AI and ML: Key to predicting network performance issues and enabling automation
5G is increasingly being eyed as a productivity booster across the globe. According to telecom analyst firm IHS, 5G’s full economic benefit could produce up to $13.2 trillion worth of goods and services enabled by 5G mobile technology. The opportunity is immense – and so are customer expectations. The stakes are now far greater for communications service providers (CSPs) as mission-critical services such as healthcare, unmanned aerial vehicles (or drones) and smart grids take 5G use cases to the next level. This is why CSPs must look to artificial intelligence (AI) and machine learning (ML) capabilities to transform their network performance management from a reactive stance to a proactive, preventative position. Let’s outline the reasons why.
5G creates a more complex network – and a torrent of alarms
In a 4G world, it may be typical for a Tier 1 mobile operator to have hundreds of thousands of active alarms being managed by their network monitoring systems. In 5G, the event and alarm volume is expected to grow significantly due to increased network complexity, the combination of physical and virtual resources, and a more dynamic, frequently changing network. It will be outside the realm of possibility for network operations teams to manually go in and triage each alarm in the case of a major network failure. In the consumer context, a network outage or service quality degradation will create a mountain of bad press and complaints. And in the enterprise sector, an hour of downtime can cost a business anywhere between $1M and $5M.
AI/ML can improve how network failures are managed by:
- classifying alarms according to historical behavior and automatically identifying their root cause,
- extending traditional rules-based analysis with adaptive mechanisms to correlate trends and locate the source of the problem more quickly,
- automatically detecting and identify patterns across all data sources,
- automatically deriving the relationship between network resources and events for faster root cause analysis.
5G requires switching from defense to offense
What we know for sure is that 5G will unleash new services that touch almost every industry – unveiling capabilities we can’t even imagine. And many of these 5G use cases will comprise stringent, contract-backed service level agreements (SLAs) that CSPs must comply with – especially for enterprise customers who want to protect their brand by ensuring their connected products and applications will be experienced with the level and quality of service they were designed to deliver. Companies responsible for selling things like connected cars, eGame competitions, drone services, augmented reality and even remote surgery will want to be assured their customers are getting what they’ve paid for; and these are just the tip of the 5G iceberg.
5G enterprise SLAs will come with strict financial penalties that will cost the service provider dearly if they fail to deliver as promised. Therefore, the ability to predict end-to-end service availability and prevent network performance issues before they occur will be imperative in order to avoid high penalties- and the loss of valuable, big brand enterprise customers.
With the increasing number of available data sources, such as network alarms, device measurements, and customer usage records, AI and machine learning capabilities can be leveraged to help quickly repair potential network failures and other issues. This new automated assurance capability allows service providers to detect network patterns as they arise, prioritize and filter alarms, and then identify the root cause of network outages, performance degradation and congestion issues – instantaneously. It can even identify evolving trends and network anomalies and predict impacted services and customers. It’s almost like having a crystal ball in the network operations center.
Being able to detect non-trivial issues at an early stage, or even predict them before they occur, gives operators the time to apply preventive measures so that preemptive actions can be taken to avoid service performance issues.
Managing the slices
One of the key features of 5G is the ability for service providers to create individual network slices, which become individual networks in their own right – each with their own service characteristics and requirements.
By integrating performance management functions with NFV service orchestrators and 5G network slice managers, network engineers can perform sophisticated impact analysis on current and forecasted anomalies, as well as correlate network function issues with their related services and slices. But monitoring and managing slices requires sophisticated tools that can manage all the network layers, including physical and virtual, along with the various network functions and applications. It’s a complex, complicated process that isn’t for the faint of heart.
Introducing Helix 11
Advanced anomaly detection, alarm prioritization, automation and root cause analysis and predictive analytics are just a few developments that feature in TEOCO’s flagship service assurance platform Helix. The new Helix 11 release is a major product expansion designed to address 5G by allowing service providers to collect, process and analyze the rapidly growing volumes of data necessary for managing the end-to-end performance of 5G networks and services.
The updated platform gives operators the necessary tools to automatically predict, identify and resolve service problems quickly and efficiently, fixing faults where possible, predicting outages and performance issues, and producing actionable insights that can be shared across other systems and teams.