From chiplets to open slicing—developments giving hope for ‘true 5G’

The many promises made for 5G boil down to two categories – cost and versatility. On the one hand, they must deliver significantly reduced total cost of ownership (TCO); on the other, the flexibility to support hundreds of use cases with differing network behaviors, and to amend or replace them rapidly in response to changing requirements – thus generating new revenue streams for operators, across many industries.

In these very early days of commercial 5G, there is a broad spectrum of opinion about how, when or even whether the platform will be ready to fulfil its promises. Nobody would argue that the current iterations can do so, in any but the most limited way. The next release of standards from 3GPP will add new functionality and flexibility, but far more important is the way that operators actually deploy these standards. At the heart of it all will be an open, disaggregated and virtualized RAN and core, capable of making futuristic systems like network slicing a reality.

Many operators believe that these new networks must be multivendor, to end lock-ins, bring down costs and ensure platforms remain open in the real implementations, not just in the standards. Telco support for open source initiatives like those housed in the Linux Foundation’s LF Networking Fund, or open industry groupings like Telecom Infra Project, have forced the large vendors to participate in projects like the ORAN Alliance. Even if the large OEMs, in the end, still supply most of the virtual network functions (VNFs) for the virtualized RAN baseband and cloud-native core, they will still have had to adjust their approach, and even their pricing, to a more open environment.

Openness is even coming to hardware – to white boxes can be programmed to support server, switch and other personas; even to radios and to the chips which underpin it all. Vendors are slowly moving away from proprietary silicon to merchant chips; and from single-purpose semiconductors to a combination of general purpose processors and programmable accelerators. This opens new opportunities for mainstream chip vendors to get into the heart of base stations and other fortresses, and to build broad platforms to support the diverse needs of a virtualized RAN and core. But it also encourages uptake of licensable technologies like ARM processor IP or CEVA DSP cores; and even a rising interest in open source architectures like RISC-V, or the emerging ‘chiplets’.

Some of the week’s most interesting events relate to the trends outlined above – towards virtualized RAN and core platforms that are fully multivendor, and open from chip layer upwards, thus tantalizing us with the prospect that radical changes like dynamic network slicing might be achievable, and that 5G might deliver its key promises.