At W.Media’s Interconnect World in Melbourne in April, two businesses with very different approaches to buying connectivity services – one opex driven and one capex driven – outlined the sequence of decisions that need to be followed when looking to venture into, change or upgrade their network services.
Net Number’s opex-only approach to connectivity
US-based Net Number Global Data Services provides phone number intelligence data and services to help global businesses and communications service providers manage their networks, improve security, and combat fraud. Its CEO and President Steve Legge said that Net Number traditionally did this using a very large global network that was based on its own technology and buying capacity, but in 2017 it began migrating to AWS.
“We’ re exclusively AWS globally,” he said. “We have a large AWS network and footprint and a global customer base. With AWS, it’s a very different model, no capex, right? All of our money in terms of network costs go into operating the network and actually sourcing data. So they’re our largest cost.”
Net Number also runs its own technology in AWS. “We’ve got a couple of patents on the network architecture and the technology that we use to distribute the data that we provide,” he said. “The biggest learning or the biggest concern that has never gone away – and it’s always something I test the team with – is making sure we understand our costs. When you’re relying on something like AWS or Azure or GCP, the cost can get away from you very quickly.”
Are networks resilient enough?
Legge pointed to the increasing complexity of interconnects and network infrastructure which means that network administrators can even get caught short despite having dual redundant paths. “We had some infrastructure in one large capital city…the AWS infrastructure was fine, but the physical point of presence where there was some physical network came in via a large, well known carrier in the US who you would all know. This infrastructure had been there for years,” he said. “It had run flawlessly via diverse paths, two physically diverse paths out of the data centre, different conduits, the whole lot. One day, both links went down for everyone.”
It turned out the data centre converged in a different capital city, at the same point of presence, on a single physical card. It worked perfectly for many years. “So the networks have got so complicated, the planning can be poor,” he said. “Whatever has led to this, led to it to physically diverse paths, and they’ve gone. So we encouraged our customers at that point to obtain triple, even quad diversity.”
He added that at that point, SLAs don’t matter with this type of outage. Over time that infrastructure had evolved despite what it originally looked like. There was a complete disconnect between what was happening with their customers and then what was happening in network planning. Despite a contract being in place the service was no longer in the service provider’s catalogue.
Legge said the carrier market is moving into the cloud. “More and more we see carrier customers now wanting to meet in the cloud, which is great. So, if you’re in AWS, we can VPC peer…The configuration is almost automated,” he said. “What we do notice is that it has become much more complex, where the virtual world meets the physical world. So what you get is, you’ll talk to a provider about being able to get connectivity. They’ll tell you how long it will take to deliver the physical connectivity. But unfortunately, it’s usually more than just the physical connectivity into the data centre or into the rack or the cross connect.”
He said that service delivery in North America was becoming increasingly complex due to the involvement of multiple vendors behind major carriers. While fibre connections may be delivered on time by a lead telco provider, such as AT&T, the virtual network provisioning often depends on several third-party vendors whose timelines are not always aligned. This has led to delays and mounting frustration, with services sometimes arriving months later than promised. As a result, customers are now being forced to engage more deeply with providers to understand the full supply chain and delivery process. In contrast, cloud-to-cloud connectivity remains relatively straightforward.
NextDC’s model centres on capex
One of Australia’s largest data centre operators and builders, NextDC, buys connectivity and sells it to customers so the company looks for partners who sees what they are packaging for customers and are prepared to be part of the backbone solution, according to Head of Network Operations Sean Rinas.
In contrast to Net Number, connectivity for NextDC – a capital-intensive bricks and mortar business – is treated as a capital expense as well, especially when it is building a new metro data centre. “We use our partners to build those private networks between sites, and we outlay that all as capital expense,” he said. “Even if we might put up a 24 year IRU on this service ahead of time, we pay that all up front. So we’re the polar opposite, really trying to minimize opex.”
NextDC does run some software in the cloud, incurring opex, but Rinas says connectivity for data centres follows the capex build model. “We know that the building is going to be there for more than 25 years. We want that network to be there for more than 25 years,” he said.
Rinas contrasted new dark fibre providers in Australia with the more traditional, larger suppliers, adding that this can impact resilience. With the new players, NextDC has transparency into their providers’ networks. “The transparency that they give us with their KMLs and physical infrastructure – you can look at how they’ve laid out the network,” he said.
Traditional carriers have a lot of redundancy built into their network, he said, “but [not] the transparency at a physical level, when we want to look at purchasing and installing a fibre solution in the metro area, for example, or even an inner capital solution.”
“You used to be able to go ‘I’ll buy from carrier one and I’ll buy from carrier two, and it’s redundant,” he said. “We know that’s not necessarily the case anymore. So we need, when we’re purchasing connectivity, to underpin our network. We need to have providers and partners who are really transparent at the physical layer of what they’re delivering for us.
Rinas said NextDC benefits from being able to focus on delivering high-quality services rather than rushing to deploy solutions. While capable of rapid service delivery, NextDC is afforded the time and capital to carefully plan and test its connectivity solutions, including lab evaluations of vendors before deployment. This approach, supported by NextDC’s management, reflects its commitment to premium service standards, including pushing for 100% uptime not just for power but also for connectivity.
Steve Legge and Sean Rinas spoke on a panel discussion at W.Media’s Melbourne Interconnect World 2025 in April. To see future Interconnect World events please visit: https://interconnectworld.com/
For future Australia & New Zealand Cloud & Datacenter events, please visit: https://clouddatacenter.events/event-region/australasia-events/
