VMware has achieved a strong position in early deployments of cloud-based 5G networks, with many operators partnering with the Dell-owned company for their telco cloud platforms. Now it is pushing further into the RAN, to ride the wave of operator interest in Open RAN, and to try to stay ahead of rivals like IBM’s Red Hat and Wind River. It also needs to counter developments that would see 5G functions running on bare metal in the public cloud; or tier one OEMs packaging their own cloud platform with their RAN network functions.
The recent announcement that Dish would run its 5G Open RAN in the AWS cloud prompted speculation that VMware’s deal with the US operator might be ending, but in fact, Dish will run its functions on the VMware layer on top of AWS. Dish is using VMware’s platform as an abstraction layer running across multiple network domains.
To keep such deals coming in, the supplier has announced an expansion of its VMware Telco Cloud Platform for the RAN, productizing some of the work it has done for Dish, and with the US new mobile entrant as the launch customer. It has worked with Intel FlexRAN on the RAN offering.
TCP RAN is commercially available today, said Sachin Katti, a strategic advisor at VMware and co-chair of the technical steering committee at the O-RAN Alliance. The platform has two layers, infrastructure and automation, which have been extended for the specific and demanding requirements of the RAN. Katti said that, as a result, the VMware cloud infrastructure is now able to manage the vRAN’s real time workload requirements and compute-intensive processes.
The new vRAN stack also includes VMware’s vSphere ESXi for virtualized and containerized RAN functions, its Kubernetes-based Tanzu for Telco, and its Telco Cloud Automation for network orchestration across network functions, services, and management. It also features Photon OS, a real time Linux container host for DUs running on vSphere; Precision Time Protocol (PTP) accuracy; optional offloading of forward error correction (FEC); and support for Intel Mount Bryce eASIC and N3000 FPGA accelerators.
The platform also enables operators to deploy and operate non-RAN workloads on the same platform, and to run a DU on a real time Linux kernel without requiring that the CU also runs a real time Linux kernel.
“To really bring the cloud to the RAN it has to solve a bunch of pretty important challenges,” Katti told SDxCentral. “Primary among them is the fact that the RAN workloads are extremely real time, and so you have very tight latency constraints … and there are “very unique requirements around access to bare metal hardware” for time synchronization and hardware acceleration.
RAN is also highly distributed, “so unlike the core of the network where you put down maybe 10 data centers for an entire network and collect, aggregate all of the processing needed for the 5G core in those 10 data centers, the RAN by definition will look like 10,000 to many more small cloud deployments or small data centers,” Katti added.
“There’s always this tension between abstraction that a cloud provides, and the performance requirements of highly unique workloads like the RAN, and that’s the tension that we’re solving,” Katti claimed.
He sees vRAN as a “stepping stone” to Open RAN, and the company will provide support for O-RAN interfaces. But operators will have to go through a step-by-step process. Lakshmi Mandyam, VMware’s VP of product management for service provider and edge, said the initial focus was on virtualizing the centralized unit (CU) and distributed unit (DU) functions.
“The first phase is to get those RAN installations virtualized so you can deliver differentiated applications,” she said, along with enabling virtual machine and container management, monitoring and automation. The next phase would be to support open 5G applications that can be run on the RAN element of VMware’s Platform.