6.7 Energy Efficiency (EE) KPI

28.5543GPP5G end to end Key Performance Indicators (KPI)Management and orchestrationRelease 18TS

Tools: ARFCN - Frequency Conversion for 5G NR/LTE/UMTS/GSM

6.7.1 NG-RAN data Energy Efficiency (EE)

6.7.1.1 Definition

a) EE_MN,DV.

b) A KPI that shows mobile network data energy efficiency in operational NG-RAN. Data Volume (DV) divided by Energy Consumption (EC) of the considered network elements. The unit of this KPI is bit/J.

c) EE_MN,DV

– for non-split gNBs;

– for split gNBs;

d) SubNetwork

e) The Data Volume (in kbits) is obtained by measuring amount of DL/UL PDCP SDU bits of the considered network elements over the measurement period. For split-gNBs, the Data Volume is calculated per Interface (F1-U, Xn-U, X2-U). The Energy Consumption (in kWh) is obtained by measuring the PEE.Energy of the considered network elements over the same period of time. The samples are aggregated at the NG-RAN node level. The 3GPP management system responsible for the management of the gNB (single or multiple vendor gNB) shall be able to collect PEE measurements data from all PNFs in the gNB, in the same way as the other PM measurements.

6.7.2 Network slice Energy Efficiency (EE)

6.7.2.1 Generic Network Slice Energy Efficiency (EE) KPI

where:

– ‘Performance of network slice’ (P_ns) is defined per type of network slice;

– ‘Energy Consumption of network slice’ (EC_ns) is defined independently from any type of network slice.

For one unit of EC_ns, the higher P_ns is, the higher the generic network slice EE KPI is, i.e. the more energy efficient the network slice is.

6.7.2.2 Energy efficiency of eMBB network slice

a) EE_eMBB,DV

b) A KPI that shows the energy efficiency of network slices of type eMBB. The P_ns for a network slice of type eMBB is obtained by summing up UL and DL data volumes at N3 interface(s) of the network slice.

, where SNSSAI identifies the S-NSSAI.

This KPI is obtained by the sum of UL and DL data volumes at N3 interface(s) of the network slice, divided by the energy consumption of the network slice. The unit of this KPI is bit/J.

d) NetworkSlice

e) In case of redundant transmission paths over the N3 interface for high reliability communication (cf. TS 23.501 [7] clause 5.33.2), it is expected that the data volume is counted once. In particular:

– In case of Dual Connectivity based end to end Redundant User Plane Paths (cf. TS 23.501 [7] clause 5.33.2.1), in which a UE may set up two redundant PDU Sessions over the 5G network, the Data Volume related to only one PDU session is to be considered;

– In case of redundant transmission with two N3 tunnels between the PSA UPF and a single NG-RAN node (cf. TS 23.501 [7] figure 5.33.2.2-1) which are associated with a single PDU Session, the Data Volume related to only one N3 tunnel is to be considered;

– In case of two N3 and N9 tunnels between NG-RAN and PSA UPF for redundant transmission (cf. TS 23.501 [7] figure 5.33.2.2-2) associated with a single PDU Session, the Data Volume related to only one N3 tunnel is to be considered.

For the measurement of the energy efficiency of the 5G core network, the 3GPP management system in charge of collecting the data volume measurements listed here above shall consider them only once in case of redundant transmission over the N3 interface.

6.7.2.2a Energy efficiency of eMBB network slice – RAN-based

6.7.2.2a.1 Definition

a) EE_{RANonlyeMBB,DV}

b) A KPI that shows the energy efficiency of network slices of type eMBB based on NR measurements. The Pns for a network slice of type eMBB is obtained by summing up UL and DL data volumes at F1-U, Xn-U and X2-U interface(s) of gNBs, on a per S-NSSAI basis.

For non-split gNBs:

, where:

– DRB.PdcpSduVolumeUl.SNSSAI is the Data Volume (amount of PDCP SDU bits) in the uplink delivered to PDCP layer per S-NSSAI – see TS 28.552 [4] clause 5.1.2.1.2.1,

– DRB.PdcpSduVolumeDl.SNSSAI is the Data Volume (amount of PDCP SDU bits) in the downlink delivered to PDCP layer per S-NSSAI – see TS 28.552 [4] clause 5.1.2.1.1.1.

For split gNBs:

, where:

– DRB.F1uPdcpSduVolumeDl.SNSSAI is the number of DL PDCP SDU bits sent to GNB-DU (F1-U interface) per S-NSSAI – see TS 28.552 [4] clause 5.1.3.6.2.3,

– DRB.F1uPdcpSduVolumeUl.SNSSAI is the number of UL PDCP SDU bits entering the GNB-CU-UP from GNB-DU (F1-U interface) per S-NSSAI – see TS 28.552 [4] clause 5.1.3.6.2.4,

– DRB.XnuPdcpSduVolumeDl.SNSSAI is the number of DL PDCP SDU bits sent to external gNB-CU-UP (Xn-U interface) per S-NSSAI – see TS 28.552 [4] clause 5.1.3.6.2.3,

– DRB.XnuPdcpSduVolumeUl.SNSSAI is the number of UL PDCP SDU bits entering the GNB-CU-UP from external gNB-CU-UP (Xn-U interface) per S-NSSAI – see TS 28.552 [4] clause 5.1.3.6.2.4,

– DRB.X2uPdcpSduVolumeDl.SNSSAI is the number of DL PDCP SDU bits sent to external eNB (X2-U interface) per S-NSSAI – see TS 28.552 [4] clause 5.1.3.6.2.3,

– DRB.X2uPdcpSduVolumeUl.SNSSAI is the number of UL PDCP SDU bits entering the GNB-CU-UP from external eNB (X2-U interface) per S-NSSAI – see TS 28.552 [4] clause 5.1.3.6.2.4.

The final Network Slice EE KPI definition, based on Data Volume, for RAN-only eMBB type of network slice, would be defined as follows:

, where EC_RANonlyns is the energy consumption of the RAN-only network slice over the same observation period.

NOTE: Void

d) NetworkSlice

6.7.2.3 Energy efficiency of URLLC network slice

6.7.2.3.1 Introduction

This KPI is defined with two variants.

6.7.2.3.2 Based on latency of the network slice

a) EE_{URLLC,Latency}

b) A KPI that shows the energy efficiency of network slices of type URLLC. The P_ns for a network slice of type URLLC is the inverse of the average end-to-end User Plane (UP) latency of the network slice. In this KPI variant, latency are the only factor considered for evaluating the performance of network slice.

, where ‘Network slice mean latency’ is defined as the average end-to-end User Plane (UP) latency of the network slice, and where the average end-to-end User Plane (UP) latency for one S-NSSAI is defined by:

This KPI is obtained by the inverse of the average end-to-end User Plane (UP) latency of the network slice divided by the energy consumption of the network slice. The unit of this KPI is (0.1ms * J)^-1.

d) NetworkSlice

6.7.2.3.3 Based on both latency and Data Volume (DV) of the network slice

a) EE_{URLLC,DV,Latency}

b) A KPI that shows the energy efficiency of network slices of type URLLC. The P_ns for a network slice of type URLLC is the sum of UL and DL traffic volumes at N3 or N9 interface(s) on a per S-NSSAI basis multiplied by the inverse of the end-to-end User Plane (UP) latency of the network slice. In this KPI variant, data volume and latency are two factors considered for evaluating the performance of network slice. This KPI is applicable for the cases where, for example, the URLLC network slice is deployed and operators want to evaluate the Energy Efficiency of the slice at different periods of time, such as the busy hours in the morning and the idle hours in the mid night, in which both latency performance and the data volume performance can vary.

where

w_N3 and w_N9 are the weight for DV_N3 and DV_N9 respectively. w_N3 and w_N9 can be decided according to the deployment of PSA UPF. For example, in cases where PSA UPF has only N9 tunnels, such as the ones described in TS 23.501[2] clause 5.6.4 and clause 5.33.2.2, w_N3 can be set to 0 and W_N9 can be set to 1, so that only N9 interface is considered. In the cases where PSA UPF has only N3 tunnels, w_N3 can be set to 1 and w_N9 can be set to 0, so that only N3 interface is considered.

This KPI is obtained by the product of the sum of the weighted UL and DL traffic data volumes at N3 interface(s) or N9 interface of the PSA UPF of the network slice multiplied by the inverse of the end-to-end User Plane (UP) latency of the network slice, divided by the energy consumption of the network slice. The unit of this KPI is bit/(0.1ms*J).

d) NetworkSlice

e) In case of redundant transmission paths for high reliability communication (TS 23.501 [7] clause 5.33.2), it is expected that the data volume is counted once. In particular:

– In case of Dual Connectivity based end to end Redundant User Plane Paths ( TS 23.501 [7] clause 5.33.2.1), in which a UE may set up two redundant PDU Sessions over the 5G network, the Data Volume related to only one PDU session is to be considered;

– In case of redundant transmission with two N3 tunnels between the PSA UPF and a single NG-RAN node (TS 23.501 [7] figure 5.33.2.2-1) which are associated with a single PDU Session, the Data Volume related to only one N3 tunnel is to be considered;

– In case of two N3 and N9 tunnels between NG-RAN and PSA UPF for redundant transmission ( TS 23.501 [7] figure 5.33.2.2-2) associated with a single PDU Session, the Data Volume related to only one of the multiple N3/N9 tunnels for redundant transmission connecting to PSA UPF is considered. The main reason for this is that, if the traffic is counted more than once, it will increase artificially the EEURLLC,DV,Latency KPI.

The 3GPP management system in charge of collecting the data volume measurements listed here above shall consider them only once in case of redundant transmission over the N3/N9 interface.

6.7.2.4 Energy efficiency of MIoT network slice

6.7.2.4.1 Based on the number of registered subscribers of the network slice

a) EE_MIoT,RegSubs

b) A KPI that shows the energy efficiency of network slices of type MIoT. In this case, the P_ns for a network slice of type MIoT is the maximum number of subscribers registered to the network slice.

, where SNSSAI identifies the S-NSSAI.

This KPI is obtained by the maximum number of registered subscribers to the network slice divided by the energy consumption of the network slice. The unit of this KPI is user/J.

d) NetworkSlice

6.7.2.4.2 Based on the number of active UEs in the network slice

a) EE_{MIoT,ActiveUEs}

b) A KPI that shows the energy efficiency of network slices of type MIoT. In this case, the P_ns for a network slice of type MIoT is the mean number of active UEs of the network slice.

, where SNSSAI identifies the S-NSSAI.

This KPI is obtained by the mean number of active UEs of the network slice divided by the energy consumption of the network slice. The unit of this KPI is UE/J.

d) NetworkSlice

6.7.3 5G Energy Consumption (EC)

6.7.3.1 NF Energy Consumption (EC)

6.7.3.1.1 Definition

a) EC_NF

b) This KPI describes the Energy Consumption (EC) of a 5G Network Function (NF). This KPI is obtained by summing up the energy consumption of PNF(s) and/or VNF(s) which compose the NF. The unit of this KPI is J.

– How a 5GC NF is composed of VNFs and PNFs is implementation specific. In particular, whether a VNF instance (respectively PNF) is shared or not between more than one NF is implementation specific. Hence, the case where a VNF instance (resp. PNF) is shared between multiple NFs is out of scope of the present document;

– EC_PNF represents the Energy Consumption (EC) of a PNF;

– EC_VNF represents the Energy Consumption (EC) of a VNF. It is obtained by summing up the Energy Consumption (EC) of all its constituent VNFCs;

– In the present document:

# EC_PNF is measured according to ETSI ES 202 336-12 [10],

# it is considered that EC_VNF cannot be measured hence is estimated. Therefore the resulting EC_NF KPI is defined as:

6.7.3.1.2 Estimated Virtualized Network Function (VNF) energy consumption

a) EC_{VNF,estimated}

b) A KPI that gives an estimation of the energy consumption of a VNF. This KPI is obtained by summing up the estimated energy consumption of its constituent Virtualized Network Function Components (VNFC). The unit of this KPI is J.

d) ManagedFunction

e) In this version of the document, the energy consumption of the VNFC is estimated as per clause 6.7.3.1.3.

6.7.3.1.3 Estimated Virtualized Network Function Component (VNFC) energy consumption

a) EC_{VNFC,estimated}

b) A KPI that gives an estimation of the energy consumption of a VNFC. In this version of the document, this KPI is obtained by taking the estimated energy consumption of the virtual compute resource instance on which the VNFC runs. The unit of this KPI is J.

d) ManagedFunction

e) In this version of the document, the energy consumption of the virtual compute resource instance is estimated based on its mean vCPU usage, as per clause 6.7.3.1.4. The method for calculating EC_{VNFC,estimated},estimated is described in TS 28.310 [9] clause 6.3.2.2.1.

6.7.3.1.4 Estimated virtual compute resource instance energy consumption based on mean vCPU usage

a) EC_{virtualCompute,estimated,VCpuUsageMean}

b) A KPI that gives an estimation of the energy consumption of a virtual compute resource instance. The energy consumption of a virtual compute resource instance X is estimated as a proportion of the energy consumption of the NFVI node on which the virtual compute resource instance X runs. This proportion is obtained by dividing the vCPU mean usage of the virtual compute resource instance X, by the sum of the vCPU mean usage of all virtual compute resource instances running on the same NFVI Node as X. The unit of this KPI is J.

d) ManagedFunction

– VCpuUsageMean is the mean vCPU usage of the virtual compute resource instance during the observation period, provided by ETSI NFV MANO (see clause 7.1.2 of ETSI GS NFV-IFA 027 [11]),

– is sum of the vCPU mean usage of all virtual compute resource instances running on the same NFVI Node during the same observation period, all separately provided by NFV MANO (see clause 7.1.2 of ETSI GS NFV-IFA 027 [11]),

– EC_{NFVINode,measured} is the energy consumption of the NFVI node on which the virtual compute resource runs, measured during the same observation period, as per ETSI ES 202 336-12 [10].

6.7.3.2 5GC Energy Consumption (EC)

6.7.3.2.1 Definition

a) EC_5GC

b) This KPI describes the Energy Consumption (EC) of the 5G Core Network (CN). It is obtained by summing up the Energy Consumption of all the Network Functions (EC_NF) that compose the 5G core network. For the Energy Consumption (EC) of Network Functions, see clause 6.7.3.1. The unit of this KPI is J.

d) Subnetwork

6.7.3.3 Network Slice Energy Consumption (EC)

a) EC_ns

b) This KPI describes the Energy Consumption (EC) of the network slice. It is obtained by summing up the Energy Consumption of all the Network Functions (EC_NF) that compose the network slice. The unit of this KPI is J.

As a network slice may be composed of a RAN network slice subnet, a Transport Network (TN) network slice subnet and a 5GC network slice subnet, they all participate to the energy consumption of the network slice. However, the definition and way to measure the energy consumption of the TN segment is not in the scope of the present document.

The definition of EC_ns based on the following principles:

– For all gNBs in the network slice, clause 5.1.1.19.3 (PNF Energy consumption) of TS 28.552 [6] applies. This measurement is obtained according to the method defined in ETSI ES 202 336-12 [10] – clauses 4.4.3.1, 4.4.3.4, Annex A;

– In case a 5GC NF is composed of Virtualized Network Functions (VNF) and/or Physical Network Functions (PNF), clause 6.7.3.1 of this document defines the NF Energy Consumption (EC);

– In case a NF is dedicated to a network slice, the energy consumption of the NF is entirely attributable to the network slice;

– In case a NF is shared between multiple network slices, the participation of the NF to the energy consumption of the network slice has to be estimated, as it can’t be measured:

– In case of a gNB shared between multiple network slices, the energy consumption attributable to each network slice is estimated as a proportion of the total gNB energy consumption, where the proportion is calculated as the data volume of the network slice relatively to the total data volume carried by the gNB,

– In case of a AMF shared between multiple network slices, the energy consumption attributable to each network slice is estimated as a proportion of the total estimated AMF energy consumption, where the proportion is calculated as the mean number of registered subscribers of the network slice relatively to the overall mean number of registered subscribers of the AMF during the same time period (see TS 28.552 [6] clause 5.2.1.1 for the definition of the mean number of registered subscribers),

– In case of a SMF shared between multiple network slices, the energy consumption attributable to each network slice is estimated as a proportion of the total estimated SMF energy consumption, where the proportion is calculated as the mean number of PDU sessions of the network slice relatively to the overall mean number of PDU sessions of the SMF during the same time period (see TS 28.552 [6] clause 5.3.1.1 for the definition of the mean number of PDU sessions),

– In case of a UPF shared between multiple slices, the energy consumption attributable to each network slice is estimated as a proportion of the total estimated UPF energy consumption, where the proportion is calculated as the data volume of the network slice relatively to the overall data volume of the UPF during the same time period.

– In case of a UPF with N3 interface(s), the data volume of the UPF is obtained by summing up, for all N3 interface(s), the number of octets of incoming GTP data packets on the N3 interface, from (R)AN to UPF (see TS 28.552 [6] clause 5.4.1.3) and the number of octets of outgoing GTP data packets on the N3 interface, from UPF to (R)AN (see TS 28.552 [6] clause 5.4.1.4)

– In case of a PSA UPF with no N3 interface(s), the data volume of the UPF is obtained by summing up, for all N9 interface(s), the number of octets of incoming GTP data packets on the N9 interface for PSA UPF (see TS 28.552 [6] clause 5.4.4.2.3) and the number of octets of outgoing GTP data packets on the N9 interface for PSA UPF (see TS 28.552 [6] clause 5.4.4.2.4)

– The case of other 5GC NFs shared between network slices is not addressed in the present document.

d) NetworkSlice

6.7.3.4 NG-RAN Energy Consumption (EC)

6.7.3.4.1 NG-RAN EC

a) EC_NG-RAN

b) This KPI describes the Energy Consumption (EC) of the NG-RAN. It is obtained by summing up the Energy Consumption of all the gNBs that constitute the NG-RAN. The unit of this KPI is J.

d) Subnetwork

6.7.3.4.2 gNB EC

a) EC_gNB

b) This KPI describes the Energy Consumption (EC) of the gNB. It is obtained by summing up the Energy Consumption of all the Network Functions (NF) that constitute the gNB. For the Energy Consumption of Network Functions (EC_NF), see clause 6.7.3.1. The unit of this KPI is J.

d) ManagedElement

6.7.4 5GC Energy Efficiency (EE)

6.7.4.1 Generic 5GC Energy Efficiency (EE)KPI

, where:

– ‘Useful Output of 5GC’ (UsefulOutput_5GC) is the useful output of 5GC. It can be defined differently, depending on which 5GC network functions are considered;

– ‘Energy Consumption of 5GC’ (EC_5GC) is the Energy Consumption of 5GC.

For one unit of EC_5GC, the higher UsefulOutput_5GC is, the higher the generic 5GC EE KPI is, i.e. the more energy efficient the 5GC is.

6.7.4.2 Energy Efficiency of 5GC based on the useful output of 5GC user plane

a) EE_5GC,UO,UP,DV

b) A KPI that shows the energy efficiency of 5GC. This KPI is based on the useful output of 5GC user plane. The useful output of the 5GC user plane is obtained by summing up UL and DL data volumes at N3 interface(s).

This KPI is obtained by the sum of UL and DL data volumes at N3 interface(s), divided by the energy consumption of 5GC. The unit of this KPI is bit/J.

d) SubNetwork

– In case of Dual Connectivity based end to end Redundant User Plane Paths (see TS 23.501 [7] clause 5.33.2.1), in which a UE may set up two redundant PDU Sessions over the 5G network, the Data Volume related to only one PDU session is to be considered;

– In case of two N3 and N9 tunnels between NG-RAN and PSA UPF for redundant transmission (see TS 23.501 [7] figure 5.33.2.2-2) associated with a single PDU Session, the Data Volume related to only one N3 tunnel is to be considered.