Notes

  1. The interface to Near-RT RIC to update the maxNbrOfConns per cell per slice (NSSAI) is also O1. Beyond Guilin, this will be updated to A1.
  2. Messages in Brown are sent over O1 interface.
  3. It is assumed that the Near RT-RIC shall send the initial setting of the maxNbrOfConns when a new NSSI is created, or when a new NSSAI is associated to the NSSI in a cell. This has impact in RAN-Sim. If this is not desired, then SDN-R should prepare and send this info to the CU, which will have an impact on the yang models, as well as in SDN-R.
  4. The FileReady notifications and the PM data will be sent by each of the CUs, resulting in DFC fetching the file from the CU, etc. The flow diagram shows this flow (Steps 7 - 13) only for 1 CU.
  5. It has to be decided whether a single Control Loop message from Slice Analysis MS shall be for:
    1. All Near-RT RICs (in this case, SDN-R has to separate out the message to be sent to each Near-RT RIC)
    2. One Near-RT RIC (i.e., for all the CUs associated to it)
    3. One CU
  6. Example of logic in Slice Analysis MS in Guilin:
    1. Based on (5), wait for sufficient duration to collect all the relevant inputs.
    2. "Moderate" the increase proposed by ML MS using a simple logic (e.g., don't increase more than 20%). Later on this can be done by checking config policy.
    3. If ML MS does not check on minimum changes, ignore changes (i.e., don't carry out the control loop actions) if the changes are, for example, < 5%.
    4. Check if the NSSAI is available and active.

Intelligent Slicing


Offline Training phase

Training


Obsolete

  • No labels

2 Comments

  1. Hi all, 
    I am trying to reproduce this Intelligent Slicing flow ... 

    I did not find any documentation about ML-MS and it is not clear how to deploy it. Can you provide me with a pointer to the documentation or some guidelines to deploy it?


    I think it is something like the Slice-Analysis ms ... but I have no found any blueprint for it.

    Thanks,

    Aniello

    1. Hi Aniello Paolo Malinconico,

      ML MS is not an ONAP component. We can discuss further in mail if you are still interested to know about this.