Your browser is not supported. Please update it.

11 October 2022

Telco cloud struggles to deliver on early promise amid recreation of silos

Frustration among operators over unfulfilled promises of the telco cloud have been simmering for at two years and have now boiled over for some. Orange has become one of the harshest critics, after Laurent Leboucher, group CTO at its Innovation Networks division, condemned the current state of the telco cloud as a “nightmare” during a keynote at the NGMN Industry Conference and Exhibition 2022 held in Paris in September.

Leboucher accused the industry of allowing competition to create new silos as vendors sought to establish their own vertical combinations of hardware and software in complete contradiction to the original concept founded on virtualization, as was already well established in the non-telco world of enterprise data centers. This was denying operators the very operational benefits and efficiencies they had been promised cloud platforms would deliver. Leboucher added that the task of simplifying the telco cloud was “a very big challenge” for Orange but was a “nightmare for everybody”.

Leboucher’s gripes condense down to two main complaints, firstly that the telco cloud has failed to achieve the promised escape from vendor lock-in, and secondly that it has also if anything increased rather than reduced complexity. If true this is concerning for operators in the context of Open RAN, which relies on telco cloud concepts.

The telco cloud evolved before Open RAN around 2012 with the promotion of Network Functions Virtualization (NFV), an idea already proven in the IT world, in which core operations such as firewall processing and routing are migrated from dedicated hardware to common off-the-shelf (COTS) servers where they run as virtual machines or containers (or containerized VMs).

Operators could then employ lower-cost generic hardware, which could be separated in location from the overlying processes and therefore deployed anywhere in the cloud. Workloads could then be shifted and located judiciously to minimize latency or bandwidth consumption for example. Data -intensive user functions could be executed in a hub closer to the user, so that less traffic went to and from central control points.

Open RAN  arose later, driving common interfaces between various technology elements in a bid to break the stranglehold of a few big vendors and reduce both capex and opex. The first step in this direction came with Centralized RAN, initially still single-vendor and featuring centralized and occasionally virtualized baseband units (BBUs), but still with proprietary interfaces closed to competitive providers.

Then came virtualized RAN (vRAN), which introduced the fully virtualized baseband, which could sometimes be split into two parts – centralized units (CUs) for higher layer network functions and distributed units (DUs) to run Layer 1 and 2 functions close to the cell site. But these were still integrated by single vendors. This then provided the foundation for Open RAN, with fully open, disaggregated, virtualized and cloud-native platforms featuring a complete split between the CU and DU.

Leboucher was not talking specifically about Open RAN, in which, ironically, Orange had just claimed to be making significant progress after striking a new agreement with NEC and Mavenir to extend its standalone end-to-end 5G cloud-native network with Massive MIMO technology (see separate item). Orange joined the two vendors in claiming this was a major stepping stone towards Open RAN deployments, by enabling interoperability between radios and virtualized distributed units over the O-RAN Alliance Open Fronthaul Interface.

But that is a technological rather than strategic milestone, and Leboucher’s complaint was aimed at other major vendors weighing in with what he saw in effect as proprietary offerings. Such vendors might be collaborating in development of the platform but still in effect shutting out others not involved.

He may even have had in mind major IT vendors such as Dell, which has been collaborating with California-based Wind River for its 5G edge development system called Studio, based on the Kubernetes container framework, to develop its modular package called Telecom Infrastructure Blocks. As Dell explained, this combines validated and pre-packaged hardware and software developed specifically for telecom workloads and use cases, available from November 2022. Among these will be foundational building blocks for virtual DUs and CUs, and the site controller management cluster used in Dell’s MultiCloud Foundation.

Wind River Studio provides the software components for expanding networks to encompass 5G edge capabilities, integrating with Dell’s Bare Metal Orchestrator designed to automate as much deployment and lifecycle management of the complete hardware and software stack as possible.

The companies make little secret of their objective to provide a one stop shop for deployment of a 5G telco cloud embracing edge compute. “Setting up a system is simply a matter of adding more Infrastructure Blocks,” said Dell. But these blocks can only come from Dell with Wind River.

In the short term at least, this may well be what operators want, as Kevin Dallas, Wind River president and CEO, indeed indicated. “Our collaboration with Dell will help address complex CSP challenges in deploying and managing a physically distributed, ultra-low-latency cloud-native infrastructure for intelligent edge networks,” said Dallas.

There is the risk though of missing cost savings or capabilities associated with other vendors’ components. In its defence, Dell can point to its recently announced collaboration with Fujitsu to help operators accelerate adoption O-RAN systems globally. But again, this is bi-partite, including integration of Fujitsu’s carrier-grade O-RAN compliant radio units (RUs) with the Dell O-RAN Accelerator Card. Operators will then be able to implement a package including multiband RUs, vRAN and lifecycle management software from Fujitsu, integrated with vRAN and those O-RAN Accelerator cards from Dell.

There are signs though of Leboucher’s complaints being at least partially addressed in the latest announcements by VMWare, one of the pioneers of virtualization in the modern IT era.

Expansion of its partner ecosystem to provide more choice and more automation to ease the operational complexities associated with roll out of vRAN and Open RAN were at the top of the company’s recent telco cloud enhancements.

VMWare highlighted the challenge of embracing and continuing to support vertically integrated legacy platforms while introducing Open RAN and vRAN with better automation and orchestration tools not just for bringing new network functions on board but also integrating with multiple clouds, and edges. Operators are crying out for a consistent management operations toolset that allows them to do this, according to VMWare, which claims to have done its utmost to fill this void.

This toolset is designed to enable greater automation across the protocol stack from the physical infrastructure through network functions to storage management, up to but not including the higher-level applications on top.

The latest enhancements also expand the VMWare partner ecosystem to include at least 275 virtual network functions (VNFs) and containerized network functions (CNFs), including as it happens Mavenir’s Converged Packet Core. The aim is to reassure operators by including only components that have been certified under Telco Cloud Service Assurance, VMware’s integrated service monitoring and network management service for its telco cloud offering that was actually acquired from Dell in 2018. This now supports additional use cases, including communications-as-a-service (CaaS), which is the concept Dell is peddling with Wind River, as well as 4G and 5G mobile cores, network slicing and RAN.

Again, like Dell, VMWare is pushing the idea of bare metal automation, sometimes referred to as metal-as-a-service. This is a step in the virtualization journey where the link between a computer’s operating system controlling vital functions and the underlying hardware is finally severed. Instead, the computer is shipped as hardware and basic functions to support virtualization, without any operating system or applications. The aim is to enable greater flexibility for deployment since operating systems can then in principle work above a flexible platform of distributed COTS machines.

Bare metal automation has become particularly relevant in the context of 5G and edge compute, for at least two reasons. First, under 5G, applications are increasingly being developed around cloud-native design principles, in software containers separated from the underlying hardware, with the telecom network being transformed to a distributed infrastructure. The idea then is that applications can be deployed at the most optimal location according to bandwidth cost and latency, among other considerations. Bare metal automation is required to enable this to happen as smoothly as possible with minimal configuration effort.

Secondly, cloud-native technology with bare metal automation addresses some previous shortcomings with virtualization, essentially achievement of the continuous integration/continuous delivery (CI/CD) model for ongoing software development and revision. This concept, underpinning microservices, enables applications and processes running on the network to be upgraded and revised in small chunks that can be tested and validated more easily as an ongoing process tied in with software development.

VMWare could certainly argue that its latest announcements go some way to allaying Leboucher’s concerns, at least indicating that key participants in the industry are listening.