[powerpress]

Peter Rysavy returns to the podcast to discuss his recent paper, How “Title II” Net Neutrality Undermines 5G. Rysavy is a well known wireless engineer, consultant, and educator. He discussed the promise of 5G with us a year ago, emphasizing the new applications unleashed by a network with high basic speeds, low latency, and lower cost per gigabyte.

The 5G architecture – still a work in progress – builds on a set of design goals based on use case analysis. Every network is good at doing at least one thing, and for many architectures that one thing is a far as it goes.

But 5G aims to do many things: provide basic Internet service, support Internet of Things devices, carry payloads 100 times larger than those we use today, serve crowds assembled temporarily for major social events, and enable cyber-physical systems such as self-driving cars with the ability to react to real-world events faster than humans can.

We don’t know how to support every 5G use case today

We know how to do many of these things today – carrying massive payloads with high reliability it well understood. And we know how to provide extremely cost-sensitive applications with network access. But we it comes to life-or-death applications that need extremely low latency, we have a ways to go.

It’s not that we don’t know how to do this on dedicated networks; we do. But providing time-sensitive applications with the service they require on a network infrastructure that’s also serving other apps of the same kind can be tricky, especially when machine vision is involved.

Rysavy’s answer is “network slicing”, a technique that makes one physical network look like multiple virtual networks to applications. He explains this concept in his paper:

5G needs QoS management, not only for traffic prioritization to support mission-critical applications, but also to enable a fundamental capability in its architecture: network slicing. Network slicing, implemented through virtualization, will allow an operator to provide different services with different performance characteristics to address specific use cases. Each network slice operates as an independent, virtualized version of the network. For an application, the network slice is the only network it sees. The other slices, to which the customer is not subscribed, are invisible and inaccessible. The advantage of this architecture is that the operator can create slices that are fine-tuned for specific use cases. One slice could target autonomous vehicles, another enhanced mobile broadband, another low-throughput IoT sensors, and so on.

Instead of solving coordination problems one at a time, network slicing solves them in bunches. First, it isolates applications of different types from one another, and then it resolves contention among equal applications. In the first instance, it can use prioritization; in the second, harm mitigation is key.

What does this have to do with Title II regulation?

Title II provides regulators with tools to resolve contention among users of the same application, such as telephoning or browsing the web. But it lacks tools to resolve contention between users of different applications drawing resources from a common pool.

The norms for the new kind of resource contention will be developed in the field as we learn from 5G networks. It’s vital that regulators allow these new norms to develop in real marketplace conditions.

Recent rulemakings by the FCC haven’t given us much confidence in the regulator’s ability to study new systems before imposing traditional procedures.

Highlights:

• The idea of 5G is the provide an extremely flexible networking fabric that can address an extremely wide range of use cases.
• Network slicing isn’t new, but it is a lot more sophisticated that crude time-division multiplexing.
• Video transcoding is a change in the form and content of video communication.
• The Internet is a huge information processing system. Even simple routing requires information processing at each router node.

Enjoy. And tell your friends, tell your mom, tell your dog.