Content for  TS 23.288  Word version:  18.4.0

Dispersion analytics identifies the location (i.e. areas of interest, TAs, cells) or network slice(s) where a UE, or a group of UEs, or any UE disperse most (if not all) of their data volume and sessions transactions (i.e. MM and SM messages). The NWDAF shall be able to provide dispersion statistics or predictions and shall be able to collect UE dispersion related information from NFs. Dispersion is the percentage of activity that a UE, or group of UEs, or any UE, generates at a location or a slice during a period of interest. This enables the operator to rank hot locations or slices by the various activities and to identify the top contributors (i.e. UEs) for that activity. When heavy UEs at a particular location or slice are detected/predicted and reported to a policy entity, their QoS attributes, or access to an area of interest or a slice can then be limited.

• Data volume dispersion - The percentage of data traffic volume that a UE, or a group of UEs, or any UE, generated at a location or in a slice during the period of interest.
• Transactions dispersion - The percentage of MM and SM messages that a UE, or a group of UEs, or any UE, generated at a location or in a slice during the period of interest.

The operator may classify and assign one of the three mobility classes per dispersion characteristics for a UE or a group of UEs, or any UE as fixed, camper or traveller. In addition, the operator may classify a UE as a Top-Heavy UE in case its dispersion percentile rating, at a location or a slice, is higher than a threshold value defined by the operator (e.g. 0.9 percentile), see percentile ranking in the output Tables, clause 6.10.3. The classification is based on either Data volume dispersion or Transactions dispersion.

• Data-Classification - fixed, camper, traveller data-classification per thresholds assigned by the operator.
  EXAMPLE:
  When a UE disperses, during the period of observation, most (threshold=95%) of its data at a location or a slice, the dispersion data-classification of the UE at that location or slice is "fixed".
• Transaction-Classification - fixed, camper, traveller transactions-classifications per thresholds assigned by the operator.
  EXAMPLE:
  When a UE disperses, during the period of observation, (threshold=40%) of its session transactions at a location or a slice, the dispersion transaction-classification of the UE is camper at that location or slice.

Based on the dispersion analytics, the analytics consumer can determine that a data volume or transactions hot spot is formed when the volume of data or amount of transactions dispersed by most of the UEs at the area of interest or slice exceeds a certain data or transaction threshold established through trend analysis or operator policy. The consumer of dispersion analytics may indicate in its request:

• Analytics ID = "Dispersion Analytics";
• Target of Analytics Reporting: a single UE (SUPI), or a group of UEs (an Internal Group ID), or any UE. "Any UE" is only supported in combination with Analytics Filter Information (S-NSSAI, Area(s) of Interest, Top-Heavy UEs indication, Fixed indication, Camper indication or Traveller indication) and Dispersion Analytics type "Data Volume Dispersion Analytics" and/or "Transaction Dispersion Analytics";
• Analytics Filter Information:
  • optional list of TA(s), Area(s) of Interest, Cells, or S-NSSAI(s), Application ID(s), Top-Heavy UEs, Fixed UEs, Camper UEs, Dispersion Analytics type (Data Volume Dispersion Analytics (DVDA) or Transactions Dispersion Analytics (TDA) or both); and
  • optional list of analytics subsets that are requested (see clause 6.10.3);
• Preferred level of accuracy of the analytics;
• Preferred level of accuracy per analytics subset (see clause 6.10.3);
• Preferred order of results for the list of UE Dispersion Analytics information:
  • ordering criterion:
    • For DDA, one of the following: "time slot start", "data dispersion", "data classification", "data ranking", or "data percentile ranking";
    • For TDA, one of the following: "time slot start", "transaction dispersion", "transaction classification", "transaction ranking", or "transaction percentile ranking"; and
  • order: ascending or descending;
• Analytics target period indicating the time period over which the statistics or predictions are requested;
• Optionally, maximum number of objects and maximum number of SUPIs;
• Optionally, preferred granularity of location information: TA level or cell level;
• Optionally, Spatial granularity size and Temporal granularity size; and
• In a subscription, the Notification Correlation Id and the Notification Target Address are included.

The NWDAF shall be able to collect UE dispersion information from NF(s) and AFs. The information collected by the NWDAF is network data from 5GC NFs and service data from AFs:

• Data related to UE transactions dispersion and bound by location collected from AMF and SMF are defined in the Table 6.10.2-1 and Table 6.10.2-2. This may include individual transactions or SMF/AMF state transition information stored as UE behaviour trends, which provides aggregated transaction information.

Data volume can be collected from the AF per UE across all the UE applications provided by a particular AF or per UE per application when application IDs are requested. Data related to UE data volume dispersion, bound by a location or slice, collected from the AF and UPF are defined in Table 6.10.2-3 through Table 6.10.2-6. If the Target of Analytics Reporting is set to an Internal Group ID and the AF is located in the untrusted domain, NEF translates it into an External Group ID.

• In Table 6.10.2-3, the data volume is collected per UE from the AF. The collected UE information is applicable across all the applications used by the UE during the period.

• In Table 6.10.2-4, the AF reports data volume per UE per application in relation to the start and stop of the application as indicated by the application duration.

Data volume can also be collected from the UPF per UE IP address across all applications or per UE for specific application(s).

• In Table 6.10.2-5, the data volume is collected per UE from the UPF. The collected UE information is applicable across all applications used by the UE between start and stop of the PDU session. The UPF reports volume per UE IP address across all applications.

• In Table 6.10.2-6, the UPF reports data volume per UE for specific application(s) in relation to the start and stop of the application as indicated by the application duration.

There are two modes of data collection:

• Non periodical: A mode where the data volume is requested and consequently provided for the total volume of a PDU session.
• Periodical: A mode where data volume is provided periodically between the start and stop of a PDU session. The period can be specified in the requested analytic target period or configured as a default value in the UPF.

For both modes of data collection, if there are multiple application sessions for the UE PDU session, the NWDAF aggregates (i.e. sums up) the data volume across all applications to obtain per UE information.

The task of the NWDAF is to calculate the volume of data and amount of transactions dispersed by the UE at each visited location during a period of interest. The entry time and exit time to a location is provided by the AMF. Collecting volume per UE and per specific application is optional and such information can be provided by the UPF or AF. The Correlation Information as defined in Table 6.2.4-1 is used to correlate the input data from AMF and SMF, the input data from SMF and UPF and the input data from UPF and AF.

• Slice information related to UE transactions dispersion from NF(s) are defined in Table 6.10.2-7 and Table 6.10.2-8. This may include individual transactions or SMF/AMF state transition information stored as UE behaviour trends, which provides aggregated transaction information.
• The Slice data volume collection is the same as listed in Table 6.10.2-3 to Table 6.10.2-6.

The task of the NWDAF is to calculate the volume of data and amount of transactions dispersed by the UE at each slice during a period of interest. A time stamp of a slice assignment is provided by the AMF. The Correlation Information as defined in Table 6.2.4-1 is used to correlate the input data from AMF and SMF, the input data from SMF and UPF and the input data from UPF and AF. Data collection may be provided on samples (e.g. spatial subsets of UEs or UE group, temporal subsets of UE location information).

The NWDAF can provide Dispersion analytics, in the form of statistics or predictions, to an NF or AF.