Summary Bullets:
- The ETSI NFV MANO model currently suffers from too rigid functional decomposition and too light interface definition; however, there’s plenty of innovation around to help.
- Vendors shouldn’t be naïve in claiming vIMS multi-vendor interoperability based solely on 3GPP conformance; NFV introduces a very challenging vertical dimension.
One of the main takeaways from our coverage of TM Forum Live! 2015 in sunny Nice earlier this month was around the (im)maturity of the ETSI NFV MANO stack. In short, the ETSI MANO model has been described as having “boxes which are too thick” (defining some functions and omitting others) and “interfaces which are too thin” (lack of implementable interfaces to support multivendor interworking).
So, what are we to make of this? Is NFV doomed? After all, it’s been well said that without MANO, the NFV dream is largely unrealizable.
Thankfully, the answer is ‘no.’ However, it does create some interesting takeaways that both operators and vendors alike need to consider as they move the market forward.
- Nobody should be surprised.
To start with, it’s good to remind ourselves that we’ve been in this situation before. Software moved us away from seeing the ‘boxes’ on network architecture diagrams as simply being hardware ‘boxes.’ That started the move to talk about network ‘functions.’ If you want evidence of this, just look at the endless list of four letter acronyms ending in ‘F’ (for function) in the ETSI 3GPP network design specifications for IMS. In these specs, there was already the concept of network functions living independently of hardware, which is also the central tenet of NFV. So, nothing has changed here – and in the development of the physical IMS architecture, there was as much debate over interfaces and functional content as there now is for NFV.
- But vendors shouldn’t be naïve.
Nevertheless, NFV does introduce some real differences into the mix, but it’s more about network integration than network design. A case in point is the virtual IMS (vIMS). Given 3GPP’s IMS ‘functions,’ it shouldn’t be that difficult to virtualize IMS and it isn’t. Despite its complexity, IMS has been an early NFV use case; the logic being that it is so well defined in 3GPP that multivendor interworking is all ‘done and dusted.’ But, vIMS nicely illustrates what NFV can do even to a stable 3GPP network design when it comes to integration. The problems hinge on the fact that there are now two dimensions to network integration: horizontal (3GPP-defined) and vertical (the NFV impact). Now, there has always has been vertical integration – downwards to hardware and upwards to OSS – but NFV upsets all this. NFV gets in between the IMS and the hardware by way of the NFVI and between the IMS and the OSS via the MANO. The result is immense vertical integration complexity.
Vendors should think twice before claiming ‘NFV multi-vendor compliance’ solely on the basis of inherited horizontal integration credentials.
- And there are ways round it.
That said, the future of NFV isn’t all dark and stormy; there are ways round the vertical integration problem. Solid progress is being made towards both standardizing and productizing the NFVI with the first OPNFV release this month, and the ETSI ISG Phase 2 standardization process of MANO is ongoing with a full program of work stretching into 2016, but with multiple interim deliverables along the way. Meanwhile, some have suggested that MANO be open-sourced; others, like Nokia, have recommended that the OSSii initiative be extended to telco cloud management. Those are debates in their own right, but whatever the route to resolving these critical issues, our view is that vendors must adopt an ‘open API’ strategy and be prepared to iterate those APIs constantly as the standardized interface specifications mature.
In any case, the challenges around NFV mean that the supplier community has little time to waste. Happily, the momentum around NFV demonstrated at TM Forum Live! indicates that vendors were, almost without exception, doing ‘anything and everything’ necessary to make the vertical dimension work, as multiple Catalyst projects showed. There’s no lack of will or innovation; the issue is how best to harness this work into ETSI.