Your browser is not supported. Please update it.

14 April 2020

TMO eyes $30m savings from cloud-native, but most MNOs remain shy

While Vodafone touts the role of old-style network functions virtualization (NFV) in protecting its networks from collapsing under the weight of lockdown data traffic, other operators are already in the process of migrating to the newer cloud-native architectures, and the process of network functions cloudification (NFC).

In general, for deployments of the huge scale of Vodafone’s, there are good reasons to use something tried-and-tested, despite its downsides (see separate item), but for those with the confidence to progress to NFC, the results may be more dramatic, and of course, there will be no need to engage in another major migration for many years to come.

T-Mobile USA is not always a radical thinker in terms of the mobile network itself – especially compared to its new acquisition Sprint – but it has pursued a low cost, lean network model which is increasingly software-driven, in order to retain its status as the USA’s most cost-competitive and disruptive national MNO. It is already a major user of Kubernetes, the dominant open container management system, and currently operates 20,000 containers on its virtualized cloud infrastructure.

It is starting to put some figures on the expected results of going cloud-native and introducing microservices and containers. Thom McCann, a senior software engineer at TMO, recently presented some results of the carrier’s seven-year migration to the cloud in a virtual summit run by D2iQ (Cloud Native Virtual Summit featuring Kubernetes). He said that, within the coming year, TMO will reduce its cloud costs by $30m through a combination of centralized management of Kubernetes containers and a normalized approach to clusters.

The tipping point came in late 2018, when the number of containers TMO was running in the cloud overtook the number of virtual machines (VMs), sparking accelerated growth in containerization. “We were on a path toward what I call ‘spending many millions of dollars’ to a path where we spend fewer millions of dollars,” McCann said.

The cloud team is achieving such results through at-scale containerization plus  automation of application development. The efforts are based on a central Kubernetes platform called Conducktor, which offers a common way for any TMO applications developer to access the cloud, security and enabling apps. Conducktor is also supported by the MNO’s telemetry platform, which gathers all the data in its clusters for use by different apps teams.

TMO believes its use of containers enables some of its disruptive ‘Uncarrier’ moves in terms of consumer propositions and pricing, because it reduces the cost of running networks and developing applications, but also makes it more agile.

That agility will be sorely tested during the period of integrating Sprint into the organization. Sprint has made some advanced moves in virtualized networks but has been far less successful in reducing costs or winning new customers than TMO, and the very different cultures of the two organizations may prove hard to reconcile in the short term.

That will test the efficiency of the cloud platform to its limits, to see whether it can still deliver its benefits in a telco that is going through a difficult change process, and therefore defend revenues and customer experience from the effects of that process.

It will also test the cultural shift that McCann says has already happened in TMO as a result of the cloud migration – a highly coordinated approach across many groups.

“It’s taken a coordinated effort across teams at T-Mobile to be able to realize these savings, and to make sure they’re optimized for container environments,” McCann said. Centralized cluster management has been key, as it is hard to track changes and ensure quality if clusters are running in separate workgroups or application teams. But the same cross-function communication filters through to other aspects of the organization, and will now need to be applied to the incoming Sprint units too.

“When you have a normalized approach to clusters and you own the platform that you’re running and most of your application teams are consuming Kubernetes in a normalized way, you then have the opportunity to optimize how you purchase, from a cloud perspective, the underlying resources,” McCann added.

Meanwhile, other telcos are starting to eye containers and cloud-native as they start to plan their implementation of a full 5G core, with greenfield, fully cloud-based deployments like Rakuten’s in mind. This could be based on virtual machines, like most virtualized 4G cores, but since the majority of operators will not deploy a full 5G core for at least two years – often far longer – it will make more sense to adopt a more modern technology.

By the time they are going live, current fledgling vendor offerings should be more stable and proven, though for operators going through procurements processes now, it is not just the products that are immature. The 5G core market itself is fluid – Microsoft’s acquisition of virtualized core specialist Affirmed brought a cloud giant into the market, probably to compete with Nokia (in particular) for the enterprise 5G market. It is unclear how quickly open source efforts like Open Networking Foundation’s OMEC (open mobile evolved core) will find their way into carrier-grade, battle-hardened offerings. And even choosing between the established vendors’ core networks now risks seeming hasty, if a different supplier leapfrogs its rivals in getting a fully-fledged cloud-native product to market.

But the Kubernetes community are eager to bring many stakeholders together to test solutions and prove that the framework is robust enough for 5G core and even RAN implementations. “Containerization in 5G is not only possible, but it can be here today,” said Bill Mulligan, Kubernetes advocate at Loodse, a multicloud enterprise software provider. “The industry is changing and if telcos want to keep up they’d better start examining these technologies.”

The move to edge computing, combined with 5G, will be a catalyst for some operators to take the plunge and adopt cloud-native technology and containers, which allow for functions to be located and moved flexibly, and for lightweight cloud architectures. Several vendors, such as CDNetworks, Edgeworx and Rancher Labs, are repackaging the Kubernetes core to allow it to operate in resource-constrained environments at the edge, which will be important for many enterprise and IoT-focused 5G business models.

Last autumn’s KubeCon + CloudNativeCon event in San Diego, California demonstrated the ambition of this community to penetrate a huge market which has, until now, remained relatively immune to the charms of open cloud technologies.

Pat Gelsinger, CEO of VMware, has said that telco cloud and 5G represent a “great opportunity to participate in an enormous market”, but has also acknowledged the need to reassure telcos that they can achieve the same performance and robustness from a cloud-native network as a traditional one.

At the event, there was a ‘live’ 5G call between San Diego, Montreal in Canada, and Sophia Antipolis, France, presented by Heather Kirksey, VP of community and ecosystem development at the Linux Foundation; Azhar Sayeed, chief architect for service providers at Red Hat; and Fu Qiao, project manager at China Mobile. Behind those public faces, the demo was enabled by over 100 volunteers from 15 organizations including A10 Networks, Alibaba, Altran, Commscope, Eurecom, Intel, Kaloom, Lenovo, Loodse, NetScout, OpenAirInterface and Turnium.

The team were using a prototype 5G core running in a containerized environment, developed as part of the Kubernetes-based virtualized central office project. They used 5G NR radios, with commercial handsets in a custom-built Faraday cage, and spectrum borrowed from an unnamed MNO.  The network connected a live call over points-of-presence in the three cities.

The huge numbers of nodes that a multinational operator will need to control and orchestrate will make virtual machines, and relatively heavy infrastructure like OpenStack, unsustainable for at-scale vRANs and edge clouds – hence the move to more agile, automated cloud-native and containerized approaches.

Such cooperations, whether in open groups like Linux Foundation or between suppliers, will be vital to help address the daunting challenge of running a RAN in the cloud, and trusting the most expensive and vital part of the MNO’s network to the immature technology of containers.

“In a 5G world, new services and applications will not be driven by massively scaled, centralized data centers but by intelligently distributed systems built at the network edge,” said Kevin Shatzkamer, VP of enterprise and service provider strategy and solutions at Dell EMC, in a statement. “