The Open Networking Foundation (ONF) announced the formation of the SD-RAN project (Software Defined Radio Access Network) to help create open source software platforms and multi-vendor solutions for mobile 4G and 5G RAN deployments. The first O-RAN project the ONF is going to take on is building an open source controller for open RAN networks, aka a RAN intelligent controller, or RIC. Functionally, this controller will mediate between open RAN hardware and application software modules that have been dubbed “xApps.”
Network operators have been adopting open systems for some time. It started with software-defined networking (SDN), which brought the open-systems approach to the network core. That trend that was furthered by network function virtualization (NFV).
The last major parts of communications networks that are still built almost entirely using proprietary technology is the RAN. This is the part of the network dominated by Huawei, along with Nokia, Ericsson and, to a lesser but growing extent, Samsung.
“The vendors selling boxes? They don’t want to cede control,” Timon Sloane, ONF’s VP of ecosystem and marketing, told EE Times. “But the operators feel very strongly about this. They want a powerful RIC, and they want powerful xApps to sit on top of the RIC.”
Some of the world’s largest network operators have been agitating to migrate to open RAN systems. Some of those are founding members of the ONF’s SD-RAN project; they include AT&T, China Mobile, China Unicom, and Deutsche Telekom. Other founding members of the project include Facebook, Google, Intel, NTT, Radisys and Sercomm.
The vendors are controlling the argument – and the market – at the moment, however, largely because a suitable RIC doesn’t exist, nor do any xApps. The vendors say it can’t be done, Sloane said, “and until someone proves they can do it, the vendors will keep saying it can’t be done.”
Most of the very many companies interested in getting a piece of the RAN market are members of the O-RAN Alliance. Where the O-RAN Alliance is a standards and trade group, and the brand new Open RAN Policy Coalition appears to be a lobbying operation of some sort, the ONF is largely an engineering organization. The ONF has been integral in developing SDN/NFV technology, and it found itself in a position to remain involved by defining a near-real-time controller unit that is intended work with any OEM’s hardware and with software from any applications developer.
ONF calls that controller the µONOS-RIC (pronounced micro-oh’-noss-rick). One of the advantages is supposed to be that it isn’t being built from scratch. It starts with µONOS, which the ONF described as a microservices-based SDN controller created by the refactoring and enhancement of ONOS, the leading SDN controller for operators in production tier-1 networks worldwide. µONOS-RIC is built on µONOS, and hence features a cloud-native design supporting active-active clustering for scalability, performance and high availability along with the real-time capabilities needed for intelligent RAN control.”
Inherent in open communications infrastructure is disaggregating, well, everything that can be. Certainly, hardware and software, but also the hardware itself. The industry has decided that the Radio Unit (RU), which is basically the antenna and associated electronics, can be a discrete system. This could potentially be a big boon to companies with RF expertise which will be able to do business with greater independence.
What had been one monolithic baseband unit has been split into two functional units: a centralized unit (CU) and a distributed unit (DU). The simplicity of the nomenclature belies the difficulty of finding the most effective way to split a baseband unit in two; the industry is still hashing out the details.
The µONOS-RIC will use the O-RAN Alliance E2 interface to connect with vendor-supplied RU, DU, and CU RAN components.
xApps running on top of the µONOS-RIC are responsible for functionality that traditionally has been vendor-proprietary, implanted in proprietary basestations. The foundation said that xApps will also apply AI and ML to optimize RAN functions.
The SD-RAN project already has a working skeleton prototype of the µONOS-RIC controller above a RAN emulation platform through the E2 interface. This implementation is demonstrating handover and load balancing at scale, supporting over 100 base stations and 100,000 user devices with less than 50ms handover latency (less than 10ms latency for 99 percent of all handovers). It’s a highly encouraging development suggesting that advocates of open RAN will, in fact, be able to do what proprietary RAN vendors say can’t be done.
Sloane pointed out that open RAN is eagerly awaited not just among network operators, but in the enterprise sector as well. He noted that the FCC is working to open new (and robust) unlicensed spectrum
Enterprises aiming to create private networks (to run factories, for example), won’t have to shell out billions of dollars for spectrum, Sloane, observed. “We see that as a huge opportunity, and we’re seeing an explosion of activity there.” With all of the extras built into 5G, enterprises will be able to do so much more with artificial intelligence, machine learning, quality of service – “they can do things hadn’t been possible before,” he said.
Separately, the FCC is currently wrapping up an auction for Citizens Broadband Radio Service (CBRS), a 150 MHz wide chunk of spectrum in the 3.5 GHz range; that spectrum is likely to go to phone and cable companies.
Those involved in the SD-RAN project are shooting for a field trial by early 2021.
Sloane said that the ONF’s activities will remain completely complementary to work taking place within the O-RAN Alliance, the O-RAN Software Community and the TIP OpenRAN Project Group.