T-Mobile US is regarded by many as one of the telecoms success stories in recent years, underpinned by an ‘Un-carrier’ strategy that has allowed it to consistently increase subscriber figures at the expensive of rival operators.
From Hadoop to multi-cloud: T-Mobile US spells out its data platform modernization journey
T-Mobile US is regarded by many as one of the telecoms success stories in recent years, underpinned by an ‘Un-carrier’ strategy that has allowed it to consistently increase subscriber figures at the expensive of rival operators.
In its financial statement for 2022, for example, the Deutsche Telekom-owned operator reported 6.4 million postpaid net customer additions and a 5% increase in revenue. It has also been building its base of 5G-based fixed wireless access (FWA) customers, which stood at 2.6 million by the end of last year.
Vikas Ranjan, Senior Manager, Data and Analytics Engineering at T-Mobile, has been with the operator for about 18 years and says he is “privileged” to have been “part of the beginning of the journey at T-Mobile.”
That’s not to say the journey has been easy, of course. Indeed, a key challenge has been learning how to manage, and derive insights from, the vast amounts of data that is generated by communication service providers every day.
During a TM Forum webinar to discuss the modernization challenge of data platforms — including common approaches, how to track and manage costs in a multi-cloud environment, and structure and governance issues — Ranjan describes T-Mobile’s own data platform experiences “from our inception to where we are today” and highlights some of the challenges and the best practices it learned along the way.
Connecting the dots
As Ranjan explains, telcos are no longer typical network companies. “When we look at business domains within a CSP or telco internally, we see network as the core of the business because everything is built around the network,” he says. In essence, the telco has become a far more complex ecosystem “where we want to correlate and harmonize data” across interconnected enterprise domains including billing, supply chain, finance, customer, support, marketing and more.
Although this brings new opportunities, it also brings challenges, Ranjan says. “In order for us to connect the dots … or to derive the value and the benefits from all of these business domains and opportunities, there has been a lot more focus in the last 10 to 15 years on data platforms and the power of data.” He adds: “All of this data is only good if we see the value of data. And in order for us to derive the value of data, data reliability [and] observability … are becoming extremely critical and important to be successful.”
T-Mobile’s data platform journey began in 2015 when it migrated workloads from data warehouses to Apache Hadoop, an open source framework that is used to efficiently store and process large datasets ranging in size from gigabytes to petabytes of data.
However, by 2019 it reached a tipping on what Hadoop can do in terms of scalability. T-Mobile also found that Hadoop, “just like any new platform” was not easy to learn.
“You need really, really smart people to manage the complex ecosystem of different tools on Hadoop,” reflects Ranjan. On the plus side, that helped ease the next phases of the journey: the move to the cloud, hybrid cloud, Kubernetes and multi-cloud.
Ranjan says the benefits of the cloud in terms of elasticity and agility were immediate, but also brought some very complex challenges in terms of cost and data governance, especially in a multi-cloud environment.
In terms of other key learnings over the years, Ranjan emphasizes that collaboration is key to the success of data platform modernization. “You can only be successful if you bring people together; you can only be successful if you bring ideas together,” he says.
“I think the second thing we learned is what I call a fail fast, pivot fast approach and mindset. Do not rush into doing things as you hear about things. Spend enough time on understanding and identifying the problem,” he says.
Ranjan also points out that you won’t be successful if you don’t nurture your team. “Invest in your people, train people, bring them opportunities. Invest time in hackathons, give them opportunities to think beyond normal, give them opportunities to look at how can we do things differently,” he says.
“Last but not least, focus on your data observability and data quality from day one,” he says. If you don’t, he warns, you will encounter problems when trying to solve issues and business problems further down the line.
He concludes: “I will say that this is a really exciting world we are in right now. When we were looking at data platforms a few years back, the whole integration of AI, ML into data platforms was very, very naive. It’s a lot more mature now”, as the hyperscalers invest billions into making AI and ML a core integral component of data platforms.
At the same time, he notes, not every problem has to be solved with AI and ML. “It all comes down to responsibility, he says, which he sees as critical to success “with all these cool technologies.”
Watch the webinar on-demand to find out more.