Policy and Machine Learning in Virtualized Network: Not if, but When (and how?)

by Keith Brody Nov 07, 2017

Our last blog started our review of Heavy Reading’s OSS in the Era of NFV/SDN conference in London. Here, we continue with our report which, as already mentioned suggests that the state of the two virtualisation technologies in telco presently raises many more questions than it provides answers.

Here are some more takeaways from the event, the second part of a three blog series:

OSS is dead (no one actually said that. I did. But that’s the bottom line.) The term is outdated, particularly in the context of Service Assurance. NFV is a game changer here because it’s so much more complex…the possible points of failure multiply exponentially. Virtualisation will also (eventually) see OSS move into the control plane and become automated. This doesn’t change end customer assurance requirements but it does present an enormous network management challenge.

Domain orchestration (or cross-demain orchestration) is becoming a big deal. An ‘ecosystem’ of domains is emerging and controlling them is going to be a big issue. Some vendors are talking about a ‘God box’ solution. Then again, some people believe in unicorns.  The reality is that different orchestrators will (probably) be needed for different domains but without a common control surface, well, Houston, we have a problem.

There’s a need for real-time analytics to drive Service Assurance. Artificial Intelligence (AI) is everyone’s favourite buzzword but few in the industry (if any) understand what it actually means.  Right now, the real value of data is forensic, not real-time management (according to one telco.)

Telcos managing network IT in virtualised networks is a seriously problematic issue. This is partly cultural-engineering teams are not innovation friendly. Operators need common solutions, desperately need to move away from siloed approaches, and they need vendor offerings that support and reflect this.  OSS and network vendors have to come together as a result but even if/when they do, the lack of standards (see above) will make it difficult to drive operational automation. However, is you can (eventually) automate everything then there’s no need for an OSS (as we know it now.)  With that said, a warning from BT: ‘take it slowly. DON’T automate what you don’t understand.’

Design surface (how to push new, virtualised services into the network) is a new part of OSS. This underlies the reality that the network/IT distinction in the operator’s business has to change.

Operators presently are almost exclusively focused on the near-term challenges of NFV (such as onboarding VNFs) but this problematic. They need to shift to prioritising Service Assurance. However, this is particularly difficult to manage/do in hybrid networks, which are the reality most telcos face today (simple reason: Increased complexity.) Also, the standard telco software licensing model is a barrier to NFV.


Keith Brody & Ola Billinger 



Subscribe to DigitalRoute news and blogs

Topics: OSS, NFV, SDN, OSS Mediation, Telco, Machine Learning, Open Source, NFV/SDN, AI, Artificial Intelligence

Keith Brody

Subscribe to our Blog Notifications

Follow us:

Press Releases

Browse Blogs by Topic

see all