Slightly tired by hype around NFV and SDN – and by the fact that everybody has to find some way to add these acronyms to their slide decks – I decided to write a report on the topic to give at least to myself some peace of mind. Is it really worth all the fuss? And how pervasive will it be? Will it be a piecemeal, gradual change that operators have to go through just to keep up with increasing traffic loads and monetization targets? Or a deeper evolution of the overall network architecture that extends far beyond the functions that are being virtualized?
I am still looking for the answers – and would love to get your take – but my initial observations and feedback suggest that the evolution towards virtualization is going to be more enticing than technological, performance and cost advantages alone would predict.
The path to virtualization will be necessarily gradual – perhaps painfully so – because much is at stake and the transition has to be well executed. Operators cannot sacrifice reliability for the sake of agility and flexibility. But even though initially it will be limited to a few functions, the impact from the beginning will be felt across the entire network, because it fundamentally changes the way mobile operators plan, deploy, test, monitor, operate and expand their networks.
You can make a good financial case for NFV and SDN just on the basis of cost savings (especially on the capex side), but I doubt that this is going to be the main driver – or a sufficient one to justify the pain that operators have to go through as they move from legacy to virtualized environments. The use of commercial-off-the-shelf hardware is highly desirable, but could turn out to be a double-edged sword if deployed too hastily.
A main driver for NFV and SDN is the realization that current networks are too rigid, unyielding and slow to extract all the potential that new technologies and solutions contribute, and to adapt service creation to subscribers’ changing usage models and to revenue-generating targets. The flexibility and agility that virtualization brings – and that increasingly operators recognize as its main benefits – reduce the amount of fine-grained control that operators currently have over individual functions. This change is going to be challenging for operators (and, possibly, for vendors as well) from a cultural and operational perspective, even when the technology works flawlessly.
Sometimes virtualization is presented as a deus ex machina – the network will run itself magically, automatically optimize resource allocation, and effortlessly create new revenue. This is unlikely to happen. It will be a challenging journey to learn how to leverage the new freedom and to balance tradeoffs that increase the network potential but also have unexplored effects – in order words, to manage and orchestrate networks in which at least initially a mix of virtualized and legacy functions will have to coexist seamlessly. This is why – I have come to realize – it is important for all vendors to have an NFV and SDN story: no matter what they sell to operators, their products will sooner or later have to operate in a virtualized environment.
In legacy networks, it is relatively straightforward to identify what does not work and whose fault it is. Virtualized networks will make it more difficult to identify the source of performance issues and possibly to fix them. In turn this increases the demands and complexity of network testing and monitoring, but it also adds to their relevance. Mobile operators cannot test or monitor individual elements, they need to look at the end-to-end performance (which is what matters even in legacy networks, but it is not what performance metrics today capture), and when they spot an issue, they must try to identify the source in an inherently dynamic network.
The impression I have from talking to operators and vendors and that I found intriguing is that the overarching impact of virtualization is to complete the transition to a full-fledged, open IP infrastructure that will increasingly make mobile networks behave and look like any other network. Even though requirements that are specific to mobile networks and their capacity-constrained RANs will remain, the exceptionalism of mobile networks is bound to disappear as the share of wireless traffic increases. Mobile and Wi-Fi traffic currently account for about half of all IP traffic today, and mobile operators can no longer afford to treat wireless traffic as a luxury item. Virtualization will be both a driver and an enabler of the full transformation of mobile networks into IP networks.
I look forward to exploring these issues in my upcoming report. If you would like to contribute to the report or in sponsoring opportunities please contact me at firstname.lastname@example.org