Here’s the third (and final) part of our review of Heavy Reading’s recent OSS in the Era of NFV/SDN conference in London. Even if the event raised many more questions than it provided answers, that alone is a pointer to the present state of virtualisation in telco networks.
Here are the remaining takeaways from the event, the final part of a three blog series:
Telcos historically like to take a ‘one size fits all’ approach to network management but this doesn’t work with NFV Service Assurance. For instance, alarming is a problem because faults tend to trigger alarms at both hardware and software levels making resolution time consuming and unclear. Accurate root cause analysis in a dynamic network topology is not straightforward. Self-healing algorithms are badly needed. So are tight integrations with the orchestrator and accurate service impact analysis between layers.
Integration of NFV and legacy OSS monitoring is another problem area, especially in multi-service, multi-vendor environments. Everything should become far more policy-driven, but that means hard-to-manage service changes in mid-lifecycle. Here, the problem isn’t the lack of standards but the fact that existing standards need to be harmonized and converged together. The Holy Grail is moving towards being to react to faults as they happen.
Policy is a major challenge for Closed Loop Service Assurance. Lack of standards and lack of ways to impose policy rules across domain barriers mean any ‘master orchestrator’ can’t effect closed loop actions without inter-domain standardisation. Until that changes, policy will remain an attractive concept but not an ‘implementable’ one.
There’s a lot of talk, but (arguably) there’s paucity of NFV Use Cases of, if not that, a lack of clarity about which or what sort of Use Cases might have the most immediate impact. (Author’s note: I was struck that bottom-line financial impact wasn’t once mentioned in the discussion.) That said, ‘predictive’ Use Cases like Churn Management seem to be among the most popular. There’s a lot of talk about Use Cases driven by AI and Machine Learning (ML). Right now, it’s just that. A lot of talk.
Automation is where the telco community wants to see AI/ML in action. For fairly or semi-static environments this is good, but “event based” environments like a mobile networks (soccer games, youtube releases, “house of cards”, “accidents”), where predicting will be hard means ML will have a harder time due to a lack of data. Eventually it will catch up (as more data is collected).
An interesting point that gained common agreement: In Machine Learning Use Cases, Proofs of Concepts are useless. You have to put the case into the real network to find out if it works. Another obvious thing; There is really very little common agreement about AI or ML in telco, what they are, how they work, what they might be used for.
There were a lot of “bottom-up” vs “top-down” discussions; general agreement that both are needed. One speaker mentioned that ML is used for Frequency Hopping (bottom-up) while to change base station or RAT could be a top-down decision based on policies etc. It’s always a balance between business driven and tech driven interests. The general feeling is that most operators see ML more as enablement for monetization of the OTT services, which may be a more mature way compared to trying to catch the train by bundling or self-built initiatives.
Lastly, a couple of definitions of digitzation and what is a DSP? With regard to the former, it’s “changing your business to become more digital (no kidding.) It’s a modernization of business processes to enable telcos to compete with OTT players.” Telcos like to blame legacy OSS and BSS as barriers to achieving this but according to at least some speakers, the problem is their mindset, not their infrastructures. The fear of large transformation projects continues to predominate and slow the pace of the digital shift.
As I said the beginning, NFV and SDN in telco? Clear as mud. If you want to comment on any of the points above, please e-mail me.
Keith Brody and Ola Billinger