Network virtualization promises to radically change the way that telecoms networks are designed, built and operated. In doing so, virtualization technologies will profoundly change communication service providers (CSPs), making them more agile, more automated, and better able to compete. This blog post considers whether the same technical advances could also create more flexibility for CSPs that use wholesale access in the last mile.
Could Network Virtualization Transform Wholesale Access in the Last Mile?
By Michael Dargue
Today, wholesale access in the last mile includes regulatory remedies such as local loop unbundling (LLU). Under LLU, competing CSPs serve their customers using the copper access lines of the former incumbent and co-locate their network equipment in its exchange buildings (central offices) to do so.
However, whilst it is feasible to physically unbundle copper access lines, the approach is unsuitable for next-generation fibre broadband over passive optical networks (PONs): physically unbundling individual lines on a PON is both operationally unpleasant and cost-prohibitive. Regulators have therefore fallen back to a mix of access to low-level passive infrastructure (ducts and poles), and active (‘lit’) services such as Virtual Unbundled Local Access (VULA).
In theory, VULA enables CSPs to co-locate additional hardware in the exchange to implement network features such as quality of service, policy enforcement, multicast and local content caching.
VULA was devised by Ofcom and adopted by the EC to address the difficulty of unbundling PON (and FTTC) networks. It embodies many of the characteristics of LLU, but crucially, doesn’t require physical unbundling. Instead, the access network operator provides CSPs with active last mile access services that aim to emulate an unbundled service. There are several flavours across the EU, with the most faithful allowing CSPs to interconnect with the access operator at its exchange and take the traffic from there.
In theory, VULA enables CSPs to co-locate additional hardware in the exchange to implement network features such as quality of service, policy enforcement, multicast and local content caching. To the best of our knowledge, this hasn’t yet happened due to the cost of deploying and maintaining this equipment in a distributed network. However, network virtualization could potentially overcome these barriers.
In network function virtualization (NFV), the physical network equipment is replaced by virtual network functions (VNFs), i.e. software applications that can run on generic x86 servers. This represents a huge change for an industry that has previously revolved around hardware as we discuss in our industry report.
Initiatives such as CORD (Central Office Re-architected as a Datacenter) and TIP (Telecom Infra Project) are seeking to apply the NFV concepts in central offices / exchanges. In this new world, the proprietary hardware appliances for voice, broadband and other services would be stripped out and replaced by racks of generic servers; the functionality of the old boxes replaced by software.
A defining feature of NFV infrastructure (NFVI) is that the servers and storage are a shared infrastructure which can be used by any VNF. Furthermore, as with cloud computing infrastructure, there is no fundamental reason why this shared infrastructure cannot be multi-tenanted, assuming appropriate steps are taken to manage security and resource allocation. In fact, ETSI has already considered the needs of multi-tenancy in its requirements specification for NFV management and orchestration.
In this new paradigm, rather than charge for power and co-location, the access operator would charge for compute and storage – in much the same way as AWS and other cloud services do.
This then raises the potential – at least from a technical perspective – for competitive CSPs to run their own VNFs on a shared infrastructure in the exchange. Doing so would be far less expensive than installing and maintaining their own dedicated, physical boxes and would provide much greater flexibility – services could be instantiated and torn-down in a matter of seconds and minutes rather than weeks and months.
In this new paradigm, rather than charge for power and co-location, the access operator would charge for compute and storage – in much the same way as AWS and other cloud services do. A further benefit is that there would be little or no need for CSPs to physically access the site, improving security and reducing operational cost.
Perhaps even more interestingly, this approach can also be extended to the access network itself. Here, we would still have physical OLTs aggregating the PON connections, but because network virtualization separates the control plane from the data plane, the OLTs become far more open and flexible. Such an approach is described by the Broadband Forum in its recent Technical Report, TR-370: "Fixed Access Network Sharing - Architecture and Nodal Requirements".
The report describes two approaches that enable Virtual Network Operators (VNOs) to share the physcial access network of an Infrastructure network Provider (InP). The intention is to enable the VNOs to "operate, control, and manage their own virtual network [so that they can] provide customized services". This is about as close as VULA could get to physical unbundling.
Returning to our initial question, network virtualization certainly has the potential to transform wholesale access in the last mile, but only if industry chooses to embrace this new model of access competition. Clearly, this isn’t going to happen in the short-term. Network transformation programs can take years to complete, and with network virtualization, network operators are just starting the journey. However, as the network and system architects start designing the network of the future, the mode of competition shouldn’t be ignored.<>