Cloud-RAN (or Centralized-RAN) can inject new efficiency and flexibility by virtualizing the functions of a RAN and allowing large numbers of large or small cells to share a common baseband resource in the cloud. That, however, has limitations when it comes to low latency, and will often be complemented by an edge-based topology such as ETSI’s Multi-access Edge Computing (MEC), which will allow both the access point and the IT platform to be located close to the user, and potentially in the same housing (though that is controversial).
The upshot is that operators may be turning their baseband network functions into software, but they will need far larger numbers of cell sites, fibre links (long and short range) and new quantities of IT infrastructure (owned or outsourced), in order to move ahead with the new networks.
Some are already starting to plan for this dichotomy of a virtualized, software-centric network which nevertheless needs huge numbers of physical assets, which therefore need to be readily affordable, accessible and deployable.
AT&T said recently, when it was showing off its ‘5G Evolution’ (aka LTE-Advanced) technology in Indianapolis, that 5G services such as self-driving cars will simultaneously use functions which are processed inside the car (braking), on roadside servers (advanced driver assistance), or in the cloud (navigation). It said these various types of updates will involve “tens of thousands of central offices, macro towers, and small cells usually never farther than a few miles from our customers”.
One way to create an architecture which can support so many types of processing for just one end device – in that case a car – is to virtualize heavily, as AT&T is doing. That enables resources to be assigned dynamically where they are needed, and to be swapped in and out easily as requirements change. Gordon Mansfield, AT&T’s VP of RAN and device design, said a virtualized and centralized RAN is important – in parallel with a virtualized packet core. Together, these allow functions to be created and then dispersed through the network wherever they are needed.
Like many operators, AT&T sees this as a two-phase project, centralizing first and then virtualizing. This should help to address some of the particularly thorny issues around virtualizing the RAN. While AT&T aims to have 55% of its network functions virtualized by the end of 2017, the RAN is the toughest element because it still involves so many physical components such as antennas and cell sites.
“We’re doing a lot of proof of concept work in that area,” Mansfield told reporters. “As quickly as the technology is ready then we’ll start that pivot. A lot of the work we’re doing right now is kind of a stepping stone. You’ve got to go build out the centers where you’re centralizing the capability, and once you’ve centralized, it makes it easier to virtualize. So they go hand-in-hand.”
While virtualization eases the process of densification because updates can be created once in software and distributed to all the sites, it does not solve the issues of siting and backhaul. Mansfield acknowledged that, while high traffic requirements were the natural top criterion for choosing a location to densify, sometimes it also depended on how easy it would be in terms of site acquisition, permitting and construction.
“It’s not just the permitting aspect. That’s a big part of it, but it’s also the ability to go and construct and quickly get the transport,” he said. “You’ve not only got to build the radio infrastructure, but you’ve got to get the fiber fronthaul or backhaul to deliver on the higher speeds we’re targeting with this.”
AT&T has been trialling a Centralized-RAN small cell network in San Francisco and is now repeating that in other cities, to support LTE densification and to lay the foundations for a smoother migration to 5G in future. “The main reason is it helps us to shift the topology to prepare for millimeter wave and 5G capabilities that will follow,” he told FierceWireless. AT&T plans to deploy 5G New Radio first in mmWave spectrum at 28 GHz, initially for fixed wireless access.
It is not just the well-known 5G frontrunners which are openly grappling with the challenges of sites and fiber. The first public test of an emerging technology called RAPID 5G has been conducted by Polish service provider Exatel, in a live data transmission in the Blue City shopping center in Warsaw. RAPID 5G is one of many projects which are trying to think beyond current backhaul and fronthaul technologies in order to prepare for very high data rates and very low latency, expected in second- or third-wave 5G deployments. A collaboration between the EU’s Cordis program and Japan, it is targeting 10Gbps data rates with ultra-low latency and very large numbers of devices, making it suited to high density environments like stadiums or malls. The first Japanese trial will take place in a football stadium.
The project – whose name stands for ‘Radio technologies for 5G using Advanced Photonic Infrastructure for Dense user environments’ – is proposing a C-RAN architecture which could mix and match 3G, 4G, 5G and millimeter wave access points, but using photonics to connect these to the centralized baseband (fronthaul). This would result in a radio-over-fiber approach which the participants in the project say would reduce the cost of transmitting very wide bandwidth and high frequency radio signals.
The Polish field test transmitted 4K and 8K video streams through 5G antenna modules installed around the shopping mall, over mmWave radio and optical fiber. The video was received by a moving terminal and a PC with virtual reality goggles. The goal was to test interoperability of all elements, especially when converting optical baseband signals into the mWave frequency range by optical means. The trial used the Optical Network Evolution (ONE) platform from DAS vendor Corning, an all-optical backbone to support cellular, WiFi and Ethernet backhaul.