- Data centers, connectivity and a playground for innovation are the key pillars of Mastercard’s cloud strategy
- The company's CTO of Operations George Maddaloni explained how its bits and bytes get where they need to go
- And yes, they are still using MPLS
Mastercard may not be the first name that comes to mind when you think of connectivity, but it turns out that it takes one heck of a network to connect three billion cardholders to hundreds of millions of points of service and thousands of banks in the blink of an eye. In an exclusive interview with Fierce, CTO of Operations George Maddaloni explained how the company gets all those bits and bytes where they need to go. And man, is it something to behold.
The short version? Mastercard’s network is a sprawling mass made up of 60 data centers, multiple 100-gig backbone transport links running MPLS technology and cloud sandboxes. Those three things – data centers, connectivity and a playground for innovation – are the key pillars of Mastercard’s cloud strategy, he said.
The easiest way to think about Mastercard’s data centers is as one giant private cloud, Maddaloni said. The company still runs some of its own data centers but noted the majority are housed in colocation facilities. While some of these deployments are larger to handle data processing, some are smaller cage-sized server deployments designed to act as transit gateways that can funnel traffic onto other portions of its network.
“We need to operate in every region that we have consumers processing transactions, so we establish regional data centers where needed,” he explained.
Colocation facilities give Mastercard access to another key piece of its network puzzle: interconnection to both the internet and hyperscale public clouds. While Mastercard primarily uses Amazon Web Services and Azure for internal application deployments, its outward-facing software needs to reach customers operating in all the major hyperscaler environments.
“AWS, Azure and Google, we’ve seen all three have customers operating in the cloud that we have to connect to. That has scaled up dramatically in the past couple of years,” Maddaloni said.
“We’ve just hit where we’ve had over 250 million transactions that have been processed through those cloud connections," he continued. "So, public cloud represents a part of our network that has scaled up dramatically over the past three years since we’ve really taken that on.”
Mad about MPLS
As for how it connects all of these data centers, Maddaloni said Mastercard runs a backbone network of multiple 100-gig transport links using multi-protocol label switching (MPLS) technology. Wait, MPLS? Yep.
Though other enterprises are moving away from the technology, Maddaloni said MPLS remains the best option for Mastercard, in part because it’s the most efficient for the type of traffic the company is running – think messaging rather than beefy video streams – and allows it to weave together different carrier networks so they look like one giant network.
“We’re using that protocol to establish the same kind of virtual fabrics within that network so that we can carve out different components of bandwidth for different functions,” Maddaloni said.
IDC Research VP Dave McCarthy told Fierce it’s not surprising that Mastercard is still using MPLS. Why?
“The things that financial services does has such a high emphasis on ensuring uptime and security — and so those kind of traditional ways of connecting things together makes a lot of sense for them,” he said. McCarthy added that the financial services sector has been on a long journey of moving to cloud, but hasn't quite gone all in yet.
That said, things are changing. McCarthy noted that IDC has forecast that cloud spending in the financial services sector is set to increase at a five-year CAGR of 22.3% to reach $335 billion in 2028.
AI and the future
But what about innovation — and the use of artificial intelligence (AI)? Well, Maddaloni said Mastercard uses public cloud sandboxes internally to allow developers to tinker with greenfield projects and provide its fintech startup partners with robust tools for innovation.
And on the AI front, Maddaloni said Mastercard has been using the tech for over a decade to run its fraud detection engine. Now, he added, it is looking at ways to leverage Generative AI to improve the performance of its fraud detection models and reduce false positives. It’s also tapping into AI internally via tools like Microsoft Copilot, Amazon Q and Github Copilot.
But it’s forging ahead cautiously.
“The thing that we always think about is in terms of data,” he said. “We’ve got a very solid set of data principles in terms of how we protect consumers’ data and other principles associated with that including the elimination of bias, social impact and maintaining the privacy associated with that.”
McCarthy noted that Mastercard’s approach is reflective of the reality that while GenAI and large language models (LLMs) sparked massive market excitement, enterprises are finding implementation tricky. More specifically, they’re grappling with how best to use GenAI and LLMs while still protecting their data.
In terms of what’s next on the horizon, Maddaloni said Mastercard is looking to get edgy. And it’s not the only one. IDC predicted edge spending for Financial Services will reach $29 billion in 2028 with a 5-year CAGR of 13.9%,” McCarthy told Fierce.
“While everybody is focused on the chips and the cloud and things of that nature, I think what we can do on the edge of our network and by leveraging edge computing” will really be transformative, Maddaloni concluded. “We look to remove latency, we look to do more intelligence on the edge, we look for new connectivity patterns that make our technology work better and I’m actually really excited about what may come from an edge computing perspective.”