upon infrastructure making extensive use of technologies such as software-defined networking (SDN), network function virtualisation (NFV), and advanced data analytics. These will be required to enable a service model that can slash the cost of carrying the huge volumes of data that are expected to be generated by billions of connected devices and more widespread use of high-definition video, according to Intel.
“One of the things that is important to keep in mind for 5G is that thebackhaul network, from the base station back into the core, is already in a transformation, and this is because, as those devices have more and more data passing through them, the network itself has to transform,” said Lynn Comp, director of market development at Intel’s network products group.
Not surprisingly, Intel sees this an opportunity, as it has been putting much development effort over the past few years behind SDN and NFV initiatives, as well as working to optimise its processors for handling analytics workloads.
But it is not just in the cloud or datacentre, Intel’s traditional area of expertise. The firm is now talking about the need to have intelligence distributed throughout the network, and especially at the network edge, to support some expected use cases for 5G networks.
“What you’re seeing is cloud-like capabilities going from the datacentre, into the network and out to the edge, helping to process and add intelligence and insights about what data should go where,” said Comp.
The need for this will be driven by applications where a rapid response is called for, according to Intel. For example, when monitoring industrial processes or controlling autonomous vehicles, sending all the data back to a remote datacentre for analysis may simply cause too much of a delay. Having some local analytics capability would be a better solution.
If Intel is right, then processing power is not going to be concentrated inside mega datacentres operated by the major cloud providers, but will be spread across the network with base stations and other nodes becoming miniature datacentres in their own right, perhaps caching content locally to cut latency. And naturally, this will also see more Intel chips being sold.