Unlocking the Secrets to Next-Generation Service Assurance
Service assurance, broadly speaking, faces three significant challenges as it looks ahead. The first is a continual growth in network traffic. As it stands, operators are seeing a growing demand for delivering a diverse range of data-rich services. 4G is already widely available with some subscribers being in remote parts of the world and still achieving access to video content and OTT services through their smartphones.
According to figures obtained from GSMA, in 2017 there were 5 billion unique subscribers, with this figure expected to reach 5.9 billion by 2025. This means that 75% of the world's population will be connected to the Internet. Of those connected to the Internet, much of the data consumed is attributed to video. In fact, by 2021 it is estimated that 82% of all traffic on the Internet will be related to video services. Now, with the introduction of 5G, this will only expand the range of services offered and increase demand on the network.
The second major challenge comes from a realization of the importance of security, which is in part due to an increase in breaches to information on the cloud, and new privacy regulation (GDPR) are driving the trend towards encryption. Indeed, over 50% of all data is now encrypted, and YouTube sends 97% of its traffic over encrypted connections, with an expectation for this to be the goal for all video traffic across the Internet. Operators are finding themselves in a blind spot as a result of increased encryption, which poses a major challenge to operators who are looking to understand their customer's quality of experience (QoE).
This has given rise to the third challenge for operators which is finding a way to differentiate themselves from the competition. As prices for services are at an all-time low and have appeared to have leveled out, operators need to find ways to stand out from the crowd, and they do so by delivering superior service. Understanding the QoE, therefore, becomes a higher priority, and thus the fact that operators find themselves in the dark due to an increase in encryption is proving a challenge to monitor and assure. Added to this is the fact that video services are becoming more and more popular and operators want to find a way to tap into this new revenue stream by offering streaming services as part of an operator package.
These challenges come together to create a perfect storm for operators who must take a new approach to monitor their networks.
Cracking the code
Operators need to think out of the box as to how to combat these challenges. To start with they must realize that with the introduction of 5G and the next generation of services and technologies that come with it, there will be too much traffic on the network to monitor it all. Not only is it a nearly impossible task, but it would also be a waste of an operator's resources, both for their workforce and financially, to try and achieve this. Operators, therefore, need to take an on-demand approach to service assurance and network visibility, using smart sampling and filtering on specific data sets. Doing so allows operators to decide, on the fly, what areas of the network they want to perform in-depth analysis. This enables the ability to zoom in on high priority issues or selected subscriber groups over a specific or extended period, as well as to give the option to slice the network by location, service, application or device. On-demand doesn't mean that operators lose sight of certain areas in the network. Instead, it offers the operator a smart perspective giving them the option to focus on their highest priority issues, creating a more efficient network.
The challenge of understanding encrypted traffic requires an entirely new approach to measuring a user's QoE. If we consider that most data consumed is attributed to video, and over 90% of video traffic is encrypted, this leaves the operator in a blind spot as to their subscriber's experience. With operators offering monetized solutions to streaming services and having a growing interest in outperforming their competitors with the service they deliver; the onus is on the operator to ensure the QoE is of a certain standard. This compounds the need for operators to have full network visibility into encrypted traffic. Without being able to decode the session, operators need to be able to find a way to understand and monitor the quality of service delivered. The solution to this is to adopt a heuristic approach to the data which is based on a number of assumptions and uses these together with Machine Learning to analyze patterns of data. Once these patterns are understood, the operator can begin to set Key Quality Indicators (KQIs) and measure the QoE.
To collect the necessary data required to understand the patterns, service assurance providers can gather statistics on specific metrics alongside a crawler which calculates KQIs and at the end of the session will prompt the user to enter the real customer experience observed. Thousands of these samples are required to generate enough information so that Machine Learning can develop algorithms for fine-tuning these KQIs.
Video is just one example of how heuristics can be applied to deliver next-generation service assurance. Gaming, which for enthusiastic gamers requires ultra-low latency and high throughput for real-time reactions for multiple players at once is also a monetizable service. Operators can offer packages which will guarantee a minimum and maximum latency and therefore will want to monitor the speed and quality of the latency. This will enable the operator to meet their agreed SLA as well as deliver a higher QoE for the gamer.
Monitoring tethering connections, which enable one device to access the network via another connected device, is another area where Machine Learning can be applied to deliver next-generation service assurance. The operator has an interest in knowing the volume of traffic that is being carried through tethering, whether this is a mobile device connecting to a WiFi hotspot or a laptop connecting to a mobile device. Using Machine Learning, operators can detect traffic originating behind a hotspot, and the type of traffic, so whether it is from a mobile device, fixed line broadband or a WiFi/ Bluetooth hotspot. This is important to know as operators can then target the user who is tethering and offer them a tailored solution to encompass their tethering needs. In addition, operators can then see which OTT applications are being used through tethering, enabling the operator to understand the demands being placed on the network, contributing to the operators maintaining full network visibility.
This brings us to the third challenge of operators finding a way to differentiate themselves. The need for the operator to demonstrate their added value lies in delivering a higher level of service. QoE management has become a more significant challenge, between squeezed prices and higher demands on the network, compounded by the latest hurdle of understanding encrypted traffic, making the point once again that a new approach needs to be adopted. Operators need to adapt their techniques for monitoring the QoE, whether that is taking an on-demand approach to traffic monitoring or using heuristics and machine learning to understand encrypted traffic.
A vision for the future
As we move towards a new age for connectivity, operators are transitioning to a 5G ready cloud. Part of that transition is moving services towards the edge so that they can provide the lower latency which will be a requirement for many of the services offered with 5G. Smart sampling, filtering, and load balancing is also a necessity in moving towards an on-demand approach. There will also be the challenge and cost of finding a data center large enough to store all of the historical data. Operators already realize that trying to monitor and assure such high volumes of traffic is not resourceful or necessary, and so an on-demand approach must be part of their strategy going forward.
Together with a heuristic approach to data monitoring, which utilizes machine learning, operators can start to gain insights into various types of traffic, from OTT to encrypted traffic. Machine learning, while still in its infancy for monitoring solutions enables operators to apply the necessary KQIs to help measure performance. There is always room for improvement, fine-tuning the heuristic models together with the algorithms for the KQIs, but the more data the operator can gain on the network performance, the faster they will be at identifying problems and proactively troubleshooting, ensuring a superior QoF for the customer.