Guaranteeing network stability has become more and more complicated with the advent of advanced networks. Each new generation of technology and infrastructure has brought new challenges that make providing assurance increasingly challenging, from new air interfaces to virtualized infrastructure. This is now such a big problem that humans alone cannot monitor and assure performance. Now as 5G is rolled out worldwide, networks are about to become more complicated once again, with new edge compute, network slicing and open platforms to engage a wider ecosystem. And it brings with it a huge promise of always-on devices, with many billions of machines and sensors deployed across many different industries. Operators must look to new assurance technologies if they are to provide the network stability required for 5G.
The visibility challenge
Networks today are doing a lot more than they did two decades ago. There are more customers, huge variations in devices, and more data is being transmitted - including non-mobile (IoT) connections, different applications, and countless other strains. This increasing number of variables makes managing network performance very challenging for operators. Combined, there is so much wild data swimming across a network that it has become increasingly difficult to get a full and detailed picture of a network’s performance. In fact, the vast majority of network outages go unreported and undetected.
This issue is best illustrated in users’ ‘on the ground experience’ when a device displays a full signal or connection to the network, though they are seemingly unable to connect. To the user, the network isn’t working, but the operator is unaware of the issue, thus leaving the user to report the problem.
Last year Heavy Reading published a report highlighting the scale of this problem. A key insight from its findings revealed that the majority of network degradations are identified by calls to operator customer support and not by existing monitoring systems. While some customers flag network outages to operators, this feedback isn’t a reliable or a sustainable means of assuring a network as it only provides the parts of the picture a user can see. This is especially true as many outages don’t last a great deal of time. Yet it takes customer experience monitoring on average 15-45 minutes to alert operations teams to customer-impacting degradations. Faults lasting a few minutes or even a few seconds are missed because they are so small that they fall through the cracks. This leads to a skewed picture of what is happening across a network, and can accumulate over time before triggering a catastrophic event.
Aside from this reliance on customer feedback, outages are also difficult to detect because operators now oversee a plethora of network performance data. There are tons of information travelling around a network at any given time, but it requires substantial human “horsepower” to analyze all of it, and in reality, is too much for humans to process and manage. The root of this issue isn’t the quantity of data, but rather the current processes and toolsets for detection of faults.
The ‘little data’ problem
Operators do already have many of the pieces of the puzzle to efficiently detect an outage, but they lack the human resources to put that puzzle together, with the data needed to provide a full picture of the network scattered and siloed across different parts of an organization. Information about customer care, marketing and the network are stored separately, but to get an accurate picture, all need to be interlinked in one central interface and their interdependencies fully understood. As a result, operators are facing a ‘little data’ problem; at the moment they can get an approximate view of the health of the network, but specific faults are hard to find because those are often hidden in small pieces of data that are tucked away from their sightlines in an unruly mountain of “big data”.
That’s not to say operators are completely blind - far from it, there is analytics working behind the scenes to make sense of this data. But this analysis takes time. ‘Scaling’ these current capabilities isn’t enough as it won’t necessarily provide the required correlations, understanding of interdependencies and detailed view of a network’s true performance across all its nodes and layers. Operators require an overview of resources, operational domains and performance data to provide an assessment of true capability.
The advent of 5G
Up until this point, the issue has been a workable challenge for operators, with approximations, averages, and “experience-based” assessments providing the core of performance management. But now with the advent of 5G, this challenge is about to become much larger and completely unavoidable. 5G is much more than just a new generation of technology. 5G is expected to enable applications like real-time patient monitoring in the healthcare sector, distribution and logistics management and product-line robotics for the manufacturing sectors, and innovation in public transport systems. Almost every sector is set to be touched by the reach of 5G and they will all require an uninterrupted network that runs smoothly according to the specific operating parameters each of them need. To put this in perspective, the GSMA estimates that there will be five times more machines than humans connected to networks by 2025.
This huge increase in machines is set to be the most significant challenge for network assurance over the next few years. Why? Because machines have far different needs to human network users. If a human has a poor connection, they’ll contact their operator and let them know, or simply put up with a short-term outage until it gets fixed. But a machine is intolerant of failure, and if its connection fails, then it will too - and the service it supports. This could be devastating in genuinely “mission critical” applications like personal health monitoring or autonomous vehicles.
With so many machines expected over the next five years, operators need to be rethinking service assurance if 5G is to work for connected machines as planned, and capture the revenues they’re expected to generate. But since operators often lack the human horsepower to make sense of network performance data, they must look to solutions that can automatically detect, recognize and resolve faults as they happen. Network assurance must become automated on a large scale to avoid service interruption and ultimately protect these machines from failure.
Building self-sufficient service assurance
In the era of 5G, networks need to be self-sufficient and able to detect outages without relying on humans. Self-sufficiency is entirely possible, though it must be all-encompassing to deliver the desired results.
To create these self-sufficient networks, operators need to implement automated assurance that can combine problem detection, diagnosis, and the means to resolve problems and outages. These systems must be rigorous and offer features like prioritized assurance flags which provide alerts for more critical and pressing user experience problems i.e. the most urgent disruptions. And critically, these automated networks need to go beyond detection and flagging and also become predictive.
Predictive assurance uses artificial intelligence and machine learning to find and resolve problems before they happen. It can do this by being fed data of past outages and user experience problems, then using this information to identify and resolve potential flaws before they occur. These building blocks are the foundation of a truly intelligent and self-sufficient network.
This level of protection is essential for 5G-connected machines, where network outages must be prevented before they happen to protect machines from failing. Operators need this deeper level of insight to deliver on the promise of IoT, connected machines and smart cities.
There is also a big opportunity here for operators. By gaining more control over the network, operators can deliver a better customer experience, increasing customer retention. This leads directly to increased revenues and reduced customer acquisition costs for operators through reduced churn.
Assurance needs to move into this proactive model if it is to work for connected machines and guarantee the service quality they need. As the world moves ever-further into the 5G era, assurance has to be self-sufficient, self-learning, and automated. A new approach has been needed for some time, but as 5G arrives, this approach is now essential for operators to deliver true network stability.
