Scope
Since 3G, 3GPP has been standardising mobile networking standards, enabling true world-wide interoperability. Currently, 3GPP is further developing the 5G standard—“beyond 5G” (B5G)—and starting to develop the 6G standard. Energy efficiency is increasingly becoming a topic within 3GPP. This summary discusses energy efficiency-related specifications from the (recently) completed release 18 and before.
Summary
This section provides an overview of five R18 (and before) documents that (partially) relate to energy efficiency.
TS 28.310 Energy Efficiency of 5G (releases 16-18)
Describes use cases, requirements, and solutions for energy efficiency.
These relate to collecting data volume and PEE measurements (see TS 28.552), and to orchestration to improve energy efficiency.
With regards to orchestration, capacity booster cells can be switched off if their capacity is not needed (and the candidate cells taking over may need to increase their coverage area) and edge User Plane Functions (UPFs) can be switched off if no low latency services are required in an area.
Relevance: energy measurements, orchestration
TS 28.552 5G Performance Measurements (releases 15-18)
Describes many performance measurements for RAN and 5GC, dividing them into “families”.
One of these families concerns the so-called Power, Energy and Environmental (PEE) measurements. Section 5.1.1.19 describes the PEE measurements with regards to gNodeB (gNB). The document does not discuss PEE measurements regarding 5GC.
These, Physical Network Function (PNF) related, measurements are energy, power, temperature, voltage, current, and humidity, of which energy is most relevant (used in TS 28.554). They are obtained according to ETSI ES 202 336-12.
Note that the other measurement families may also be relevant when attributing energy consumption to a particular user, session or slice.
Relevance: energy measurements
TS 28.554 5G End to End KPI (releases 15-18)
Describes key performance indicators (KPI) based on TS 28.552 performance measurements.
In particular, the section 6.7 describes the Energy Efficiency KPI for Next Generation Radio Access Network (NG-RAN), slices and 5G Core (5GC), and the Energy Consumption KPI for NF, 5GC and NG-RAN. In case of virtualised resources (for slicing or otherwise), also the way of attributing hardware energy consumption is described.
Section 6.7 of TS 28.554 provides detailed KPIs (metrics) for Energy Efficiency and Energy Consumption. In particular, the following KPIs have been defined:
- NG-RAN data Energy Efficiency: the data volume divided by energy consumption of the considered network elements. The unit of this KPI is bit/J. The KPI relates to a certain period of time, which is not further specified. The calculation is detailed for both non-split and split gNBs.
- Network slice Energy Efficiency: generically defined as the performance of the network slice divided by its energy consumption. The performance of the network slice is defined differently depending on the type of the slice:
- eMBB slice performance: the total data volume transported, that is the sum of UL and DL over the N3 interface for the S-NSSAI of the slice. An alternative, RAN-based, metric sums up UL and DL over F1-U, Xn-U and X2-U instead.
- uRLLC slice performance: the reciprocal of the network slice mean latency i.e., 1 / (UL delay + DL delay). An alternative, more complex, metric factoring in both data volume and latency.
- mIoT slice performance: the maximum number of registered subscribers of the network slice or, alternatively, the number of active UEs in the network slice.
- NF Energy Consumption: accumulated energy consumption of the PNFs and VNFs that compose the NF, assuming that the individual PNFs and VNFs are not shared between more than one NF. PNF energy consumption can simply be measured (it is hardware), but VNF energy consumption needs to be estimated i.e., its share of the energy consumption of the underlying hardware needs to be attributed to it. The latter is done by breaking down to individual VNFCs and then to the virtual compute resources that these VNFCs run on. The VNFC energy consumption is attributed by computing the ratio of either mean vCPU usage or mean vMemory usage, or mean vDisk usage or I/O traffic volume divided by the total amount of such.
- gNB: fraction of data volume.
- AMF: fraction of mean number of registered subscribers.
- SMF: fraction of mean number of PDU sessions.
- UPF: fraction of data volume (counting octets of GTP data packets for DL+UL over N3 or N9).
- The share of other NFs shared between network slides is not yet specified.
NG-RAN Energy Consumption: accumulated energy consumption of all gNBs that constitute the RAN. For each gNB, the energy consumption of all its constituent NFs is accumulated.
5GC Energy Efficiency: generically defined as the useful output of the 5GC divided by its energy consumption (as defined by “5G Energy Consumption” above). Specifically, the useful output if defined at user place level by summing UL and DL over the N3 interface.
Relevance: energy metrics, energy measurements
TS 28.541 5G Network Resource Model (releases 15-18)
Describes many data types and attributes. Among these, Sections 6.3.30 and 6.4.1 describe “energyEfficiency” which relates to enhanced Mobile Broadband (eMBB), (ultra-Reliable Low Latency Communications (uRLLC) or massive IoT (mIoT) slices, as defined by TS 28.554. It can apply to RAN, core, and/or the entire network.
Relevance: data collection
TS 28.622 Generic Network Resource Model (releases 13-18)
Mentions the energy efficiency attributes of TS 28.541 being part of “management data”.
Relevance: data collection
Relevance for EXIGENCE
The metrics defined in TS 28.554 are very relevant to energy metrics. This specification also describes aspects of attributing energy consumption in case of virtualisation, which is very relevant to energy measurements. TS 28.552 links the underlying hardware measurements to ETSI TS 202 336-12, which is relevant to energy measurements as well. TS 28.310 links data collection to the architecture, which is relevant to data collection and it also describes some orchestration use cases that might be of relevance to orchestration. TS 28.541 and 28.622 are related to the 3GPP data infrastructure and therefore may relate to data collection but are believed to be less relevant for the work in EXIGENCE.
