Cisco's picks for top 4 emerging networking technologies

Why machine reasoning over machine learning

Cisco is directing a significant portion of its development dollars on machine learning (ML) algorithms. The idea is to stuff networks with the intelligence necessary to secure and manage them as billions of mobile and IoT devices come online over the next few years.

But more potent than ML is machine reasoning (MR), which promises algorithmic software capable of performing tasks and debugging problems like network engineers do today, Apostolopoulos said.

In general, MR means programming a computer with logical techniques, such as deduction and induction. The MR system has an inference engine that draws intelligence from the relationships between facts and rules within a specific domain, such as an area of network management.

The information loaded into an MR system typically comes from the practices of experts in a particular field. Computer scientists are developing ways to turn that data into a language that computers can read and then use to draw intelligent conclusions.

Proponents argue that MR is more effective than machine learning. That's because the latter uses algorithms to perform only statistical analysis on facts to come up with a recommendation.

"The computer can try to debug a problem in the same way that the experts would do it," Apostolopoulus said of MR. "It can do a much better job at identifying the root causes of problems than machine learning can."

Advanced technologies like MR will become critical to engineers trying to troubleshoot networks serving numbers of devices exponentially higher than today. IDC predicts that just the number of IoT devices connected to the internet will grow almost 29% annually, reaching 42 billion by 2025. Together, those devices will carry 79 zettabytes of data.

Optical networking

The $660 million acquisition of silicon photonics company Luxtera in December 2018 showed that Cisco wants to move faster into optical networking. Luxtera's transceivers for data center switches and routers turn their electronic traffic into much faster light beams to run over optical fiber.

Cisco's interest in optical technology is to eventually use it within its line cards and application-specific integrated circuits. Without faster data speeds, both will become bottlenecks in data centers struggling to keep up with increasing traffic flows.

Apostolopoulos said widespread optical networking would contribute, for example, to making multi-user augmented reality mainstream. Today, the enterprise network is too slow for multiple people to see and interact with the same augmented reality (AR) content at the same time.

However, in the future, "we'll have very high-quality, high-performance, low-latency wireless to many of these devices at the same time," he said. That will make multi-user augmented reality possible.

Multi-user AR will help students visualize concepts in math and science while also providing corporate workers with better diagrams while repairing in-house or customer equipment.

"CEOs at different companies talk about how they believe this field will be very important and transformational in many ways -- similar to how the iPhone transformed mobility," Apostolopoulos said.

CBRS and cognitive radio emerging networking technologies

Apostolopoulos is optimistic that trends in the communications industry will deliver efficiency in radio spectrum management. Spectrum is finite and shared, so effective management is the only way to serve the billions of devices worldwide that enterprises will add to wireless networks.

The technologies that will help make future networks more robust include cognitive radio and the Citizens Broadband Radio Service.

Apostolopoulos expects the industry to deliver cognitive radio advancements that will significantly increase the number of devices on a wireless network. Cognitive radio is a form of wireless communication engineers use to plant intelligence into a transceiver.

The smart device can detect unused communication channels and instantly move traffic over them. The result is the optimal use of available radio frequencies and minimal interference among network users.

"That will become more important over time," Apostolopoulos said, particularly as the number of internet-connected devices grows.

CBRS will add to the amount of spectrum monitored by intelligent transceivers. The service operates in a 3.5 GHz band from 3550 MHz to 3700 MHz the prime spectrum for 5G services. The Federal Communications Commission set the rules for shared commercial use of CBRS in 2015. The FCC plans to hold an auction of 70 MHz of the CBRS spectrum on June 25.

The FCC expects carriers that license the spectrum to make their 5G networks much better at penetrating buildings and walls. Today, the emerging next-generation wireless technology operates at a frequency that can't pierce objects easily.

Apostolopoulos expects enterprises to license CBRS spectrum or contract with a service provider that does to set up private wireless networks on the factory floor. Those networks will provide the necessary security to run manufacturing robots and other machinery.

"In your manufacturing plant, you'll know that you're the only person who'll be using that spectrum because you're the only one who has access to it," he said.