Like other massive computing tasks, signaling is moving to the cloud. But because it involves real-time communication between base stations and nearby radios, it can't be done on giant regional data centers like the ones Google and Amazon operate, one or two to a country. If signals have to travel halfway across a continent, the tower's conversation with the device will time out and start over again.
Instead, the computing needs to be virtualized and spread out among facilities closer to the cell tower. Things will get even more strict with 5G networks, where the industry consensus is to keep latency to 1 millisecond. That's necessary for things like remote video monitoring and control of equipment in near real time. So signals and traffic will probably be processed in various data centers about 10 kilometers and 100 kilometers away from a base station, Weldon said.
Cells themselves will get distributed, too. Smaller cells closer to users let carriers reuse the same frequencies many times over in the same area that one big tower serves now. The age of the small cell has been delayed since it was first predicted a few years ago, partly because the complicated work of mounting a base station has to be done many times over. But the proliferation of new connected devices will make it necessary, Weldon says.
Small cells will do one more thing to improve networks: They're packed in close enough to use very high frequencies that only work at short ranges of 100 meters or less. Researchers at Alcatel-Lucent and other companies are studying these millimeter-wave frequencies and the U.S. Federal Communications Commission will debate a proposal for opening up some millimeter-wave bands for mobile service at a meeting next week.
Sign up for CIO Asia eNewsletters.