Both SDN and NFV are terms strongly hyped throughout the industry at the moment. Will these replace the need for Small Cells? Are they complementary, competition or just a distraction? We ask if the promised benefits are the best way to harness evolving technology.
It all started with the core network
Most core network controllers such as MSC, SMSC, HLR all run software which is platform independent. Functions that handle signalling messages, co-ordinate and allocate resources and route calls are all easily hosted on any brand of server. Conversely, those functions dealing with the actual voice and data flows, such as internet content filters, media transcoding and packet routing are implemented in dedicated hardware.
In the past, network operators have a multiplicity of different dedicated telecom appliances for many of these common functions. These had to have very high resilience and uptime to avoid outages. Commercial servers now match that availability at a much lower cost.
Standards have evolved to separate the signalling/control and media processing aspects of the network. Major equipment vendors now offer core network software under licence that can run on most major server brands. Operators bulk buy servers themselves, standardising on the same product lines which further reduce cost, streamline management/diagnostics/repair/upgrade cycles and simplify staff training requirements.
Backoffice servers used for billing, planning, operation and customer service management can also be consolidated to use the same server family. Further efficiency savings come from co-locating these server farms in a few locations. Sometimes operator groups often combine those internationally – that text message to your office colleague may be routed half way across the continent and back without anyone realising.
Virtualisation separates software function from hardware platform
With so many functions capable of sharing the same large colonies of server farms, it then becomes possible to move functions by allocating them to different servers - even in different countries. This NFV (Network Function Virtualisation) contributes to efficiency savings by sharing peak traffic across all available servers. Ultimately, this means a further savings in total server capacity. High capacity datalinks are needed between datacentres to make this work well. The incremental cost of a few higher speed links is more than offset by hardware equipment CAPEX and OPEX.
There are clear commercial benefits from this approach for all parties:
- Operators can streamline and consolidate core network operations in a few locations equipped with common parts bought directly from server suppliers.
- Core network vendors can focus on the applications rather than each trying to build their own high-availability platforms. Frankly, that's the more profitable and differentiating part of their business.
- High availability platform vendors, from HP, Sun, Radisys etc., can invest in developing industry standard products sold throughout the industry.
A different case needs to be made for NFV in the Radio Access Network
A similar approach is being promoted within the RAN. You centralise the common, heavy duty processing in a few large datacentres connected to relatively dumb remote radio heads or antenna. It's particularly aimed at outdoor sites, especially in denser urban areas.
There are a few different flavours of this approach:
- Use remote radio heads (RRHs) connected by CPRI to a "Baseband Hotel". This splits up the traditional macrocell into its two constituent parts, reducing the amount and complexity of equipment at each cellsite. CPRI typically requires about 1-2Gbps per sector, meaning that dark fibre is the preferred solution. Sites can be located several kilometres away, allowing a central location to serve entire metropolitan districts. In some cases, these so-called "Fronthaul" connections could also use wireless – such as demonstrated by eBlink.
- Use Outdoor DAS, where the RF signal is piped to each remote antenna. That would definitely require dark fibre. As with indoor DAS, it can be shared among operators and requires a DAS radio node per site.
This contrasts with the traditional macrocell and small cell approach which requires a much lower backhaul throughput that only needs to carry user data and a little bit of signalling traffic.
All of these architectures require the same three critical elements at each site:
- Approval to install equipment and antenna (Landlord consent, physical access by staff, regulatory approval, planning permission)
Proponents of the NFV approach point to benefits of much higher spectrum efficiency. Tight co-ordination between remote radio heads can achieve a denser level of traffic and higher service quality (e.g. fewer dropped calls/data sessions, faster connection/setup times etc.).
There is also an argument that those expensive baseband resources can be redirected to serve traffic in different places at different times of day. This should reduce the peak capacity required, in a similar way that has been applied in the core network. I don't really buy that story because I think the cost of baseband processing ($100 or less for a Small Cell SoC) is now so inexpensive that it's insignificant in the overall budget.
The bottom line
The trade-off is therefore between:
Any difference in cost between
- the radio node, remote radio head and a small cell (equipment, site rental and installation)
- spectral efficiency of the system
- benefits through equipment sharing
The major saving from NFV in the RAN arises if you could achieve the same throughput and deliver the same service from fewer sites. Each site could use more bandwidth and use the spectrum more efficiently.
However, the cost of each site would (in my view) be lower per small cell because of inexpensive hardware and simpler backhaul requirements.
The centralised baseband equipment and fibre infrastructure to each radio node will cost significantly because of the high bandwidth fronthaul needed.
The lowest cost for spectrum is when you don't need more
Spectral efficiency is a very popular topic when the cost of spectrum is taken into account. Investing in in-building small cell solutions offer great potential in that respect because they don't need any new spectrum.
Outdoors, many operators may still have some LTE bands (especially TD-LTE) as yet not fully utilised.
What the Small Cell industry has to champion is that the incremental low cost of deploying each new small cell is significantly less than that of fibre-connected radio heads or outdoor DAS, and that the spectral efficiency gains using Cloud RAN don't deliver such substantial benefits to compensate. That would benefit greatly from productivity gains from slicker and more efficient deployment and commissioning processes.
Other factors include long leads times for site acquisition, availability of fibre and other practical constraints that are operator and country specific.
The Small Cell Forum is currently working on a study report on this topic, and we await its publication in the coming months.