Times are a-changing in the mobile network industry. While the last decade saw the evolution of networks focused largely on the twin KPIs of capacity and speed for soaring data demands, this could only keep pace until a multitude of various end users, devices and service types came into the picture and began demanding their own custom resources. Now with burgeoning mobile broadband traffic, an uptake in private networks and more industrial IoT devices around, 4G networks no longer stand a chance in combatting network congestion issues. Luckily, 5G is here to save the day - but the big question is, how?

Will fitting in some new 5G base stations set things right? Unfortunately, no. With current 4G networks facing stifling rates of RAN congestion, 5G networks can only make headway if they are equipped with transformed network optimization capabilities. And with new obstacles such as network slicing and guaranteed SLA contracts cropping up, current optimization models are becoming all the more impossible to manage. Of course, operators may attempt to remedy this on the probe-end and by deploying analytical engines to optimize TCP (Transmission Control Protocol) and packet buffering, but it may be the case that these efforts will never be enough.  

Getting out of your comfort zone: optimizing the RAN, core, application and user experience together

In fact, the true solution to this complexity lies behind the coordinated management and optimization of RAN resources, the core network, and user level configurations simultaneously – in essence, it necessitates coordinated intelligence. This in turn, calls for a new and holistic approach to network optimization altogether. An example?

Let us take the case of a network optimization model focused solely on the RAN – here, an optimization engineer may be able to fiddle around with cell level actions like optimizing mobility load balancing. But in a coordinated intelligence model, we can incorporate cell level actions equipped with rich insights populated from user level metrics and take action based on precise geo-located user experience statistics. This provides us with an impressively deep level of granularity that can, eventually, deliver on any SLA requirement. 

In such a coordinated modular system, we can now look further into the host of functions that can complement each other in achieving an ultimate end-to-end network optimization model. For a start, a centralized SON solution could search the network for RAN bottlenecks and take action based on network usage levels. An NWDAF (Network Data Analytics Function) would be able to analyze the traffic and resource usage per slice/service statistics, reporting it back to the required network functions. Alternatively, in absence of an NWDAF, network probes can be installed to provide detailed information into the applications running on the network. And finally, a PCF (Policy Control Function) can dynamically adjust policies at either a user or slice level to meet varying demands and SLAs. As the name suggests, the key here will be to coordinate and create a perfect symphony between components, by enabling them to seamlessly communicate and interact with each other. In this, we are creating a complete coordinated intelligence model - with an end-to-end view of the network, even the absence of an entire functional component will not halt regular network operations, as they continue to optimize forecasted user experience. Unarguably, coordinated intelligence will be the future of mobile network optimization, and will continue to pave the way towards zero-touch networks.

From a practical perspective, we now zoom in on an interesting use case of coordinated intelligence: user level predictive congestion management. A coordinated intelligence model would be able to capture network level measurements and feed these, along with historical metrics, into a machine learning engine in order to forecast network traffic for the next few hours. When it does predict a congestion in this timeframe, it will immediately communicate the issue to core network functions, requesting them to detect for data-hungry applications and users. And with minute-level throttling instantly applied to those applications and users, bottlenecks can be circumvented, and congestions completely steered clear of. All of this would be done virtually, with zero human touch, as a result of the seamless interconnection between the cSON, NWDAF, PCF functions and probes on the network. 

While use cases may vary between deployments, network usage levels and user behaviors, a perfect synchronization between smart network functions is all it takes to create the intelligent networks of tomorrow. And with its latest products and cutting-edge solutions, P. I. Works is well-positioned to not just meet, but exceed such automation-driven network optimization requirements. To learn more about our portfolio of offerings and other opportunities, please get in touch with us at marketing@piworks.net.