Why a centralised-distributed architecture is the only viable long-term RAN solution
Where are we now?
Initial 5G deployments are referred to as “Non-Standalone (NSA) 5G” and based on Dual Connectivity. This is where the device connects to a 4G “Master” LTE base station first, and then, under the direction of the master, establishes 5G connectivity.
These early 5G deployments use spectrum in Frequency Range 1 of 5G. In Europe, this is likely to be in the band 3.4-3.8 GHz (but not exclusively). In terms of coverage, the 5G cells that can be supported at these frequencies within towns and cities, allow 5G “secondary” base stations to be co-located with the 4G “master”. These base stations are invariable well-sited in secure locations and equipped with good backhaul connectivity, mostly in the form of fibre links. In effect, 5G is being used to add capacity to the existing 4G site, whilst bringing additional efficiency in terms of more advanced antenna techniques.
What happens next?
As more 5G spectrum is released, operators will start gaining access to millimetre wave frequencies. These are much higher (set to be 24.25 GHz to 27.5 GHz in Europe) than traditional cellular frequencies. The characteristics of the millimetre wave radio channel are very different, with much greater attenuation (loss of signal) over distance, and decreased diffraction effects meaning any “signal shadow areas” behind buildings aren’t filled in as well as they are at the lower frequencies. This all means that the 5G millimetre wave cells’ coverage is limited practically to around 200m, and largely line-of-sight coverage.
However, at these higher frequencies, there is a lot of additional available spectrum (and therefore capacity). Also, the high attenuation with distance means cells can be deployed in isolation (single or small groups), very much reducing the overall network interference. Hence, the greatly increased spectrum can also be re-used at much reduced minimum distances compared to the traditional cellular spectrum. Finally, smaller antenna sizes mean antenna arrays, rather than single antennas, will be used. This opens up a whole new world of highly efficient antenna techniques – including advanced beamforming. Therefore, overall, current industry thinking is that small cell (millimetre wave) 5G is very much worth the investment.
How do we maximise investment?
There are a number of considerations that will drive the way in which 5G will be deployed once millimetre wave spectrum is available. All of them have an impact on ROI, and in the main, point us towards a Centralised-Distributed 5G RAN architecture:
The smaller Millimetre Wave Cells need to work in a master-secondary arrangement with the larger LTE cell - or with lower frequency 5G cells in the “Stand Alone” 5G deployment mode. This will act to combine good coverage and reliable connectivity provided by the Master, and additional capacity available through the Millimetre Wave small secondary cells. A number of secondaries can be suitably located to work with a master to provide Dual-Connectivity for different devices in the overall coverage area of the master cell. The devices’ connectivity will be switched between master and secondary (and back again) as radio signal strength varies. Statistically, if enough devices are using the secondaries, there will be enough overall capacity, even if some devices are temporarily using the capacity of the master.
The existing LTE sites are generally well-equipped in terms of backhaul connectivity – usually in the form of fibre. If a secondary site can self-backhaul to the master, the data can be aggregated through the existing fibre connection towards the Core Network. With so much spectrum available at the millimetre wave frequencies, it should be relatively straightforward to use some of it to form a microwave link (or to beam-form) from the secondary to the master in order to provide the self-backhauling connection. Of course, line of site would be required, but in general this is a much cheaper option than cabling out to each and every secondary.
The link from the base station itself to the antenna(s) can take many forms. However, the final connection is traditionally in the form of digital representations of the actual analogue wave that is to be applied to each antenna. With 5G, the massively increased number of antennas mean the amount of digital information is simply not feasible to move over any kind of distance (even on fibre). The obvious solution is that the final stage of processing (very lowest layer in the protocol stack) is likely to be in the antenna unit itself.
This essentially means that for millimetre wave (small-cell) base stations, the base station functionality will inevitably be separated. What form the separation of functions takes, and where the rest of the functionality is located is open to question. Should it be at the small-cell site (acting as a distributed unit – DU), at the master site (now acting as a centralised unit - CU), or a mix – with higher layers at the centralised master site and lower layers at the distributed site (with the very lowest sub-layer in the antenna unit)?
Ease of deployment
If we choose to deploy most of the functionality at a centralised site (possibly the master base station site, or a centralised RAN site), it means additional small cells can then be deployed cost-effectively as small units of hardware. The capacity would need to be sufficient, but any additional cell deployment would require minimal upgrades to the CU. The deployment could be further simplified if robust Self-Optimisation, or Self-Organisation techniques are used as part of the small cell deployment process.
Security is of growing concern for any operator, and 5G has made significant improvements in this area when compared to 4G. However, the actual 5G radio interface encryption and integrity protection (the two primary radio interface security mechanisms) operate between the device and the base station. The actual security functionality is part of one of the protocols located higher in the protocol stack. Hence, it makes sense to locate that part of the base station functionality in a more secure location – in this case a centralised unit, rather than the distributed unit.
There are other security mechanisms in place to protect the Centralised Unit / Distributed Unit connection, but nevertheless, security is best served in terms of reliability by locating the encryption function itself in a more secure location.
To reduce latency in support of a wide range of new use cases and services, location of computing resources at the edge of the network is very much a requirement. This could be at a suitable RAN site, or co-located at the base station for especially low-latency service support. This potentially throws up all sorts of concerns, with two major ones being security, and the physical positioning of the actual processing hardware.
The balance is to gain access to low latency support whilst reducing the risk and maintaining ease of deployment. Again, a centralised-distributed architecture seems to be a likely solution, where the computing resources are located at a relatively secure and well-equipped CU.
Finally, and a bit further along the time-line, 5G potentially brings in a completely new way of thinking about radio coverage – one that could significantly improve connection efficiently and reliability. Based on cell-free coverage and extensively using beams, it requires specific radio unit positioning in order to work optimally - ideally, very regular positioning in specific patterns. Small units would be much easier to position than larger units. This again points towards a distributed architecture, where the DU is kept small and light.
It all points to a centralised-distributed 5G RAN architecture
Although these considerations (above) are not new, the 3GPP 5G specification documents include newly standardised interfaces that allow the gNodeB (the technical name for the 5G base station) to be deployed very flexibly. It could be as a complete integrated base station, or separated into the different functions as discussed in this blog. At first glance, it just seems like we have options. In reality, the way in which 5G is deployed – the spectrum we use, the use cases that we support, concerns around security, and ease of deployment, amongst other concerns – will all dictate the actual architecture. For the reasons set out in this blog, it seems to point firmly towards a centralised-distributed architecture for the base station and RAN in the medium to long term.