In the distributed edge, the processing power needed for many applications will be done at the network edge instead of a centralized site. With data not needing to travel as far, latencies are reduced, and network bandwidth is improved.
I talked recently with Michael Recchia, Global Telco Solutions Architect at Red Hat. Recchia has decades of experience working at AT&T and Verizon and was one of the original architects for the radio edge-cloud—particularly for the open RAN intelligent controller. He’s an expert with hands-on experience in strategy, architecture, and the design and testing of programmable network deployments with cloud-native edge solutions in the open RAN environment.
Why Process Data at the Edge?
Most of the data that’s processed now needs to be done at the edge of the network or as close to the enterprise data center as possible. We don’t need to burden the network itself with that data. If you’ve got a data set and 99.9 percent of that data is being processed locally, you’re wasting network resources by sending it to a centralized cloud—(Recchia likes to call this the distributed intelligent edge, where 99.9 percent of the data is processed at the edge of the network because all the benefits accrue there). For example, take video rendering at the edge of the network. We can do it within a sub-millisecond when we need to at the edge. We know that things like 360-video virtual reality perform poorly in the 50-millisecond range, which we often have now with LTE. It works great with 5G, edge processing, and sub-millisecond communications.
We can do things like deep learning and machine learning at the edge of the network where the data is. And then, if somebody needs to know about it, we can aggregate that data and then send insights northbound. Another example is IoT devices. With 5G, we can have a million devices per square kilometer versus 10,000 in prior generations. Because of the processing power of the cloud environment, we’re able to process so many more IoT devices per square kilometer.
Network Slicing Life Cycles
Enterprises should be greatly interested in network slicing. You’re able to build your own private, isolated network that you could either manage yourself as an enterprise or you could outsource the management. You have, in essence, your own virtual network. You’ve got a lot of flexibility in how you conduct business. Things can change from day to day. You can bring down a particular virtual network or a slice. A new slice can be commissioned that offers different types of service than the slice you had the previous day. In essence, these slices take on a life cycle. They’re commissioned, they have a life span, and then they’re decommissioned, and then perhaps you have some other things that you want to do, and hence, you build another slice.
Network Slicing Business Models
As a mobile network operator (MNO), I can contract out to an entity interested in providing some class of telco services or subclass of telco services. They’ll engage with the MNO. They’ll work out a business agreement, and a slice will be commissioned. The enterprise that operates the slice will be an MVNO. They come into it with full knowledge that they’re getting a virtual network, they’re going to be operating that network, and there are certain costs associated with it.
There are also other business models where an enterprise brings up a particular slice as a service. They’ll use that for a period of time. They may not necessarily operate it. There are models under development where they can bring up a slice, and another entity manages it. They just provide the services from that slice.
Battles Get the IoT
Many use cases will need the low latency 5G provides. Take the Department of Defense, where there’s a lot of interest in private gNodeB. The gNodeB would be mobile, and IoT devices would be placed on other vehicles that would be part of the field. And there would be several soldiers and other personnel equipped with IoT devices on their uniforms and equipment. That will be a lot of data coming in from all these devices, and there’s a lot of data to determine what’s actually happening and what could happen.
Distributed Neural Networks
With up a million IoT devices per square kilometer, there will be a lot more data than is traditionally being processed. With the ultra-low latency offered by 5G and the higher bandwidths available, a burgeoning area in the distributed intelligent edge is the Distributed Neural Networks (DNN). DNN, in addition to deep learning and machine learning, will help us crunch the massive amounts of data that are being generated. We will need algorithms that can process large amounts of data and then help make decisions about what’s going on now and what to do next.
To learn more about the distributed edge and how it is impacting private 5G innovation, listen to the podcast hosted by Ashish Jain, CEO and Co-founder, PrivateLTEand5G.com and KAIROS Pulse. Our podcast guest is Michael Recchia, Global Telco Solutions Architect at Red Hat.