The business case for IT cloud running enterprise and/or web applications is widely understood. Enterprise cloud takes advantage of the centralization, consolidation and standardization of commodity IT hardware in large-scale data centers. It relies on the very high utilization, based on oversubscription, of this hardware to wring out capex benefits and the automation of infrastructure management tasks to achieve opex savings. IT workloads are by nature discrete, "cookie cutter" executables that can be packed onto standard servers and managed in a uniform way.
In the early days of network functions virtualization, it was assumed that VNFs would run on IT cloud platforms with perhaps a few tweaks to the hypervisor to ensure "carrier-grade" performance. It is now becoming clear that when many VNF workloads are virtualized, they have very different requirements from IT applications. For example, VNFs can have highly specific demands on the virtualized infrastructure, needing dedicated (not oversubscribed) resources to ensure consistent, predictable performance. VNFs may be dependent on one another in service chains, so scaling up one VNF in the chain may require others to scale up, too, requiring more complex orchestration workflows than IT applications.
Operators will be able to achieve some measure of capex reduction by replacing vendor proprietary hardware with standard, off-the-shelf boxes. But resource optimization, which is the bedrock of the IT cloud computing business case -- driving server utilization up and capex down -- isn’t completely transferable to many NFV use cases.
Instead, operators are building their business cases around agility, which they believe will bring both opex and capex benefits. Agility can mean many different things, but essentially, it boils down to being faster and more efficient at delivering to customers what they want. It means cutting out a whole bunch of supply chain logistics around procuring network components, delivering them to the right geographical location -- whether that’s a network PoP, data center or customer premises -- dispatching the right engineers to set them up, which might involve thousands of miles of travel, and maintaining expensive inventories just in case the components break down. The logistics of the physical world require time and generate tremendous cost.
When network functions are virtualized, they can be delivered to production networks anywhere in the world as software blueprints, which an orchestration system can implement locally and automatically in hours rather than months. Through automation, VNF blueprints can be instantiated in multiple locations simultaneously, in contrast to the sequential network builds operators have to carry out today.
Operators have described the cost savings associated with this software-based approach to network component delivery as "dramatic." And these are over and above any savings operators can wring out of equipment suppliers as a result of buying software-only versions of their products and running them on COTS hardware in any location. As one operator put it, the business case for NFV is "blindingly obvious" when based on removing the logistical barriers to rapid network service delivery.
Operators are interested, despite the challenges, in aspects of the conventional resource optimization benefits of cloud computing, for example in rapid scaling of VNFs without having to procure additional equipment and in the opportunity to re-use common infrastructure for other VNFs. Both reinforce the agility/supply chain cost reduction argument. However, to achieve these benefits, operators expect to have to plow back at least some of their cost savings into a new cloud-oriented management and orchestration stack, new operational processes and a lot of organizational and culture change which will be disruptive and expensive in the short term.
As a result, many operators say they are still ‘struggling’ with the business case for NFV. It is not as cut-and-dried as IT cloud computing and the latter has, of course, had longer to prove itself. But the NFV juggernaut is not for halting. Agility has a huge value which operators are only just beginning to quantify. As proofs of concept turn into live deployments, the evidence the market is waiting for will emerge -- and indeed, is already there for basic VNF use cases in B2B scenarios.
— Caroline Chappell, Principal Analyst-Cloud and NFV, Heavy Reading