The world’s first, entirely cashier-less store in Seattle opened recently. AmazonGo uses “intelligent” cameras and a powerful set of algorithms to “think”. It is a grocery outfit, but not as we know it. Shoppers simply scan their phones on the way in, grab what they want — and walk out. No need to stay in line for a checkout. The mobile networks industry could soon be replicating this level of advanced, intelligent automation — with zero human touch.
Times are changing
Around 10 to 20 years ago, when most mobile networks were considered ‘young,’ dropped calls were the number one gripe for subscribers, and mobile operators worked hard to remedy the problem. Dropped calls were mainly due to poor reception. You fix the signal strength, you fix the problem.
Those were the old days.
Today, bad Quality of Experience (QoE) might comprise of buffering videos, stalling video calls and slow download speeds – as well as dropped calls. Mobile network complexity has spiraled with advances in technology like 4G LTE and soon 5G. The changes in network access and network core associated with them have also exacerbated. Today, poor Quality of Service could be down to any number of factors. It has become challenging for network administrators to find the root cause of these problems and solve them quickly.
The need for speed
For some operators, dispatching an extra workforce to fix the problem might not help either. Even the best of engineers would struggle to analyze zettabytes and zettabytes of data in real-time and make a decision on coverage blackspots and optimization – in seconds. Traditional tools are not much help either. They lack the power to analyze heterogeneous data at ultra-high speeds. Existing solutions are inflexible, and once a course of action is set, it simply cannot change its specs based on changed circumstances. It is time for change.
One approach is emerging, driven by advances in Artificial Intelligence (AI) and Machine Learning (ML) that addresses these issues: zero-touch network operation, security and user experience. Using new AI/ML tools and techniques, networks can continuously learn by monitoring and analyzing vast quantities of data. In this way, they gradually build up intelligence to self-manage, self-protect and self-heal. For mobile operators the results are tangible — simpler network operation, better user experience, and minimum subscriber churn.
How can mobile operators make this a reality?
Carriers can deploy zero-touch network frameworks using intent-based networking, AI, and semantic telemetry. The framework only requires the “what” from the network administrator. The framework would then instruct the network to take care of the “how” with simple commands delivered via a Natural Language Programming (NLP) interface. The network carries out the required actions and has the intelligence to self-correct, self-optimize and self-heal to maintain the wish-list defined by the network administrator.
There are five key features of intent-based networking:
- Semantic Telemetry – Enables the policy-driven collection of data from remote points to support monitoring, analysis, and visualization based on push semantics.
- Data-Driven Intent Modelling: This intent is derived from using a data-driven ML approach which is the brain of an intent-based network. But creating this brain needs a new and more effective AI architecture for networks that provide real-time ML capabilities. This newer architecture is called “Network AI” (NAI). NAI effectively combines smart data collection, stream analytics and advanced ML techniques to provide massively parallel Complex Event Processing (CEP) pipelines for faster model building and accurate anomaly prediction.
- Intent Federation Engine: Facilitates the creation of end-to-end intent that is federated across multiple domain controllers and translated to their respective domain level intents.
- Closed Loop Automation: Self-learning feedback loops that help the network to self-manage, self-heal and self-protect.
- Natural Language Interface: Network administrators can use human-understandable language to define intent, rather than cryptic programming languages.
Initiatives today — for the future
By adopting an intent-based approach, humans can dictate what the network needs to achieve through intent. Automation and AI will allow the network to monitor, self-learn, and run reliably. The network will understand business intent and continuously align with it. It will use context and AI to continually learn and adapt to the changing needs and conditions constantly, and a fully integrated security capability will constantly protect it.
Already there are a number of plans, proposals, and projects involving standards bodies, the open source community and commercial vendors along with mobile operators. For example, the ONOS project, an open source community, has launched ‘ONOS Intents.’ OpenStack has projects such as Congress and Heat that use self-styled convergence engines as the central component that controls the infrastructure. There are also two projects in the Open Daylight (ODL) community – Intent-based Network Modelling (IBNEMO) and Network Intent Composition(NIC). Commercial vendors such as Cisco and Juniper are also pursuing intent based networking.
The future is…cognitive
Mobile networks have come a long way since the days of siloed OSS systems, each managing one network system. The transformation and unification of OSS enabled a single platform, but it still relied on a large Network Operating Center. Programmable SDN first sparked the demand for next-generation network management, which advanced as sophisticated feedback loops.
The journey toward the intent-based networking will be challenging, but the rewards will unfold in real-time. Zero-touch networking heralds an era in where both hardware and software are aligned for running complex AI/ML, like deep learning neural networks, which improve the intelligence of feedback loops. The final frontier will be cognitive, closed-feedback loops driven with preventive actions for self-correction with automation and intelligence. The most important decision is to take the first step.
