AI and ML in Transforming Network Operations
By Ram Sridharan
The complexity of communications networks is increasing due to traffic growth, and the need to support low latency applications and millions of end user devices.
Operators need to transform their operational models to address the challenges and opportunities this brings. Networks and infrastructures must be redesigned to instantly meet the demands made by applications and devices. Not just a new kind of network — a new way of operating the network. One that dynamically adapts to meet demand; optimizes costs or power consumption; anticipates future failures and self-configures to mitigate or avoid service impact.
This means operators need to increase the intelligence of their network operations, planning and optimization. Machine learning (ML) and artificial intelligence (AI) will play a key role in automating network operations and optimizing the customer experience. Tools and techniques from AI are already finding their way into all corners of the digital landscape. AI and ML approaches are beginning to emerge in domains, addressing network automation, virtualization and cloud computing, fault management and predictive maintenance.
Let’s look at a few use cases where the principle of ML and AI can benefit the operator. “Autonomous Networking” is a continuous cycle of observing, analyzing, and improving actions with little or no human intervention. This would include SDN-based network automation. Big data and real time insights can help operators resolve real-time network events like dynamic control, congestion avoidance, sudden performance degradation, high cost traffic forwarding, etc. Using machine learning on a multi-dimensional data set, traffic redirection could be initiated proactively and ultimately help operators avoid unnecessary capital investments.
Operators are deploying cloud platforms with SDN/NFV as they move towards a virtualized world. Life cycle management of those virtual network functions across hybrid cloud environments poses a major challenge to cloud operations. AI/ML techniques can predict traffic capacity and usage trends, then auto scale and self-heal during outages. These techniques can also help determine VNF placements based on company security and compliance policies.
Faults will continue to be a fact of life when operating a network. Tomorrow’s network operations centers will apply machine intelligence techniques to map current conditions to historic conditions. Then they can perform intelligent grouping of cross domain alarms using pattern matching techniques to detect incidents, recommend actions, initiate remedies and finally identify true root causes, not mere correlation. Causation analysis or true root cause analysis will use large scale ML models to determine the causal path from the observed fault state to the root cause. The network management will become an almost autonomous operation — imminent fault conditions will be predicted, and corrective actions performed. Such a knowledge database can also help to proactively detect anomalies (a.k.a. predictive anomaly) leading to implementing predictive maintenance.
The progress of AI and natural language processing is now fostering a wave of innovations affecting how people communicate, how they work, how they interact with businesses. Natural language processing is a key enabler of autonomous customer care solutions. ML applied to historical customer care workflow data on an individual subscriber level can help care agents (humans or bots) provide a better customer experience. Similarly, ML can also be applied to trouble ticket data to group issues based on category, corrective actions taken and effectiveness of these actions and provide a closed-loop feedback to improve problem resolution methods and times. Intelligent digital assistants using machine intelligence will automate service activation and maintenance tasks to help field techs do their jobs more effectively and efficiently.
In conclusion, while technologies such as data virtualization, programmable network and predictive analytics are being deployed today to provide some level of automation, the use of cognitive AI/ML will make networks “deep learning networks.” AI techniques can help detect historical patterns, correlations in big data sets and implicit models. By combining human intellect and creativity with the computing power AI offers, self-improving intelligent algorithms can be developed to help create the networks of tomorrow. These techniques along with the current transformation in networking and infrastructure, can help realize a highly responsive and fully autonomous network — one that can self-optimize, self-manage, and self-operate.
Ram Sridharan,
CTO Office,
MSO Segment,
Nokia
Ram Sridharan has over 25 years in the ICT industry. As part of the Chief Technology Office MSO Segment at Nokia, he represents Nokia’s long term vision and current technology, products and solutions. Ram is involved to acquire new customers and forge strategic third party relationships. His areas of focus include 5G/IoT, Cloud/NFV/SDN, big data, CEM and security. Ram holds a BSCS from Concordia University, an MBA from University of Dallas and is a project management professional.
Shutterstock
