At MWC Barcelona 2023, Rakuten Symphony launched its Mobile as a Software™ initiative, designed to help the industry rethink its outdated, hardware-focused approach to building networks — one that has remained largely unchanged over 40 years — instead encouraging operators to embrace the flexibility, scalability, cost effectiveness and agility of Mobile as a Software-based networks.
To provide a variety of perspectives on the challenges and benefits of this new direction, we interviewed a number of Rakuten Symphony leaders. Here is an interview with Partha Seetala, president of the cloud business unit for Rakuten Symphony.
I look at the problem from a cloud perspective. The phrase, “software is eating the world” may be old-fashioned in data centers because the concept is embraced completely in the enterprise space. But it is not as well accepted in the telco space, and I think we need to make the case for why telcos are ready. From an economic perspective, costs are eating telcos alive, competition is stiff and consumers want more data and services for the same prices – or even less. In order to improve their bottom lines, telcos have to essentially figure out how they cut costs or develop new revenue streams. Moving to software is the only way to cut costs in a way that does not impact service quality. There is no other way. The foundation for this Mobile as a Software network is a distributed compute environment which I call a distributed stateful edge.
I don’t recommend that an MNO should ever turn their network over to a public cloud provider. And I know that is a controversial opinion, but I come from a non-telco background and so I look at the problem statement from a slightly different perspective.
MNOs have only 3 important assets – their network, subscribers and spectrum. For a business to be competitive and profitable, it must recognize its core assets and optimize to own and manage them efficiently. By giving up their network assets to hyperscalers, MNOs lose the ability to exploit organic knowledge and optimize and leverage that knowledge to offer additional services, thereby increasing their top line and reducing their bottom line. The only thing that remains for an MNO once it loses control of its network is its subscribers and spectrum.
-"I don’t recommend that a MNO should ever turn their network over to a public cloud provider."
First, the enterprise space has a saying - failure in the public cloud is cheap, but success at scale is expensive. Several enterprise customers who have reached scale have realized this and are repatriating workloads back to their own privately managed clouds.
Secondly, guess who has more subscribers than MNOs? The hyperscalers. They have hundreds of millions to billions of subscribers/users. After managing an MNO's network for several years, they would have acquired critical knowledge about operating a mobile network. When that happens, the only thing stopping them from offering their own network service to their substantially larger subscribers/users is spectrum licenses.
By this time, however, the MNO is financially weak because they have been paying the hyperscalers high recurring annual costs and are also incapable of managing their own network due to outsourcing that knowledge years ago. So how can they remain viable? One option is for them to sub-license or sell the other critical asset that they own – their spectrum. By that time, guess who is well positioned to purchase the spectrum? The hyperscalers. who now have advantage over two out of the three critical assets – the network and subscribers. Only antitrust regulations could prevent this from happening. However, as history shows, hyperscalers such as Amazon have expanded to many adjacent areas while carefully navigating regulatory hurdles.
It is extremely shortsighted for MNOs to give hyperscalers access to their networks. No self-respecting MNO will admit they are letting go of control; instead, they will use language more comforting to them, such as: we are benefiting from hyperscaler economies of scale, etc. But in reality what they are doing is letting the wolf into the henhouse.
.jpg)
That’s the key question because the problem that telcos face leads them to a very different cloud solution for Mobile as a Software. The public cloud was built to centralize the compute and storage for better economics, but the MNOs don’t have this same need to centralize their data processing. The problem they face is that they need tens of thousands of small data centers that are spread apart from each other geographically because it’s much cheaper for them to process the data close to the antennas and 5G RAN systems than it is to transport it across a high-bandwidth fiber circuit to a regional or central data center.
For example, you have many wireless signals from the same cell phone bouncing off of different surfaces before hitting the same tower. It’s better to de-duplicate all those signals closer to the antenna before sending an IP stream onto the network. This reduces the transport costs and latency in a significant way.
Now, a distributed stateful edge that is designed this way looks very different from how public cloud services are built. There is plenty of room, power and cooling in a centralized data center, so the compute capabilities can be built out in much more methodical way. Centralized cloud operators can plan ahead of time to build and expand their facilities.
But in the case of the distributed stateful edge, operators can’t do that because there are so many data centers, each with very little real estate and sometimes power limitations that can impact the number of servers that can be installed.
In addition, all of these separate server stacks are all presumably on a different IP network. This brings up the question of how an MNO can bring up the operating system and all these services across such a diverse area. That's a big problem that requires telco-strength orchestration to solve.
Another consideration is how to deploy a service that has elements running in the edge data center, and other elements in a regional or centralized cloud. Some Open RAN deployments are like this with the Distributed Unit (DU) in the far-edge, the Centralized Unit (CU) in a near-edge regional data center, and a few other network functions running in the centralized data center. In these cases, all of these elements need to work together to form one high-quality end customer service.
So again, these are not problems that centralized clouds naturally have, because public cloud services are essentially running everything in one data center.
From a cloud perspective, here’s how we’re bringing Mobile as a Software to market to help MNOs roll out and support a distributed stateful edge.
"MNOs can’t build large distributed stateful edge capabilities without full function orchestration that can scale applications and facilitate their lifecycles."
The foundation of a Mobile as a Software network is a cloud native container platform like Kubernetes that enables micro services for scalability. The Kubernetes-based Rakuten Symphony SymcloudTM Platform adds MNO functionality to the virtualization platform including one-click application onboarding and declarative and context-aware workload placement that can be activated by automated policies. While Kubernetes is designed to abstract away the details of the underlying hardware, several complex network functions used by MNOs, such as RAN or Core, actually depend on the capabilities of the underlying chipsets, smart Network Interface Cards (NIC) or Field Programmable Gate Arrays (FPGAs). How does one reconcile the needs of such applications with the abstractions that Kubernetes forces? Symcloud Platform offers this out of the box through discovery, intelligent resource management and a simple policy language so that network functions that need access to this hardware get it without breaking the underlying design principals of Kubernetes.
Keeping virtualization overhead to a minimum has always been critical in a distributed stateful edge server, and Symcloud Platform has been designed to run in a minimal number of cores. A key benefit of Symcloud Platform is policy driven automation that makes it easier to find resources.
MNOs can’t build large distributed stateful edge capabilities without full function orchestration that can scale applications and facilitate their lifecycles. Rakuten Symphony Symcloud Orchestrator offers multi-domain, multi-site capabilities that overcome the challenge of a distributed app deployment (i.e., at the edge and also in the regional/central data center). While the application is running, Orchestrator has full monitoring and planning tools and can troubleshoot and alarm the system to track problems with the services.
We also recognized the need for high-performing storage to support stateful applications at the edge, in addition to more traditional stateless applications. The Rakuten Symphony Symcloud Storage software creates a storage stack designed to provide application-aware storage and data management supporting the capability to group volumes and applications into a single logical entity. Symcloud Storage supports day zero application turn up with functionality that includes data persistency for stateful apps, always-on availability with self-healing and advanced data placement for stateful apps.
After the software is installed, Symcloud Storage supports ongoing needs such as maintaining steady performance, storage scalability and management so that the storage is easily maintained with minimal expertise or human intervention. When it comes to data management, Symcloud Storage offers a complete set of services including snapshots, clones, back-ups, business continuity and multi-cloud portability.
To answer this question, I would like to draw a parallel with Amazon and how it became widely successful in the public cloud business. Obviously, the company did a great job executing on its business plan, but they had one inherent advantage that others did not: Amazon had its own e-commerce shop, which meant they had the same set of problems that other fast-growth enterprises have.
As Amazon developed its internal cloud, they added software programmability to minimize the number of people managing their system. So, they learned from their own internal operation, and as a result of that they got better at knowing exactly what type of services their customers would want. I think the same thing happens with Rakuten Symphony because we have Rakuten Mobile which gives us first-hand knowledge of what MNOs need in a distributed stateful edge and gives us the opportunity to be close enough to the business that we can innovate very effectively. And we offer this is as pure software so that MNOs can deploy it within their own data centers and retain control of their most important asset – their network.
