.svg)
In this issue, Rakuten Symphony SVP of Technology Devesh Gautam explores what it actually takes to manage networks as software systems, drawing on his extensive experience building greenfield networks at Jio and Rakuten Mobile.
Telcos targeting growth won’t cut a clean path adding more tech layers and rapidly building out infrastructure. Networks are just too big, too complex and too expensive to scale the traditional way.
There are, however, architectural and operational commitments that can be made to redefine how the business operates and fuel a new era of growth.
That is my top takeaway from building software-powered networks at scale within Jio and Rakuten Mobile. The same principles, tested under real operational pressure and proven at national scale, are ready for adoption by telcos of every size in every region that are positioning for long-term modernization.
The race is on to scale capacity without linear cost growth. Whether pursuing more differentiated services, stronger enterprise plays, slicing, private networks, new app-driven use cases or all of the above, the focus is on reducing operational load while keeping costs contained and customer experiences high.
That means important network advancements like 5G-Advanced and AI-driven optimization must be implemented in the absence of yet another network reinvention (i.e. long planning cycles, one-off integrations that become dead ends, etc.).
As one of the earliest Rakuten Mobile team leads, responsible for rolling out a next-gen national network, we established early on that we would not be chasing specific technologies or designs within a certain launch window.
We understood it was critical to build a network that could rapidly evolve without constant reinvention.
So we set about making a series of small but hard decisions, careful to remain disciplined as we progressed. It was understood these decisions would have to hold up under real operational pressure, be it scale, change or unexpected events. They also had to make sense across vendors, domains and multiple technology generations. Perhaps most importantly, they had to support the business we wanted to operate.
It’s important to remember Rakuten is first and foremost an internet company doing telco. The guiding network and technology principles that now support more than 70 business and services were just as applicable in this legacy domain, even if we were pursuing a series of firsts.
From this work emerged a clear set of architectural and operational—and most importantly, non-negotiable—pillars that would become the foundation for everything that has happened since.
Let’s break down these pillars in the context of modern telco ambitions to understand what a clear path to growth looks like in what will prove to be one of telecom’s most decisive moments.
We knew we had to move network functions off dedicated hardware but that was not the end goal in itself. Really, we sought to remove the structural limits that were known to prevent networks from evolving once pushed into production.
Importantly, virtualization would need to be fully realized via a microservices-based, cloud-native architecture. Network functions were broken down into small, independent components that could be deployed, scaled and updated independent of anything else happening in the network. This design choice alone meant the network would actually behave like software, not fixed infrastructure. It unleashed interfaces that were more clearly defined and enforced, brought clarity to dependencies, and made automation and programmability a reality.
Above all, it shifted control.
When we break down functions this way as operators, we are no longer locked into rigid release cycles or contained by singular systems. It is the key to responding faster to new requirements, introducing incremental change on our own terms and avoiding the cascading impact that can accompany tightly coupled systems.
Our industry sometimes discusses automation in the context of an important technology achievement that is a necessary step toward enduring efficiency. We viewed it as the core operating model that would enable the network to be fully operable at scale. Essentially, a day-zero decision versus post-deployment bolt-on.
Designing automation into the system from first principles was especially consequential in the most complex parts of the network. The decision forced a different way of thinking about workflows, interfaces and failure handling, while giving confidence early on that automation could be trusted in real operational conditions.
We never viewed automation in isolation. Yes, it could accelerate behaviors, but that would also extend to potential problems in cases where underlying systems are rigid or opaque. Ultimately, this decision was inseparable from the earlier commitment to microservices and software-native design.
As the default mode of operations, the network could scale without a corresponding increase in operational burden as we benefited from repeatable change, predictable expansion and contained operational complexity, regardless of network size.
Every component brought into the network across RAN, core, transport and platform layers, had to expose data about its real-time behavior. Not for the sake of collecting for some future use case but to systematically transform the network from static infrastructure into a living system that could be observed, learned from and acted upon.
Without this visibility, automation has zero context and optimization is reduced to guesswork. Telemetry was present in every interface possible, including network performance, user context, QoE, platform health and environmental factors like power and load conditions.
This breadth was important given partial visibility would have produced blind spots that limit actions the network could safely take on its own. The data layer became the foundation for real-time decision-making and eventually, the RIC. A rich and consistent telemetry stream powered new use cases that could be quickly developed, tested and deployed with utmost confidence. Innovation was led by insight, not how complex an integration might be.
Yes, it would have been easier in the short term to accept opaque systems or limited data access so we could deploy faster. The payoff from the early growing pain is now clearly evident as the same telemetry foundation supports AI-driven optimization, energy efficiency gains and continuous learning across the network.
Telcos will win in the next era of growth by taking back control. At Rakuten Mobile, we weren’t interested in optimizing outsourced operations or vendor-led models because we wanted to operate the network ourselves, end-to-end. That also meant the way customers experienced our network was totally on us. (Relevant aside: Opensignal has named Rakuten Mobile Japan’s leading mobile network service provider across five network performance areas.)
Vendors would still be an important part of the ecosystem. This was primarily about eschewing the traditional relationship dynamics. In our view, vendors should supply components while the operator defines how the system behaves. That’s only possible when you own operations and understand the network deeply enough to operate it alone.
Every decision we took culminated at this stage. Microservices make change controllable. Automation makes scale manageable. Telemetry provided the context needed to act with confidence.
With those foundations in place, we were able to run a national network with a relatively small operations team and maintain visibility, responsiveness and control.
We detected issues early. Decisions were made based on real data rather than assumptions. Operational effort stopped growing with network size.
Most importantly, self-managed operations closed the feedback loop between the network and the business. In reality that meant we instantly felt when something broke or improved, creating an environment where the network could evolve continuously instead of through periodic, disruptive overhauls.
Modernizing a network based on these four non-negotiable principles entirely shaped the opportunities we could pursue. Gone were the fragility and unknowns of previous network tech cycles.
Today, we have the confidence to proceed with 5G-Advanced, AI-driven optimization and increasing levels of autonomy.
Nothing we deployed was revolutionary. We simply architected a network unconstrained by design or operations.
Importantly, scale never forces complexity.
The same foundation that powered a mature, commercially deployed RIC, energy efficiency gains and self-managed operations allows the network to continue learning and improving over time as innovation becomes continuous.
For telcos strategizing in this moment, committing to a small set of principles will let networks evolve as software systems. Once they are moving faster with confidence, they are already on a proven path to new growth.
Mention Devesh Gautam in the comments to start a conversation. Check out his recent interview on this topic covered in Zero-Touch.
Senior Vice President of Technology
