Back in November, The Enhanced Mobile Broadband Group in CW (Cambridge Wireless) held an event on 'Is automation essential in 5G?'. There were some thought provoking presentations and discussions but the one that stood out for me was by Dan Warren from Samsung
I consider that to be quite an achievement. I'll take it as a compliment #cwembb— Dr. Dan Warren (@TMGB) November 28, 2019
The slides are embedded below but I want to highlight these points:
- Some Network Functions will be per slice whereas others will be multi-slice, the split may not be the same for every slice
- Two slices that have the same 'per slice vs multi-slice' functional split may be different network hardware topologies
- Enterprise customers will likely want a 'service' contract that has to be manifested as multiple slices of different types.
- Physical infrastructure is common to all slices
Again, when you apply Artificial Intelligence (AI) to optimize the network functions, does it do it individually first and then end-to-end and if this is applied across all slices, each of which may have a different functionality, requirement, etc. How would it work in practice?
Agree with pretty much all of this, but there are other points. Key to success of this as a 'full industry' solution is how slicing works when roaming and across interconnect. That is fundamental to SLA support on fully mobile devices and applications (inc. NPN to MNO mobility)— Dr. Dan Warren (@TMGB) February 4, 2020
5G - Peak Complexity from 3G4G
As Dan says in his tweet, "It is hard to implement AI to optimise a point solution without potentially degrading the things around it. Constantly being pushed to a bigger picture view => more data => more complexity"
Let me know what you think.
Related Posts:
No comments:
Post a Comment