AI and 5G: Entering a new world of data

The deployment model of vendor-centric equipment cannot sustain this exponential growth in traffic.

big data messaging system / information architecture / mosaic infrastructure
Stinging Eyes (CC BY-SA 2.0)

Today the telecom industry has identified the need for faster end-user-data rates. Previously users were happy to call and text each other. However, now mobile communication has converted our lives in such a dramatic way it is hard to imagine this type of communication anymore.

Nowadays, we are leaning more towards imaging and VR/AR video-based communication. Therefore, considering such needs, these applications are looking for a new type of network. Immersive experiences with 360° video applications require a lot of data and a zero-lag network.

To give you a quick idea, VR with a resolution equivalent to 4K TV resolution would require a bandwidth of 1Gbps for a smooth play or 2.5 Gbps for interactive; both requiring a minimal latency of 10ms and minimal delay. And that's for round-trip time. Soon these applications will target the smartphone, putting additional strains on networks. As AR/VR services grow in popularity, the proposed 5G networks will yield the speed and the needed performance.

Every IoT device [Disclaimer: The author works for Network Insight], no matter how dumb it is, will create data and this data is the fuel for the engine of AI. AI enables us to do more interesting things with the data. The ultimate goal of the massive amount of data we will witness is the ability to turn this data into value. The rise in data from the enablement of 5G represents the biggest opportunity for AI.

There will be unprecedented levels of data that will have to move across the network for processing and in some cases be cached locally to ensure low latency. For this, we primarily need to move the processing closer to the user to utilize ultra-low latency and ultra-high throughput.

Some challenges with 5G

The introduction of 5G is not without challenges. It's expensive and is distributed in ways that have not been distributed in the past. There is an extensive cost involved in building this type of network. Location is central to effective planning, deployment and optimization of 5G networks.

Also, the 5G millimeter wave comes with its own challenges. There are techniques that allow you to take the signal and send it towards a specific customer instead of sending it to every direction. The old way would be similar to a light bulb that reaches all the parts of the room, as opposed to a flashlight that targets specific areas.

So, choosing the right location plays a key role in the development and deployment of 5G networks. Therefore, you must analyze if you are building in the right place, and are marketing to the right targets. How many new subscribers do you expect to sign up for the services if you choose one area over the other? You need to take into account the population that travels around that area, the building structures and how easy it is to get the signal.

Moreover, we must understand the potential of flooding and analyze real-time weather to predict changes in traffic. So, if there is a thunderstorm, we need to understand how such events influence the needs of the networks and then make predictive calculations. AI can certainly assist in predicting these events.

AI, a doorway to opportunity

5G is introducing new challenges, but by integrating AI techniques into networks is one way the industry is addressing these complexities. AI techniques is a key component that needs to be adapted to the network to help manage and control this change. Another important use case for AI is for network planning and operations.

With 5G, we will have 100,000s of small cells everywhere where each cell is connected to a fiber line. It has been predicted that we can have 10 million cells globally. Figuring out how to plan and design all these cells would be beyond human capability. This is where AI can do site evaluations and tell you what throughput you have with certain designs.

AI can help build out the 5G infrastructure and map out the location of cell towers to pinpoint the best location for the 5G rollout. It can continuously monitor how the network is being used. If one of the cell towers is not functioning as expected, AI can signal to another cell tower to take over.

Vendor-centric equipment cannot sustain 5G

With the enablement of 5G networks, we have a huge amount of data. In some cases, this could be high in the PB region per day; the majority of this will be due to video-based applications. A deployment model of vendor-centric equipment cannot sustain this exponential growth in traffic.

We will witness a lot of open source in this area, with the movement of the processing and compute, storage and network functionality to the edge. Eventually, this will create a real-time network at the edge.

More processing at the edge

Edge computing involves having the computer, server and network at the very edge of the network that is closer to the user. It provides intelligence at the edge, thereby reducing the amount of traffic going to the backbone.

Edge computing can result in for example AI object identification to reach the target recognition in under .35 seconds. Essentially, we have the image recognition deep learning algorithm that is sitting on the edge. The algorithm sitting on the edge of the network will help to reduce the traffic sent to the backbone.

However, this also opens up a new attack surface and luckily AI plays well with cybersecurity. A closed-loop system will collect data at the network edge, identity threats and take real-time action.

Edge and open source

We have a few popular open-source options available at our disposal. Some examples of open source edge computing could be Akraino Edge Stack, ONAP Open Network Animation Platform and Airship Open Infrastructure Project.

The Akraino Edge Stack creates an open-source software stack that supports high-availability cloud services. These services are optimized for edge computing systems and applications.

The Akraino R1 release includes 10 “ready and proven” blueprints and delivers a fully functional edge stack for edge use cases. These range from Industrial IoT, Telco 5G Core & vRAN, uCPE, SDWAN, edge media processing and carrier edge media processing.

The ONAP (Open Network Platform) provides a comprehensive platform for real-time, policy-driven orchestration and automation of physical and virtual network functions. It is an open-source networking project hosted by the Linux Foundation.

Finally, the Airship Open Infrastructure Project is a collection of open-source tools for automating cloud provisioning and management. These tools include OpenStack for virtual machines, Kubernetes for container orchestration and MaaS for bare metal, with planned support for OpenStack Ironic.

This article is published as part of the IDG Contributor Network. Want to Join?

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Take IDG’s 2020 IT Salary Survey: You’ll provide important data and have a chance to win $500.