Virtualisation has been a hot topic throughout the industry in recent years with questionable benefits to date. The latest initiatives look to virtualise the RAN but in practice this means choosing where to place certain network functions. But is there a risk of confusion from too many competing options? We consider what business benefits these are most likely to deliver?
Quick recap of Core and RAN virtualisation
Virtualisation in the IT industry delivers a key benefit of using the same hardware throughout large data centres. Commodity server hardware can be used for a wide variety of purposes, driving down the cost of compute power. Many different virtual servers can be supported on a single hardware platform, providing huge flexibility and scalability – whether driving a rarely accessed website, supporting a popular App or a short term highly intensive big data analytics.
Within the core network, SDN and NFV are terms used to implement similar virtualisation with the intent of providing lower hardware costs, sharing commodity hardware platforms with high scalability. Recently, we’ve discussed whether that strategy has been delivering the promised benefits.
Meanwhile, the drive for virtualisation has also been applied to the radio network architecture. In practice, this has taken the form of splitting RAN functionality between the edge and the RAN controllers. In 3G and 4G we had already seen this done with remote radio heads (RRH) connected by high capacity fibre using CPRI. While the CPRI is an industry standard, there has been little vendor interoperability.
What are the business benefits?
I can see that a standardised interface should enable two distinct benefits:
- Fronthaul vendors can provide a variety of compatible solutions that can carry the stated bandwidth with required latency and timing information. We’ve seen this include both wired and wireless solutions from quite a wide variety of vendors and used for a range of scenarios.
- RAN vendors could compete to provide equipment for each side of the interface. This could include lower cost, simpler products being used for the radio head with more of the smarts being retained in the local controller or baseband unit.
What we’ve seen is plenty of healthy competition for (1) but little impact for (2). A potential opportunity would have been to drive digital DAS systems directly with CPRI but there has been very little progress for this.
3GPP Standards for RAN virtualisation
We reported from 5GWorld that 3GPP have walked through many potential options for RAN architecture splits and are now focussing on two. One will leave much of the basestation functionality at the radio node while the other reduces it to more of a dumb radio head. These will take shape and could be used to support both 4G and 5G RAN architectures.
Meanwhile, the Small Cell Forum has invested a lot of energy in a similar effort, working through and discussing the same potential RAN architecture splits. The outcome has been the nFAPI specification. This hasn’t been adopted by 3GPP and from conversations with RAN vendors looks to me unlikely to be developed or supported by the major equipment suppliers. The concept differs and seems to me to go much further than just being a RAN architecture split. Generic commodity hardware could be used to support one or more RAN software Apps (called Virtual Network Functions), potentially allowing a radio node to be provisioned with Ericsson and/or Huawei and/or another RAN vendors software which would run simultaneously. I haven't seen that neutral hardware approach being given much more than lip service because each RAN vendor is touting their own specialist hardware with proprietary RF features such as Massive MIMO arrays.
So I suspect what 3GPP RAN virtualisation will achieve looks to me like being again mostly oriented towards a clear interface specification for fronthaul vendors to support rather than true interoperability. I can’t see that it will introduce a swathe of new radio node or small cell vendors into the mix.
nFAPI has been adopted enthusiastically by the Opensource community and you will see demonstrations and downloadable code which support it. Some may suggest this could enable a completely new ecosystem which will compete with today’s major RAN vendors.
But the maturity of Opensource RAN software is still in its infancy and will take further investment and progress to compete. It does have its uses and applications, and perhaps there are components which could accelerate progress, but it will take a very long time to become mainstream.
On the other hand, it does provide the opportunity for the academic community to experiment and perhaps come up with some unusual innovations.
Don’t forget today’s true challengers
Meanwhile there are already well-defined and clearly implemented interfaces which Small Cell vendors can use for both 3G and 4G into the core network. These are efficient, fully functional and provide for more than adequate performance. Some vendors with very high capacity solutions have developed their own proprietary architectures, such as CommScope's OneCell and SpiderCloud's E-RAN. We don’t need to wait for another standardised architecture split or technology leap to introduce them.
What is required is greater recognition within the operator community that increased competition from a wider range of equipment suppliers brings benefits by driving down costs and encouraging faster feature development. This can be done today without waiting for virtualised RAN architectures which may not live up to their promises.