As more and more consumers worldwide stream video, work remotely and access social media over mobile network, service providers face an uphill battle to maintain subscriber Quality of Experience (QoE) while meeting skyrocketing bandwidth demands. These pressures are multiplied by the need to support increasing reliance on cloud computing and the exploding Internet of Things (IoT).
At the same time, mobile providers are deploying more fiber to ensure capacity at the edge of the network to support densification and new services, such as fixed wireless, leading to more and more convergence between fixed and mobile. These converged architecture mix different technologies and architectures, adding significant complexity to network management and optimization. And as more money and technicians are needed to maintain service performance and QoE, operators are seeing profits impacted.
As a result, leading-edge service providers are seeking new real-time optimization and automation techniques and processes to maintain performance more efficiently. For example, advances in location intelligence are allowing mobile providers to focus on where subscribers are located, how they are using the network, and what their current QoE is at any given time. This intelligence can be used to enable subscriber-centric engineering and RAN planning for dynamic allocation of resources in real-time.
Overcome Optimization Obstacles
The basic tenet underlying this real-time network optimization is increased automation, allowing real-time improvements in performance while freeing up personnel, leading to reduced OpEx costs. Until recent advances in machine learning and intelligent data analytics, current network optimization methods have been hampered by three evolving network challenges: interdependency, non-uniformity and complexity. As mobile networks continue to evolve, these combined issues have created a network environment that is nearly impossible to optimize manually.
Because many of the metrics used to optimize networks are now interdependent, when parameters are changed to enhance the characteristics in one part of the network, a cascading effect causes implications in other parts of the network. For example, efforts to increase data throughput in a certain area could affect voice traffic characteristics elsewhere in the network.
This interdependency has a detrimental effect on design, since network designs that focus on one Key Performance Indicator (KPI) can differ from designs that focus on other KPIs. The traditional focus on relatively stable voice and text messaging KPIs, such as availability and quality, has been quickly surpassed by the need to optimize for subscriber QoE KPIs for social media and video streaming, such as Time to Content (TTC). Factor in newer services like voice over LTE (VoLTE), and the focus may switch to optimizing the RAN to prioritize VoLTE calls over other data in congested areas. In other words, designs that focus on a specific KPI cannot address the overall performance of the network. This is particularly relevant as networks become increasingly non-uniform.
The New Normal
As more and more consumers cut the cord and ditch their landlines for mobility, the nature of demand is becoming more dynamic. On average, 50 percent of data is consumed in less than one percent of a network area, according to a recent mobile data trends survey. And this area is constantly shifting, with dramatic fluctuations in data demand in any given cell site from minute to minute.
Moreover, the same survey also highlighted extreme non-uniformity of data consumption, with one percent of the overall subscriber population consuming 50 percent of data. So, not only do mobile operators have to contend with non-uniformity of data demand, the non-uniformity is highly dynamic. This growing trend significantly contributes to the overall complexity of mobile networks.
More By The Same Author : What Telcos Need To Do To Realize Their 5G & IoT Dreams
Evolving technologies such as LTE Advanced, VoLTE and heterogeneous networks (HetNets) compound the problem, adding more layers of complexity. As a result, changes can no longer be made to a network layer in isolation without changing the way that layer interacts with other layers. Plus, the number of tunable network parameters is now enormous. Tuning just two parameters on each of 100 cells – where each parameter has 10 possible values – creates 10200 different ways these cells could be configured!
Traditional optimization processes tend to be network-centric, where a technician locates a problem using network statistics, and adjustments are made to macro site parameters to solve the problem. This type of “blind optimization” lumps together different users at various locations, offering limited ability to enhance subscriber QoE. As the network tries to optimize an entire area, this creates an imbalance where some users have more resources than they need, while others experience poor service.
What’s needed now in an increasingly complex network environment is a subscriber-centric approach at a more granular level. Consider, for example, office workers using voice services at their desks during the morning, but relying more on data services while working remotely in the afternoon or evening. The nature of services that are needed — as well as when and where they are needed — is changing all the time. Manual optimization would require configuring the network to deliver an acceptable user experience for different usage profiles, at various times of day, and for multiple locations in the network. Automated optimization, on the other hand, allows the network to respond to the changing demands placed on it by adapting its configuration in real-time.
Let’s Get Real-Time
Of course, getting to the point of fully automated network management and service optimization will take time. However, service providers are already taking steps in that direction, thanks to advances in real-time analytics and machine learning technologies. Subscriber-centric, automated optimization is already possible with solutions that can collect, locate, store and analyze data from mobile connection events. This data is used to create a repository of location intelligence from all subscribers throughout the network that enables subscriber-centric performance engineering, proactive RAN planning and automated performance optimization.
Real-time intelligence and subscriber-centric insight can help service providers manage increasingly complex network and service ecosystems for greater operational and capital efficiency. One particular service provider even used this intelligence to optimize energy consumption, resulting in a 25 percent reduction in energy costs to save an estimated $2 million annually without disrupting service levels.
Putting the AI in Brain
As an increasingly competitive landscape compels service providers to focus on customer experience more than ever before, we will continue to see an accelerating transition from network-centric optimization to customer-centric QoE. Given the highly complex nature of today’s mobile networks, manual optimization processes are hard pressed to meet the needs of modern networks evolving to 5G.
The simple fact is that we are nearing the point of no return, when manual operations will be unable to scale sufficiently to keep up with dynamic traffic patterns. The only way forward is to increase network automation and programmability in 2018 and beyond. Customer-centric analytics and advanced machine learning will provide the foundation for this movement towards increasing automation, infusing the network with the real-time intelligence needed to provision services at speed. Over time, as artificial intelligence (AI) enables more cognitive learning capabilities, networks will evolve toward self-configuration, self-optimization and self-healing features.
But in the meantime, automated network optimization is not science fiction; it’s a vital step toward building tomorrow’s 5G infrastructure. Automation is not a buzzword for today’s network service providers — it’s a lifeline.
By : Sameh Yamany, Chief Technology Officer, VIAVI Solutions