Watch BT’s Jose Domingos and Federos’ Andrea Reichstein discuss the Catalyst project

Telcos look before they leap with automation

(This story was initially published on TM Forum Inform June 2019, contributed by Sarah Wray, Freelance Writer & Editor)

A TM Forum Catalyst proof of concept is using real data to demonstrate how communications service providers (CSPs) can improve the quality of their data to make way for more accurate automation. Early results are encouraging with BT improving data accuracy by up to 3%.

Automation offers huge opportunities for CSPs to improve services and reduce costs. It goes beyond these benefits, though – automation is essential to delivering digital services, so CSPs have no choice but to implement it.

The emergence of and advances in technologies such as network functions virtualization (NFV), software-defined networking (SDN) and artificial intelligence (AI) are helping CSPs in their digital transformation but these technologies also drive the need for automation, AI and machine learning. There is no other way to cope with the complexity and speed of virtualized services.

However, CSPs also must proceed with caution when it comes to automation to ensure they avoid the old ‘garbage in/garbage out’ adage – or worse: having it on repeat at high speed.

A 3D problem

“I think of automation as a three-dimensional problem,” says Jose Domingos, Chief Assurance Architect, BT.

  • Data quality: “Without accurate data, bad things can happen when you automate. Trust me on this one.”
  • Volume of data: “The more data you have, the better the automation outcomes but also the more computing power required and the better the tools and technologies you need to process that data.”
  • Automating the right things and optimizing: “It’s very easy to automate something and just do the wrong thing over and over again. You want to spend some time thinking about how that automation is going to work.”

A TM Forum proof-of-concept Catalyst project called AI LEAP, has carried out important collaborative work to tackle these challenges, using innovations in AI, machine learning, event analytics, automation and prediction.

The champions for this project are BT, AT&T and Orange. They work alongside participants – in this case Arago, Galileo Software, Federos and Wavelength Communications – each providing a technical piece of the solution.

Now is the time

The AI LEAP Catalyst looked at how AI can be applied to resolve some of these data issues to improve automation. And now’s the time to do this foundational work – as 5G rolls out, telco systems could be dealing with between 10 to 100 times more data. Network slices and the transition from non-standalone 5G (with a 4G core) to standalone will only increase the difficulty.

“Without automation, telcos won’t be able to deal with that complexity and deliver the quality of service that is expected,” says Steve Bowker, CEO at Wavelength Communications, an advisory firm which specializes in next-generation network strategy, design, open APIs and service assurance.

Dr. Andrea Reichstein, Head of Research, Galileo Software and a finalist for the Future Digital Leader award at the TM Forum Excellence Awards, comments that in the 5G era, AI-driven automation could help to “predict overcapacity, adjust policies and [also start] exception handling.”

Exploring patterns

Specifically, the Catalyst focused on using machine Learning to spot irregularities and patterns in Event analytics data to support root-cause analysis. From those patterns, the team explored the use of Automation to resolve issues, as well as Predicting and preventing abnormal behavior and improving customer experience. This is where the AI LEAP name comes from.

At TM Forum’s Digital Transformation World in May, the AI LEAP team presented two use cases to show their proposed approach in action:

  • Inventory detection and reconciliation – using machine learning and AI algorithms to find related events, automatically infer network topology and verify, cleanse and update inventory systems
  • Detection of anomalies – using machine-learning algorithms to monitor performance and metrics for abnormal behavior and automatically gather relevant data to identify and fix the root cause of the issue

Real-world problems

“It’s fundamental to have accurate inventory data,” says Domingos. “If you are trying to provide an assurance function on your network and you have data that is not accurate, you end up wasting resources – and it’s not just twice as many resources but a much higher multiplier, because you start a repair function on one device and by the time you go down the line, you’ve multiplied the number of resources involved.”

The team’s solution shows how AI can quickly identify event linkages. Time stamps and location data provide clues about events that are likely to have the same root cause. If the related events are detected three times or more in a row, an automation is triggered.

Tim Heywood, Head of Product Management and Marketing, Federos, explains: “We continually improve and feedback in a closed-loop way to improve the data within the topology and the inventory system. It’s using data to improve data. The data is there; you just have to use it intelligently.”

Data correlation and event analytics is done by Assure1, a product from Federos which brings together fault, performance, topology and service-level management on a single platform. HIRO from Arago uses AI to manage and automate business processes.

Galileo Software, the research division of Federos, provides its Galileo Vision platform, a map-based view of events and its Dynamic Service Relationship Modelling and Management (DSRM2) which shows ‘at risk’ services and real customer impact in near real-time. DSRM2 was developed through Galileo’s participation in various TM Forum Catalyst projects, notes Dr. Reichstein.

BT improves inventory accuracy by 3%

An important feature of this Catalyst was that it used real data. Three months’ worth of BT’s event data equaled 48,000 topology connections and around 150,000 events. The team was able to improve BT’s inventory data accuracy by up to 3%.

“Our inventory is already quite accurate but there are some manual processes involved, meaning sometimes things don’t quite align,” says Domingos. “I was very impressed that we achieved a two to three percent accuracy improvement.”

He adds: “I think if we pushed through into six or nine months of data, there would be an even higher multiplier effect.”

Thorsten Buescher, Director of Global Customer Services, Arago, notes that the potential rewards for CSPs is big.

“This is with just three months of data,” he says. “With some of our service provider clients, we have seen that if we follow the model through, they’ve gained $2 million to $3 million a year in revenue by simply having a more accurate inventory.”

Showing AI who’s boss

CSPs plan to apply AI across their businesses – not just for dealing with data quality issues, network automation or customer experience, but ultimately for almost any process or problem.

“Orange is very interested in the use of AI for cognitive network automation for fault and performance management, specifically in the area of knowledge management and decision support,” says Sophie Nachman, Standards Project Director, Orange. “We are excited to see how AI LEAP use cases focused on inventory detection and reconciliation to contribute to the Open Networking Automation Platform (ONAP), Active and Available Inventory (A&AI) project, and into TM Forum standards.”

Rob Claxton, Chief Researcher at BT and Co-leader of TM Forum’s collaborative work on AI, comments: “As we deploy AI widely and at scale across our organizations, we need to be able to maintain visibility and control, and, ultimately, accountability for all of the tasks that are being automated.”

TM Forum’s AI Program is developing standards that allow CSP to do that. These include service management standards for AI and a Data Model.

The AI LEAP Catalyst road-tested and contributed back to these standards and models to ensure they’re fit for purpose.

Domingos says: “An important aspect of this work was also to drive new standards and solutions into the industry – to link in with Federos, Arago and Galileo Software to help push new capabilities and products. Then others will join in too. It’s not only a real-world problem but also real-world solutions and standards.”

To learn more about the project, watch this panel discussion filmed at Digital Transformation World 2019 or reach out to the project partners at Federos.