5G will signify a monumental shift in communications, promising to be the first network to negotiate mobile and fixed networks, for man-to-man, man-to-machine, and machine-to-machine communications. The sheer amount of services and network complexity will require a step up of current network capabilities. Specifically, 5G networks will need to incorporate Artificial Intelligence (AI) and its offspring Machine Learning (ML).
As AI/ML continue to gain steam and the rest of the business world gets on board, current networks are suffering from the lack of capabilities needed. To be honest, today’s mostly manual, static networks are not suited for these advanced technologies. And while agile, self-organizing networks will exist in the future, service providers need to address their digital transformation efforts today, focusing on near-term solutions, to build the foundation for these networks of tomorrow.
This article will explore the near-term focus on what can be done now to help transform the networks to accommodate a future with AI, ML and big data - so that they can become more dynamic and intelligent in the future.
The Heart of the Matter
AI is being hailed as mankind’s next ‘savior,’ regardless of the industry. In telecoms, we hope that AI will help overcome the burning issues that arise from our current manual, static networks. The impetus is 5G. Unfortunately, our static and outdated equipment and frameworks aren’t equipped to take advantage of continuous and dynamic reconfigurability driven by AI/ML, and we need to catch up. But there is hope.
The 5G network needs to be assured, agile and self-organizing, made up of both a programmable transport network and an agile control layer. In order to reach that Emerald City, we’ll need 5 key components:
Artificial Intelligence: The AI hype is strong for just about every technology, and we’re now seeing this spread into telecommunications. And while AI holds many promises, there’s still a lot of confusion around what is possible and what is still out of reach.
We know that there are multiple types of users out there, with different service and performance needs. Additionally, many services need to be deterministically, assured. Until now, service assurance was about best effort, relying on heavy over-provisioning and basic monitoring. In the 5G era, this will likely result in either a vastly over-provisioned, expensive network or poor customer satisfaction. Realistically, it’s likely to be both. Thankfully, we are on the cusp of revolutionizing algorithmic development using AI techniques seen in machine learning (ML). Since ML processes continuously learn and improve, it’s likely that to implement self-organizing networks, we’ll need a combination of several ML approaches to ensure success.
A Multi-Gear Programmable Network: If our objective is now an agile, self-organizing network, we need to take a look at new approaches. Historically, networks were budgeted, implemented and operated one layer at a time, without multi-layer awareness. As we take more of a self-organizing route with ML processes, which looks at all layers of a network at once, we should equip the network with as many “programmatically configurable” technologies to optimize. This would span what we categorize as L0-L1-L2-L3 technologies providing optical, electro-optical, and packet transport functionality.
When planning an agile, self-organizing network, the network nodes need be built with multiple transport and data plane technologies, or “gears” that create multiple degrees of freedom to handle random traffic patterns and network events.
Big Data: Today, data is collected from the network only intermittently for some ‘nice to have’ business intelligence or planning applications. But there is currently no capacity or mean to acquire and process massive data streams, which will inevitably be generated from the self-organizing networks of tomorrow.
With ML processes in place, however, what might be considered as overload for human-centric procedures is the bread and butter of machine learning. It’s built to absorb and process large amounts of data because it can then construct better models for making decisions and steer toward optimized outcomes.
In this scenario of agile self-organizing networks, we need to remember to establish good data policies that relate to which data is gathered, how frequently it is reported, how it is processed, and how it is retained. It is likely we will also need to invest more in real-time data acquisition, storage, and pre-processing functions like data labeling as a front-end to the ML algorithms.
High-Level Guidance: Conventional human-developed automation relies on making assumptions. To implement self-organizing networks, we’ll need to think about a combination of several ML approaches and ensure that systems are constantly receiving feedback so they can be updated and tweaked as needed to ensure desired outcomes are met.
All ML implementations are only as good as the basis on which they are built. So, while an AI/ML brain will be at the core of self-organizing networks, it can only be useful if we create the right infrastructure. But how do we guide the ML control layer? It will require a combination of automated and human guidance by implementing key performance indicators (KPIs).
The ML control layer will use KPIs generated automatically from network data to evaluate whether its actions, say for service provisioning, traffic balancing, or preventative network maintenance, are achieving expected results. Since machines don’t come equipped with intuition, the added layer of KPIs will ensure that success can be measured and will help the humans overseeing that success to guide the ML control layer as needed.
Human Support: As we’ve touched on already, let’s not leave all humans out of the equation. Humans will need to fully understand their own customers’ evolving needs so they can best design new services to fulfill these needs, with the help of AI/ML as we take steps toward self-organizing networks.
The result will be a network of the future, ready for 5G services. To support this brain (and the beating heart of 5G) we need: a multi-gear, dynamically controllable network, continuous and massive data on how the network is performing, and an ability for the brain to evaluate its own performance and obtain corrective guidance.
The Well-Paved Path to 5G Readiness
We know that the path toward 5G will be challenging, but there is a gleaming light at the end of our path. With a new focus and dedication toward the underlying infrastructure, we can ensure that we’re paving the way, one step at a time, toward 5G success.
Knowing that agile, self-organizing networks will exist in the future, service providers must be determined in their digital transformation efforts now in building the foundation for these networks for a bright future with 5G.
It starts with a near-term focus on adding multi-gear programmable networks and putting in place the Big Data systems to report on network performance and events. This includes mapping out an incremental approach to AI and ML, switching small sections of the network where it makes the most sense. Once processes get smoother, this can expand to other functions like service provisioning and more.
We know that to ensure 5G success, we’ll need the help of a lot of AI and machines, but we can’t discount the humans completely. After all, ML will still require a set of checks and balances to ensure the right data is being captured and things are running as they should before we turn over all control for self-organizing networks. It’s true, there are already several service providers taking the necessary steps to get there - and they’re starting to lay the foundations for a network to support true 5G. By taking those incremental steps now, and once the industry gains more experience with ML network control, it’s not difficult to imagine that all the world’s networks might one day be agile and self-organizing.
