In depth: Vodafone revamps monitoring capabilities with Celfocus as it eyes billion IoT devices mark

Vodafone is preparing to extend new, “next-generation”  monitoring capabilities across its enterprise services portfolio, in a bid to improve customer satisfaction and help its own ongoing cost-cutting drive.

The plan follows on with the deployment of an artificial intelligence and machine learning-equipped monitoring solution for its Internet of Things user base, supported by systems integrator Celfocus and delivered over Amazon Web Services cloud infrastructure. The aim is to automate and accelerate problem resolution for clients of connected device division Vodafone IoT, helping improve efficiency and quality of service, and progress down the path towards “self-healing” networks.

Sampada Basarkar, Director of Products & Platform Engineering at Vodafone, presented the project at TM Forum’s recent DTW23—Ignite  industry event  in Copenhagen, calling it a “beginning”. She also guided that the operator’s engineering department is “in discussions right now” about other areas to apply new, AI-/ML-based monitoring techniques.

“ Within Vodafone Business, we have lots of different portfolios. We have to be constantly looking at what needs to be done and how we reduce our operational costs. How do we ensure that we go towards more proactive monitoring and zero-touch support? This is the beginning use-case that we are on. There will be more use-cases; there are more products. So we’ll be on a continuous journey. ”

Basarkar.

The drivers: client demands melded with need to scale and modernise

Basarkar described the business case for the IoT monitoring revamp as made up of three interlocking drivers.

For one, Vodafone’s IoT client demands are particularly focused around the need for more speed when it comes to problem identification and resolution.

At the same time, Vodafone’s network operations team was faced with the challenge of processing a huge volume of network alerts (“2,000 per day”, typically), and sifting false positive results (the vast majority) from those genuinely requiring action — which was putting a brake on velocity. The operator “wanted to change from a reactive to proactive approach to make sure we increase customer satisfaction by detecting anomalies much quicker”, said Basarkar.

Secondly, as an internal driver for change, Vodafone wanted to bring more automation into how it went about network connectivity monitoring — particularly, again, around anomaly detection.

By introducing ML algorithms into its system, the operator aimed to significantly cut down the time engineers took to validate each alarm (around 20–30 minutes) and then assess their impact (a further 20–60 minutes). It was also, more broadly, making the first steps on a longer term “journey to self-healing networks”, with the promise of significantly improved efficiency and effectiveness, said Basarkar.

“ We want to move towards zero-touch support. Because if you have to [meet] these SLAs, you can’t do it just with human resources; you have to have some sort of machine learning algorithms to be able to quickly detect and manage these situations at speed. ”

Basarkar.

The third element of the upgrade was to ensure Vodafone was positioned to support the growth it anticipates will emerge from IoT adoption. Basarkar indicated the new solution had to be capable of handling the huge volume of traffic data coming in from Vodafone IoT’s user base, which is currently processing two billion call detail records (CDR) per day.

It was also designed to scale next-gen monitoring to as many as one billion devices (significantly more than the 162.3 million IoT SIM connections Vodafone was serving when it last reported numbers, at 31 March 2023).

The challenge: Group needed to trade up on tech

Backdropping all three of the drivers for the project was that Vodafone’s legacy monitoring tech was a major constraint on development of new capabilities.

Again, the main issue was velocity, as the old solution was ill-equipped to allow “automatic anomaly detection, and move towards self-healing”, Basarkar said.

“ We were getting 2,000 alerts per day, and, if it took 20 minutes to analyse every single alert, we were spending a lot of time trying to just weed out false positives in that data. That led us to think about what needs to be done in this area from a next-generation monitoring perspective. And that’s why we partnered with Celfocus. ”

Basarkar.

Visibility in the cloud

With the upgrade, many of the functions within Vodafone’s IoT Core Services Connectivity  monitoring capabilities and operations can be switched from manual to automated processes, enabling the operator not only to resolve issues faster but also gain “increased… decision intelligence”, said Basarkar.

• In the realm of real-time processing  — the first stage in the overall monitoring process — the project team overhauled Vodafone’s pre-processing mechanisms, as well as working with Celfocus to build in new solutions based on open-source technology and AWS infrastructure. This, Basarkar said, moved the operator away from suffering “data latency” of more than 20 minutes and enabled it to gain “real-time” capability instead. “With the amount of data that we were getting, this was a massive problem that we had to go and solve”, she said. “If you look at the peak rate, we monitor about 30,000 CDRs per second at the peak rate. And it now happens with sub-second latency”.
• The second step, of anomaly detection, was where AI/ML capabilities were brought in, automating the process of identifying traffic anomalies and automatically flagging network alerts.
• The final stage, of data analytics, also saw the implementation of new software and processes to automate functions such as impact analysis to suggest service priorities.

At the same time, Vodafone increased the range of historical data that was available to analyse, to open up more insight on fault patterns over time.

The solution is designed to enable Vodafone’s ops team to “review and filter the data in a way that the data is meaningful for them and they are able to analyse these records and alarms very quickly”, according to Basarkar.

This analytics component has been key for improving visibility into patterns of anomalies and improving speed of resolution, she indicated.

“ Rather than looking at just a single instance, we’re looking at multiple instances to see if the same pattern is exhibited over and over again… It’s really just making sure that you analyse the data over a [greater] number of records. ”

Basarkar.

Basarkar: journeying into the self-healing

While effusive about the outcome of the project, Basarkar stressed that Vodafone has grander ambitions when it comes to extracting value from its network monitoring functions through AI/ML.

She described the operator as having become “proactive” when it comes to IoT network monitoring, having previously been “reactive”.

Vodafone has not yet evolved its capabilities to a state of full “self-healing”, where it reacts “autonomously” to failures. Basarkar’s slides, for instance, indicated that some areas of the fault response function (or “healing”) remain manual, such as escalation. The vision is for this to be made significantly more automated once Vodafone has “analysed enough data and built the right processes”.