Mobile Tech Explained: Cloud infrastructure

This article is part of a series breaking down the complex systems and technologies that make up the Rakuten Mobile network and Rakuten Communications Platform. In this edition, we speak with Ryota Mibu, division manager for infrastructure platforms at Rakuten Mobile, Inc.

The cloud is one of the defining technologies of the twenty-first century – it has put vast computing resources at the fingertips of billions of people, enabling them to do their jobs and live their daily lives. Many of the world’s most advanced companies, including Rakuten, run their businesses almost entirely in the cloud, which is both more flexible and more cost-effective than traditional IT.

Although it may sound fluffy, even ephemeral, the cloud depends on extensive physical infrastructure — data centers and fiber-optic cables — the factories and railroads of the digital economy. In some cases, these resources are shared between entities (a concept known as public cloud or community cloud), while in other cases they are dedicated to a single entity (a concept known as a private cloud).

One of the beauties of standardizing the platform from the central data centers to the far edge data centers, is the simplicity — we don’t need to have extra engineers or extra testing.

Ryota Mibu, division manager for infrastructure platforms at Rakuten Mobile, Inc.

More elastic and more efficient

One of the distinctive characteristics of the cloud is that its computing resources can be quickly repurposed, so the generic servers in data centers can be used to run a very wide range of software and services — a concept known as infrastructure-as-a-service. This versatility makes the cloud much more elastic and scalable than traditional IT, in which individual pieces of hardware are dedicated to a single role throughout their lifespans. For a large organization, centralized data centers are cheaper and easier to run than IT systems distributed across their facilities.

Crucially, cloud computing is also very resilient. It enables a service to quickly recover from any unexpected outages. As the cloud employs multi-purpose generic hardware, one server can quickly take over from another, should the latter fail. Moreover, each cloud is typically divided into what are known as “availability zones”, designed so that a failure in one zone won’t affect another. That means computing workloads can be immediately switched to another availability zone, as needed.

The world’s first cloud-native mobile network

Rakuten Mobile’s network in Japan is cloud-native — all the systems it uses were designed from scratch to run on cloud infrastructure. That makes the network very cost-effective and straightforward to manage, maintain and upgrade — computing resources can be dynamically matched to demand and new services can be rolled out simply by upgrading the software.

This flexible cloud architecture also enables new software to be tested in one segment of the network before being rolled out more broadly. Indeed, the dynamic nature of its cloud-native environment means Rakuten Mobile is continually testing and refining new network functions and features. 

Rakuten Mobile’s cloud infrastructure is underpinned by three central data centers. Equipped with thousands of servers, these data centers run the core network functions, as well as the services that run on top of the network, such as chat and video streaming. If an earthquake or other natural disaster were to take one of these data centers down, the others can continue to run the network.

As 5G is about collaboration and expanding the businesses, by utilizing the ultra-low latency and high bandwidth capabilities, some services, such as online gaming, sit in the regional data centers.

Ryota Mibu, division manager for infrastructure platforms at Rakuten Mobile, Inc.

Getting closer to customers with 5G

As Rakuten Mobile rolls out 5G, it is also deploying about 50 regional data centers to support the radio access network, together with applications that need to be closer to end-users to be sufficiently responsive. “As 5G is about collaboration and expanding the businesses, by utilizing the ultra-low latency and high bandwidth capabilities, some services, such as online gaming, sit in the regional data centers,” explains Ryota Mibu, division manager for infrastructure platforms, Rakuten Mobile. According to Mibu, running these services in the regional data centers dramatically reduces the latency (the time it takes for an application to respond), compared with running the services in the central data centers.

These regional data centers will also be supplemented by thousands of “far edge” data centers. In these facilities, a small cluster of servers processes the data being captured by the base stations and relay it on to a regional or central data center.

The challenge for Rakuten Mobile and other cloud-based companies is to balance the benefits of centralization with the benefits of locating some infrastructure closer to end-users. As well as making services more responsive, the distributed approach can reduce the bandwidth bottlenecks that might occur if all of an organization’s traffic had to flow in and out of a handful of centralized data centers.  A distributed cloud architecture can also help companies comply with regulatory requirements to store certain types of data sets within their legal jurisdiction.

The beauty of standardization and simplicity

Although they are very different in size and role, the three different kinds of data centers are united by the same standardized architecture. “One of the beauties of standardizing the platform from the central data centers to the far edge data centers, is the simplicity — we don’t need to have extra engineers or extra testing,” Mibu notes. The simplicity and uniformity of Rakuten Mobile’s cloud architecture also makes it easier to automate the management of the network.

Rather than relying on vendors to run its cloud, Rakuten Mobile employs experienced engineers to work in what is known as DevOps — an environment in which the traditional line between software development and operations is blurred. This approach enables engineers to develop new code in the operational cloud environment, allowing for faster refinements than if these two domains were kept separate. This approach makes it possible for Rakuten Mobile to add new functionality to its network in around six days — a significantly shorter timeframe than the six-month cycles of a traditional network operator.

“We do quick development, operation and production and improve things constantly,” stresses Mibu. “Any developer that wants to gain experience of a real DevOps environment should join us.”