The cloud-native core is certainly the best entry point for IT and data center vendors in the 5G ecosystem. Groundwork has been laid in the virtualization of the 4G evolved packet core (EPC), and even if the NFV-centric approach is rapidly looking outdated, it has been valuable to teach operators and suppliers about some of the challenges of making a large-scale core perform optimally on cloud infrastructure.
In 5G, there will be many more opportunities for vendors to make a difference to operators’ performance by distributing the core widely, in conjunction with edge compute platforms, while supporting stripped-down enterprise cores, and moving to flexible microservices that allow both services and suppliers to be swapped and changed relatively easily.
Some IT-centric suppliers will stop there, believing that the 5G RAN – with its heavy dose of physical radio/antenna technology and its eye-watering processing demands – is a step too far and should remain the preserve of the large network equipment providers, along with the new breed of challengers, which are emboldened by the opportunities of open RAN, but are generally companies with years, or decades, of radio experience under their belt.
But operator enthusiasm for open platforms such as O-RAN is also giving new confidence to some of these IT vendors, which are starting to see a better chance of a multivendor RAN in which they could play to their strengths in certain areas – managed services, integration, cloud infrastructure and so on. That depends on operators’ desire to move away from single-vendor, end-to-end solutions translating into new procurements, but if it does, there will clearly be many more entry points to the RAN supply chain that currently exist. And as the RAN itself becomes more industry-centric, with the growth of private networks and slicing, there will be opportunities for vendors with strong ties with large enterprises to leverage those to drive profitable partnerships with RAN specialists, whether established or ‘new’.
HPE made strong moves into the 5G arena last year, with its main focus on the core, and it has enhanced this proposition (see previous item). But its newly named Communications Technology Group (CTG), which made its official debut last week, also has its eyes on the 5G RAN. Led by Phil Mottram as general manager, the CTG brings together all HPE’s telecoms offerings in one unit with a heavy focus on 5G.
Like many relative newcomers to the RAN space, HPE is likely to focus initially on the emerging market for enterprise, private and rural expansion networks rather than invading the macro 5G fortress from day one. It has clear credentials in the enterprise and as it is looking to ease its entry by riding on the O-RAN wave, this is where many of the first revenue opportunities will lie. Enterprises and their mobile network providers are also more likely than macro MNOs, in the early stages of 5G, to turn to managed services from IT and cloud partners, which are a strength for HPE.
The vendors said its Open RAN Solution Stack will be commercially available before July this year. It brings together existing software technologies for orchestration and automation, optimized for the vRAN, as well as a set of RAN infrastructure blueprints to ease deployment, especially for operators new to cloud-based networks.
On the software side, it includes intent-based orchestration and AI-based automation to allow MNOs to manage large numbers of virtual network functions (VNFs) and containerized network functions (CNFs). The orchestration technology is the same that underpins the 5G core.
Importantly – and by contrast with some software-only IT players with their eyes on the RAN – HPE can also provide underlying hardware. It has announced a server that is says is fully optimized for 5G RAN workloads in an open RAN environment, the ProLiant DL110 Gen10 Plus Server, based on third generation Intel Xeon Scalable processors.
Just as some vRAN start-ups have found that it is valuable, in some scenarios, to be able to offer their own radio hardware, so the same will be true on the cloud side. Operators, at a strategic level, are vocal about their desire for fully multivendor networks, like Rakuten’s, in which each element of hardware or software can come from a different supplier. At an operational level, the same operators can be far more cautious, no doubt mindful of the huge team of engineers that was needed to make that Rakuten network function.
In reality, there will be many situations in which operators want hardware and software from the same vendor, and in the RAN, the challenges that hardware – whether radio heads or cloud – must address are often more daunting than those facing the software (virtual network functions and orchestration).
HPE CEO Antonio Neri said operators are, in reality, wanting a more integrated offering combining services and components, in order to hit maximum performance and reliability levels in the 5G RAN.
“They don’t like to bring together what I call widgets that they have to put together themselves. What they want is complete solutions that are out of the box able to be deployed and managed in a cloud-native way,” Neri said. This is the rationale behind the creation of CTG – “We felt this is an opportunity to act now and bring in every aspect of what we do into one integrated organization and architecture was the next natural step.”
The need to protect operators from the demands of optimizing hardware is why the actions of the big chip vendors such as Intel (see item below), Nvidia, Marvell and Qualcomm are so important to making the 5G vRAN work at scale; but until they produce chips that can be easily deployable by true white box suppliers, there is still considerable onus on the vendor of the server or the radio head to enable optimized performance for the whole vRAN.
The company claims its server supports multiple options for hardware acceleration of demanding RAN processes and is capable of powering a baseband unit, fronthaul gateway, and cell site router in a single short-depth server, which also ties into its edge computing platform.
Moving early in the 5G RAN server space also puts HPE in a good position to influence open specifications, a strategy Nokia has adopted in O-RAN RIC. The O-RAN Alliance has not yet ratified new server specs, but HPE is a member and says it has worked closely with operator members to be confident that the ProLiant specs will be “very close” to eventual O-RAN standards, and so should support any vendor’s O-RAN software.
“For us, cloud is not a destination, it’s an experience that we bring to all the applications and data wherever they live,” Neri said. “When I think about our vision as a company to become that edge-to-cloud platform-as-a-service company, we have the opportunity to bring that cloud experience all the way to the far edge, not just the edge of the telco, but the edge of where we live, where we work.”
The blueprints are also an interesting offering. Industry bodies such as Small Cell Forum, Telecom Infra Project and NGMN Alliance have a history of defining blueprints to ease deployment and providing common frameworks and best practice. The weight of a large vendor behind such efforts can be valuable in winning operator trust and support, provided they do not become too closed and specific to that company’s solutions.
The would-be kingmaker here is Intel, which is working on reference designs and blueprints not only with HPE but Google (see below), VMware, Red Hat and with several operators, such as Verizon and Dish. HPE said blueprints were important at this early stage of open RAN because, given the nascent nature of the technologies and specifications, the first networks have had to be built on bespoke configurations that do not drive scale. It said its Telco Infrastructure Blueprints will help lower risk and complexity, and enable more software vendors to enter the platform easily.
“We have a comprehensive set of solutions that we can actually integrate, whether it is the virtualization of the core, or whether it is at the edge with vRAN, and then ultimately even providing the last mile of connectivity with WiFi and all the multi-protocol aspects of it,” Neri said.
Mottram said HPE was currently in active discussions with more than 70 telcos and aims to deploy live commercial services with some of them from mid-2021. It currently has over 300 telco customers for various data center, software and services offerings.
An important customer in the 5G arena is Orange, which has been working with HPE to trial an open core, and is “going further and testing the concept of a 100% software-based, data and AI-driven, fully automated 5G core”, according to Emmanuel Bidet, VP of convergent networks control at Orange. “HPE brings a set of cloud-native, standards-compliant network functions, and the automation of slice management that can be replicated for end-to-end integration in each country. The first results are really encouraging,” he added.
In September, HPE claimed six customers for its then-fledgling hosted 5G core offering, one of them an (unnamed) tier 1 telco in north America. The company has developed cloud-native implementations of six 5G core network functions, which can be deployed as-a-service or licensed to telcos, and can be combined with functions from other suppliers too, on common cloud infrastructure. These have now been combined in the integrated stack described in the previous item.
HPE is aware that ending single-vendor lock-in is an important driver for operators to consider new architectures and to invest in cloud-native platforms, so it has developed a long list of partners which could also provide some of the network functions, which HPE could then integrate on its cloud infrastructure. Partners include Microsoft’s two recent core acquisitions, Affirmed and Metaswitch; Openet (just acquired by Amdocs); as well as Casa Systems, Mavenir, Samsung IBM/Red Hat, VMWare, Apstra, AMI, Tech Mahindra and World Wide Technology.
HPE has already migrated some of its well-established telecoms products, such as its Home Subscriber Server (HSS), to the cloud and with as-a-service options. The HSS is highly strategic because it will not just support existing MNOs, but can enable new entrants to deploy private or semi-private enterprise networks.
“HPE provides the underlying infrastructure as a service and then we layer on top of it the platform as a service. We layer on top of that the container platform as a service,” said the vendor. “We’ve been partnering so far with Red Hat for OpenStack and OpenShift. Very soon we’ll have VMware and Tanzu in the lab.”
However, HPE is not hedging its bets by developing cloud-based networks directly for enterprises, even though it is arguably throwing away one of its clearest advantages. Instead, it wants to go wholeheartedly after service provider business with telcos, private network operators, hyperscalers and system integrators.
“You’re not going to see us go direct to enterprise,” it said. “What we’re going to do is be an arms dealer to the CSPs and the hyperscalers.”