A method and devices for estimating energy consumption

Abstract

A method for deriving an energy consumption estimation of a contained network function in a wireless network. The method comprising: collecting energy consumption data of the network function; collecting local usage data of the network function; forming an energy consumption estimation function based on the local usage data and energy consumption data; and calculating using the estimation function, an estimation of the energy consumption of the network function based on further usage data. There is also provided corresponding network entities configured to perform the method, a system comprising the corresponding network entities, and a method of calibrating the estimation function.

Description

[0001] A METHOD AND DEVICES FOR ESTIMATING ENERGY CONSUMPTION

[0002] FIELD OF THE INVENTION

[0003] This invention relates to estimating energy consumption of network functions in a wireless network using an estimation function formed based on locally sourced data. For example, to be used when network functions or equivalent entities are isolated from the network infrastructure.

[0004] BACKGROUND

[0005] With the development of communications networks in the future, it is envisioned that they will have strong usage of energy saving technologies in order to realize green and low carbon foot-print networks. This is due to the fact that mobile systems have recently experienced a drastic increase in the energy demand of the networks. As a result, Energy Consumption (EC) tracking and control in mobile networks is becoming important. In particular, there has been huge interest within 3GPP and many companies involved in the development of mobile systems.

[0006] Currently, mobile networks provide clear access to energy consumption information.

[0007] However, this may have a number of limitations in a communications network of the future. For example, the core network assumes availability of EC information at the NFs. However, there may be limited availability of EC information at a network function (NF). This is because networks generally enforce a separation between NFs and the underlying execution platforms with the NFs running as containers. As a result, an Operations, Administrations, and Management (OAM) entity can only provide aggregated energy consumption data.

[0008] The main standardization document of 3GPP deals with the subject of energy consumption in mobile networks. Several approaches to collecting and sharing energy information have been defined in that document.

[0009] For example, one approach deals with collection of energy related information and exposure. This is implemented by an Energy Efficiency Control Function (EECF). The EECF collects the information that has impact on energy consumption / efficiency from other 5GC NFs and the OAM. The information can be collected from 5GC NFs at the granularity of per UE, per UE per service, per PDU Session, or per QoS level etc. while the information collected from the OAM can be per NF and / or per S-NSSAI. Energy Efficiency (EE): The relation between the useful output and energy / power consumption.

[0010] Existing approaches have also considered exposure of energy consumption information to the application functions, and the possibility of extending this to 3rd party entities. Energy Consumption: The Energy consumed by a device needed to achieve an intended application performance over the period of time. The Energy Consumption is expressed in Watts per hour. Usage of Network Data Analytics Function (NWDAF) analytics for energy saving has also been considered. Where it was proposed to allow more network energy related information to be exposed by the network to the authorized consumers. This was suggested for application layer / third party configurations for network energy saving and efficiency. The NWDAF provides network energy related information analytics.

[0011] Exposure of renewable energy information has also been considered. A NWDAF may provide the renewable energy ratio analytics (statistics) by one-time reporting or continuous reporting to a consumer.

[0012] Energy information extraction from the UPF has also been considered. The main idea is to reuse and enhance the existing UPF exposure mechanism, which is already used by Network Data Analytics Functions (NWDAF), to collect the data volume and / or energy consumption data for the services from UPF.

[0013] However, the existing approaches have a number of drawbacks. Firstly, there is an assumption that energy consumption information will be available locally at the NF, although this is not always the case. For example, an NF is a VM or Docker container and as an isolated entity has no access to energy information from the underlying infrastructure. Secondly, there is no method defined that can calculate the per NF or per flow energy consumption using only local NF usage information. Thirdly, the OAM EC data might be expensive or sporadic and contains only aggregate information.

[0014] SUMMARY OF THE INVENTION

[0015] According to one aspect there is provided a method for deriving an energy consumption estimation of a contained network function in a wireless network, the method comprising: collecting energy consumption data of the network function; collecting local usage data of the network function; forming an energy consumption estimation function based on the local usage data and energy consumption data; and calculating using the estimation function, an estimation of the energy consumption of the network function based on further usage data. The proposed method provides a low overhead and local data based energy consumption estimation. For example, the low overhead may be achieved by limiting the time over which the data collection takes place by using timestamps.

[0016] In an embodiment collecting energy consumption data comprises: receiving instructions at an Operations, Administrations, and Management, OAM, entity to collect energy consumption data from a first timestamp to a second timestamp; retrieving the energy consumption data by the OAM from a network element; and sending the energy consumption data to an Energy Consumption Management Function, ECMF. This allows for a low overhead and local data based energy consumption estimation. For example, the low overhead may be achieved by limiting the time over which the data collection takes place by using timestamps. In an embodiment, retrieving energy consumption data comprises receiving an energy consumption measurement from a further entity comprising any one of a virtualisation infrastructure entity, a physical infrastructure entity, or another management plane network function. This enables a flexible energy consumption estimation method which can be applied to various network types. For example, this may include physical devices that can accurately measure energy consumption as well as software devices that can measure consumption of virtualised resources.

[0017] In an embodiment, collecting usage data comprises: instructing the network function to collect resource consumption information by tracking all requests and the requests’ characteristics; collecting the resource consumption information by the network function; and sending the resource consumption information to the Energy Consumption Management Function, ECMF. This allows for the energy consumption estimations to be based on local data. For example, this may reduce the amount of information that has to be exchanged within the network. Further, this may enable a calculation without violating any data isolation rules that can be imposed by the actual software implementation or regulation.

[0018] In an embodiment, forming the estimation function comprises: receiving at the ECMF energy consumption data and resource consumption data; calculating estimation factors by the ECMF based on the received data; and forming the estimation function using the calculated estimation factors. This enables a calibration of the energy estimation function which is based on real data. This allows for an accurate method to derive the EC estimation based on resource consumption data even if such input data has never been seen before.

[0019] In an embodiment, calculating the estimation of the energy consumption comprises: receiving at the network function a request for energy consumption information; collecting resource consumption information for the network function; and calculating the estimation of the energy consumption of the network function using the estimation function based on the collected resource consumption information. This enables a calculation of the energy estimation function which is based on real data. This allows for an accurate method to derive the EC estimation based on resource consumption data even if such input data has never been seen before.

[0020] In an embodiment, the energy consumption information comprises information corresponding to any combination of processing IDs including PDU session, user equipment identifier, flow identifier, service identifier, or all traffic of the network function. This enables the energy consumption estimation to be calculated for various specific network function processes.

[0021] In an embodiment, the request for energy consumption information comes from a further network function and the estimated energy consumption is sent to the further network function. This allows for the energy consumption estimation to be generated locally based only on local data. In an embodiment, the estimation function is a linear function. In an embodiment, the estimation function is a polynomial function. In an embodiment, the estimation function is a neural network model. In an embodiment, the estimation function is a configuration policy in a PCF of the network. This allows for the estimation function to be flexible to different requirements of accuracy and data patterns that work best with different estimation function types.

[0022] In an embodiment, the resource consumption information comprises information corresponding to any combination of processing IDs including PDU session, user equipment identifier, flow identifier, service identifier, or all traffic of the network function. This enables the energy consumption estimation to be calculated based on different mobile network granularities. This enables the tracking of energy consumption of the network for various aspects which may be used in different application scenarios.

[0023] In an embodiment, the resource consumption information comprises any combination of request characteristics including request type, request length, request duration, local memory access size, response type, or response length. This enables the energy consumption estimation to be calculated based on different mobile network granularities. This enables the tracking of energy consumption of the network for various aspects which may be used in different application scenarios.

[0024] In an embodiment, the wireless network is a communication network in the future. In an embodiment, the wireless network is a cloud network with isolation at the cloud control plane. This allows for the estimation function to be used to track energy consumption of communication networks that require strict adherence to specific energy ratings and carbon emissions standards.

[0025] In an embodiment, the network function and ECMF entities are docker containers. In an embodiment, the network function is a UPF and the OAM retrieves energy consumption data for the UPF through API access. This allows forthe estimation function to be used to track energy consumption of communication networks that require strict adherence to specific energy ratings and carbon emissions standards.

[0026] According to another aspect there is provided a network function, NF, the NF configured to: receive a request from an ECMF instructing the network function to collect resource consumption information; collect the resource consumption information of the network function by tracking all requests and the requests’ characteristics; and send the resource consumption information to an Energy Consumption Management Function, ECMF, for use in determining an energy consumption estimation function. The proposed apparatus enables a low overhead for generating the estimation function as well as for the usage based only on local data for the actual energy consumption estimation.

[0027] According to another aspect there is provided a network energy management entity for deriving an energy consumption estimation of a contained network function in a wireless network, the entity configured to: receive a calibration initiation indication; receive energy consumption data of the network function from an Operations, Administrations, and Management, OAM, entity; receive network function usage data from the network function; derive an energy consumption estimation function based on the network function usage data and energy consumption data; and send the estimation function to the network function. The proposed apparatus reduces the overhead for generating the estimation function. Further, it allows for a usage that leverages only local data forthe actual energy consumption estimation.

[0028] According to another aspect there is provided an Operations, Administrations, and Management, OAM, entity for providing calibration measurements for deriving an energy consumption estimation of a contained network function in a wireless network, the OAM configured to: receive a request from an Energy Consumption Management Function, ECMF, to collect energy consumption data for a network function; request said information from a network element level entity of the network; and send the energy consumption data to the ECMF. The proposed apparatus reduces the overhead for generating the estimation function. Further, it allows for a usage that leverages only local data for the actual energy consumption estimation.

[0029] According to another aspect there is provided a method of calibrating an energy consumption estimation function, the calibration comprising deriving a model for mapping network function parameters onto energy consumption of the network function, the method comprising: initiating calibration by an energy consumption management entity of the wireless network; receiving, at the energy management entity, usage data of the network function; receiving, at the energy management entity, energy consumption data of the network function; and forming an energy consumption estimation function based on the received network function usage data and received energy consumption data, whereby the calculated estimated energy consumption is proportionate to the energy consumption data and to the usage data. The proposed apparatus reduces the overhead for generating the estimation function. Further, it allows for a usage that leverages only local data for the actual energy consumption estimation.

[0030] According to another aspect there is provided a system for deriving an energy consumption estimation of a contained network function in a wireless network, the system comprising: a network function configured to collect usage data about itself and send the usage data to a network energy management entity; an Operations, Administrations, and Management, OAM, entity configured to collect energy consumption data of the network function and send it to the network energy management entity; and a network energy management entity configured to receive the usage data and the energy consumption data, form an estimation function for estimating the energy consumption of the network function based on the received data, and based on received resource consumption information estimate the energy consumption of the network function using the estimation function. The proposed apparatus reduces the overhead for generating the estimation function. Further it allows for a usage that leverages only local data for the actual energy consumption estimation.

[0031] BRIEF DESCRIPTION OF THE FIGURES

[0032] The invention will now be described by way of example with reference to the accompanying drawings. In the drawings: Figure 1 shows a schematic of the data flow and processing between entities for implementing the calibration phase.

[0033] Figure 2 shows a schematic of the data flow and processing between entities for implementing the calculation phase.

[0034] Figure 3 shows a schematic illustration of the transfer of data between entities for the energy consumption estimation calculation phase.

[0035] Figure 4 shows a message diagram comprising the data and messages transmitted between entities as part of the proposed method.

[0036] Figure 5 shows an example schematic of the data flow and processing between entities for implementing the calibration phase.

[0037] Figure 6 shows an example of the proposed approach as implemented in a cloud system with docker containers.

[0038] Figure 7 shows an example of the proposed approach for calibration as implemented with a UPF.

[0039] Figure 8 shows an example of the proposed approach for calculation as implemented with a UPF.

[0040] DETAILED DESCRIPTION OF THE INVENTION

[0041] The following terms are commonly used in the presently considered field and have the below explained meanings.

[0042] Energy Saving feature: A feature which contributes to decreasing energy consumption as compared to the case when the feature is not implemented.

[0043] Power: The rate at which energy is transmitted. Power is measured in units of Watts.

[0044] Power Consumption: The power consumed by a device needed to achieve an intended application performance. The power consumption is expressed in Watts.

[0045] Mobile Network Energy Efficiency: Energy Efficiency of a Mobile Network.

[0046] The communication networks of the future do not currently plan to support a method for calibration and calculation of NF energy consumption based on NF usage information.

[0047] There are proposed herein architectures and interfaces for Calculating EC using a pre-calibrated local algorithm. Specifically, there are described two phases comprising a calibration process followed by a calculation process.

[0048] The calibration process’s target is to derive a EC estimation function, EC_NF, in the form:

[0049] EC_NF = F (NF_data), where F is the estimation function and NF_data is the locally available NF data.

[0050] F = wO + w1 * Featurel + w2 * Feature2 + ... + w(n) * Feature(n) may describe the structure of the function F in terms of features and associated weights. F is initially unknown and is derived during the calibration phase using real EC information from the CAM. The ECMF can derive F with different function options as selected. For example, the function may be a linear regression, polynomial regression, or neural network.

[0051] The technical details of the proposed approach will now be described.

[0052] First it is necessary to determine the estimation function using the proposed EC Calibration method. Figure 1 shows a schematic of the data flow and processing between entities for implementing the calibration method. The goal during the calibration phase is to derive a model that maps the local NF parameters to an accurate EC estimation of the NF.

[0053] The calibration is to be first be initialised. Figure 1 shows a first step 1a where the ECMF 102 informs the CAM 104 to collect energy consumption information. This may be collected from specific timestamps also provided to the ECMF 102, i.e. from timestamp 1 to timestamp 2. Enough notice is given for the ECMF 102 to start the collection before the start of this time period. At step 1 b the ECMF 102 informs the NF 106 to track all requests and their associated characteristics. Steps 1a and 1 b need not be in this order and may also be consecutive or performed in parallel.

[0054] Next in the calibration is the collection phase. Figure 1 shows step 2 where the NF 106 collects resource consumption data at the NF 106. This may be collected for a specified duration. At step 3a the NF 106 sends the collected resource consumption information to the ECMF 102. The next step 3b* includes the CAM 104 retrieving network element (NE) energy information. That is, the CAM may receive real EC information from the underlying infrastructure 108 at the network element level. The real EC energy information may be either aggregate energy information or single measurements for a specific NF. Step 3b comprises the CAM 104 sending received aggregate or single measurement NE EC information to the ECMF 102.

[0055] The next part of the calibration process is to calculate the estimation function. At step 4 the ECMF 102 calculates the EC estimation factors for all metrics and features and forms the estimation function.

[0056] The estimation function may take any one of multiple different forms. The estimation function may comprise a linear or polynomial function. The estimation function may comprise a neural network. The estimation function may be a configuration policy in a Policy Control Function (PCF) depending on the accuracy requirements for the EC calculation.

[0057] After the calibration phase the NF has the weighting factors from the ECMF that can be used to calculate an estimated EC during runtime. That is, once an estimation function is determined it can be used at a later time for directly estimating energy consumption of the network function when provided with only usage information available locally to the network function. The estimation function can be represented by the function itself or the form and weighting factors thereof. When a request is executed, the NF uses the estimation function (which it can get from the ECMF) to calculate the energy consumption for this specific request and send the EC along with the response to the requestor.

[0058] It should be understood that the calculation method is an implementation of the estimation function determined during the estimation method. That is, the calculation method may be performed using an already formed estimation function determined at a previous time in a calibration phase. However, the calibration method may be executed as part of an overall method including both the described calibration and calculation phases.

[0059] Figure 2 shows a schematic of the data flow and processing between entities for implementing the calculation method. The estimation function formed at the ECMF 102 can be provided to a network function for implementing. A further network function may then query the network function for its energy consumption estimate and may be referred to as a consumer network function 202.

[0060] At a first step 1 the consumer NF 202 can request EC information from the NF 202. The energy consumption information may be provided as a single measurement or a series of measurements over a specified duration.

[0061] At step 2, the NF 106 may then collect the local data required to calculate the EC using the estimation formula. At step 3, the NF 106 then performs the calculation of the energy consumption estimation using the collected data and weighting factors provided by the ECMF during the calibration phase. At step 4, the NF 106 sends the data downstream, to the consumer NF 202.

[0062] Therefore, there is provided a method for deriving an energy consumption estimation of a contained network function in a wireless network. The method comprises collecting energy consumption data of the network function and collecting local usage data of the network function. The method then continues by forming an energy consumption estimation function based on the local usage data and energy consumption data. Finally, the method includes calculating, using the estimation function, an estimation of the energy consumption of the network function based on further usage data. The further usage data may be local usage data of the same or a different network function. The method may cause the calculating to be performed at the network function using the estimation function. The calculation may be based on features and weights provided to the network function by an ECMF.

[0063] By referring to the usage data as local usage data it is meant that the usage data is collected locally at the network function. By referring to the network function as contained it is meant that the network function has limited connectivity to the underlying infrastructure. For example, the network function’s processes and actions may be isolated from the network infrastructure as a subnetwork and only the inputs and outputs of the NF may be known to the network. Further, the NF cannot directly get real energy consumption information from the infrastructure in contrast to the CAM. Typically, network isolation involves dividing a network into separate segments or subnets. This is also known as network segmentation. Network isolation can help to ensure reliable network performance and better manage digital infrastructure.

[0064] The process of collecting energy consumption data comprises receiving instructions at an Operations, Administrations, and Management, OAM, entity to collect energy consumption data from a first timestamp to a second timestamp. The process then continues by retrieving the energy consumption data by the OAM from a network element level. The energy consumption data may then be sent to an Energy Consumption Management Function, ECMF.

[0065] The process of retrieving energy consumption data may comprise receiving an energy consumption measurement from a further entity comprising any one of a virtualisation infrastructure entity, a physical infrastructure entity, or another management plan network function.

[0066] The process of calculating the estimation of the energy consumption may comprise receiving at the network function a request for energy consumption information, collecting resource consumption information for the network function; and calculating the estimation of the energy consumption of the network function using the estimation function based on the collected resource consumption information. The energy consumption information may comprise information corresponding to any combination of processing IDs including PDU session, user equipment identifier, flow identifier, service identifier, or all traffic of the network function. The request for energy consumption information may come from a further network function and the estimated energy consumption may be sent to the further network function.

[0067] The main benefit of the above-described proposed process is that there is no interaction with the OAM 104 required during implementation of the estimation function. This avoids a long path of interaction between the control and management planes when estimating the energy consumption. The data is locally sourced and used and therefore suitable for local network and NPN entities. This is especially important in highly private environments. Further, this is important in situations where there is a complete isolation of the NF from the infrastructure. For example, when using cloud solutions or limited exposure infrastructures. The process has low overhead since the NF can generate the EC estimation directly based on the data usage and send the EC estimation directly with the response, known as piggybacking. This drastically reduces the communication overhead since no other interaction is required to send the EC estimation. Additionally, the local calculation methods are implemented at the NF itself and does not require multiple hop interaction with the OAM and energy meters running at the NE level.

[0068] The NF local data used for EC calculation may comprise the following data. Observations from cloud systems, for example, how energy usage correlates with NF usage. Each NF has access to local information. Tracking at the NF can follow different aspects as instructed by the ECMF. That is, the ECMF can request that the NF track usage data for specific categories of behaviour, requests, and tasks. The NF calculation granularities used for the NF local information can be selected and linked to certain underlying processing IDs. For example, the NF can track requests and responses related to any of a PDU Session, a UE ID, a Flow ID, a Service ID, or all traffic of NF.

[0069] The NF local information used may comprise a list of requests along with all meta data information of each request. The meta data associated with each request may comprise any combination of: request type, request length, request duration, local memory access size, response type, response length, and response type.

[0070] Figure 3 shows a schematic illustration of the transfer of data between entities as part of an example algorithm for the energy consumption estimation calculation phase.

[0071] The network function or network element 302 may deliver energy consumption data to the ECMF 102. For example, the OAM direct measurements of a particular NF. The NF 304 may provide resource consumption data to the ECMF. For example, the NF local information used per service request comprising request type, request length, request duration, local memory access size, response type, response length, and response type. Both energy consumption and resource consumption data may be provided to the ECMF in response to a calibration trigger 306. The ECMF progresses though the corresponding steps, including obtaining OAM EC measurement 308 and MF data usage measurement 310. The ECMF may then derive the estimation function 312. The ECMF may then send the estimation function to the consuming NF 304.

[0072] The proposed method may require new interfaces and new interactions between different entities. From the ECMF to the NF there may be provided an interface used to initialize the calibration procedures between the ECMF and NF. From the NF to the ECMF there may be provided an interface used to retrieve the calibration measurements from the NF. From the ECMF to the OAM there may be provided an interface used to initialize the calibration procedures between the ECMF and OAM. From the OAM to the ECMF there may be provided an interface configured to handle the retrieval of the calibration measurements from the OAM.

[0073] The process of collecting usage data may comprise instructing the network function to collect resource consumption information by tracking all requests and the requests’ characteristics, causing at the network function, collecting of the resource consumption information, and sending the resource consumption information to the Energy Consumption Management Function, ECMF. The resource consumption information may comprise information corresponding to any combination of processing IDs including PDU session, user equipment identifier, flow identifier, service identifier, or all traffic of the network function. The resource consumption information may comprise any combination of request characteristics including request type, request length, request duration, local memory access size, response type, or response length. Figure 4 shows a message diagram comprising the data and messages transmitted between entities as part of the calibration process 402 and calculation process 404 of the proposed method.

[0074] Starting with the Calibration process 402, first the ECMF 102 triggers the calibration. The ECMF 102 then initiates the calibration process at the NF 106 and initiates the calibration process at the CAM 104. The NF 106 collects local data about usage consumption and then sends the NF local data to the ECMF 102. The CAM 104 may then retrieve the NE energy consumption data from the network infrastructure 108. The CAM 104 may then send the CAM energy consumption data to the ECMF 102. Based on the information received at the ECMF, the ECMF 102 may then calculate an estimation function for estimating the energy consumption of the NF 106.

[0075] Thus, there is provided herein a method of calibrating an energy consumption estimation function. The calibration comprises deriving a model for mapping network function parameters onto energy consumption of the network function. The method comprises initiating calibration by an energy consumption management entity of the wireless network; receiving, at the energy management entity, usage data of the network function; receiving, at the energy management entity, energy consumption data of the network function; and forming an energy consumption estimation function based on the received network function usage data and received energy consumption data, whereby the calculated estimated energy consumption is proportionate to the energy consumption data and to the usage data.

[0076] Moving on to the Calculation process 404, first a consuming NF 202 may request energy consumption information from the NF 106. The NF 106 may then retrieve the calibration factors, e.g. from internal storage. The NF 106 may then perform an energy consumption calculation using calibration weights and local information. The NF 106 may then send the energy consumption information to the consuming NF 202.

[0077] Therefore, there is also provided herein a network function configured to operate according to the proposed method. The NF configured to receive a request from an ECMF instructing the network function to collect resource consumption information. The NF then configured to collect the resource consumption information of the network function by tracking all requests and the requests’ characteristics and send the resource consumption information to an Energy Consumption Management Function, ECMF.

[0078] Accordingly, there is also provided a network energy management entity for deriving an energy consumption estimation of a contained network function in a wireless network. The entity is configured to receive a calibration initiation indication, receive energy consumption data of the network function from an CAM entity, receive network function usage data from the network function, derive an energy consumption estimation function based on the network function usage data and energy consumption data, and send the estimation function to the network function. The entity may be configured to send the estimation function in the form of features and weightings of those features. There is also provided an OAM for providing calibration measurements for deriving an energy consumption estimation of a contained network function in a wireless network. The OAM is configured to receive a request from an ECMF to collect energy consumption data for a network function, request said information from a network element level entity of the network, and send the energy consumption data to the ECMF.

[0079] The entities described above therefore combine to form a system configured to perform the proposed approach. Therefore, there is provided a system for deriving an energy consumption estimation of a contained network function in a wireless network. The system comprises the following network entities or equivalents. A network function configured to collect usage data about itself and send the usage data to a network energy management entity. An Operations, Administrations, and Management, OAM, entity configured to collect energy consumption data of the network function and send it to the network energy management entity. A network energy management entity configured to receive the usage data and the energy consumption data, form an estimation function for estimating the energy consumption of the network function based on the received data, and based on received resource consumption information estimate the energy consumption of the network function using the estimation function.

[0080] The proposed method of determining an estimated energy consumption for a network function may be implemented in a plurality of systems comprising a plurality of types of network entities.

[0081] In a first example embodiment, the EC Calibration and Calculation method proposed herein is implemented in a communication network or system in the future. This embodiment is shown in figure 5, which is similar to the schematic shown in figure 1 .

[0082] The method may be applied to a core network of the communication network in the future. The aim, as described above, is to enable estimation methods without relying on the OAMs continuous data availability. The proposed method supports estimating energy consumption based on just the NF usage data. The functionalities described above as being carried out or located at an ECMF may be executed by or at a Network Data Analytics Function (NWDAF) or a new function defined in the communications network of the future.

[0083] In a second example embodiment, the EC estimation method proposed herein is implemented in a Cloud system. The same proposed approach described above can be applied to a generic cloud system. The proposed approach is particularly useful in cloud systems as they are usually composed of NFs and / or VMs which have strict isolation characteristics. The most prominent execution paradigm is the docker or Kubernetes system. Docker containers usually have no access to the underlying resource consumption due to their isolation by design. In an example implementation of this embodiment, the Kubernetes management plan can be used to collect the underlying energy consumption and use the docker container or virtual machine to calibrate a specific NF for energy estimation following the same methods.

[0084] Figure 6 shows an example of the proposed approach as implemented in a cloud system with docker containers. The same overall approach is used as described above. However, a cloud management entity 602 provides the functions of the OAM as described above and a first docker container 604 and a second docker container 606 provide the functions of the NF and ECMF respectively as described above. Thus, in this embodiment, the wireless network may be a cloud network with isolation at the cloud control plane. The network function and ECMF entities may be docker containers.

[0085] In a third example embodiment, the EC estimation method proposed herein is implemented with a User Plane Function (UPF). This is a useful implementation due to the fact that the UPF is very energy demanding. Figure 7 shows an example of the proposed approach for calibration as implemented with a UPF. It is possible to apply the proposed approach for usage data calculation to the UPF 702. In this embodiment the OAM 104 has API access to get information about EC from the UPF 702. The OAM may collect this information in accordance with the instructions from the ECMF. The UPF 702 may also collect information about underlying resource consumption. The UPF 702 has concrete anchor data comprising the PDU session data transferred over the measurement time period. Therefore, in this embodiment the ECMF or NWDAF 704 can perform the regression on the PDU session data collected and transferred by the UPF 702 and the energy consumption data as reported from the OAM. Thus, the network function may be a UPF and the OAM may retrieve energy consumption data for the UPF through API access.

[0086] Figure 8 shows an example of the proposed approach for calculation as implemented with a UPF 702. After the calibration phase, the UPF 702 has the weighting factors that can be used to calculate EC during runtime. A consumer NF 802 can request EC information from the UPF 702. This may be requested as single measurements or over a duration. The UPF 702 may then collect the data required, e.g. the data transferred for PDU session(s), to calculate the EC. The UPF may then perform the calculation of the estimated EC using the data and weighting factors provided by the ECMF during the calibration phase. The UPF 702 may then send the EC estimation to the NF consumer 802.

[0087] The process of forming the estimation function may comprise, at the ECMF, receiving energy consumption data and resource consumption data, calculating estimation factors based on the received data, and forming the estimation function using the calculated estimation factors. As mentioned above, different mapping functions or estimation functions may be used to implement the proposed approach.

[0088] For example, a linear regression may be used. For a linear regression it is assumed that there is a linear relationship between the input features and the output, i.e. the energy consumption. The weights may then be estimated using methods like Ordinary Least Squares (OLS). The model equation would then be of the form:

[0089] Energy Consumption = wO + w1 * Featurel + w2 * Feature2 + ... + w(n) * Feature(n)

[0090] Such a function provides good interpretation possibilities, as each coefficient represents the change in energy consumption per unit change in the corresponding feature.

[0091] Another example of a mapping function or estimation function is a Polynomial Regression. A polynomial regression extends the linear regression by adding polynomial terms of the input features to capture nonlinear relationships. It enables the capturing of more complex relationships between features and energy consumption. The model equation would then be of the form:

[0092] Energy Consumption = wO + w1 * Featurel + w2 * Feature2 + ... + w(n) * Feature(n) + w_polynomial * (Feature1A2 + Feature2A2 + ...)

[0093] The degree of the polynomial can be adjusted based on the complexity of the mapping between usage data and energy data. In the case where there are more data types and sizes it is necessary to have a larger model to accurately capture the mapping.

[0094] Another example mapping function or estimation function is a Neural Network. Neural networks can be used to capture complex and nonlinear relationships. Features are fed into the input layer, and the network learns to map them to the energy consumption through hidden layers. The output layer gives the estimated energy consumption. Neural networks require a sufficient amount of data for training and may involve hyperparameter tuning to optimize performance.

[0095] The estimation function may be a linear function. The estimation function may be a polynomial function. The estimation function may be a neural network model. The estimation function may be a configuration policy in a PCF of the network.

[0096] The network function is configured to collect local usage data of the network function, wherein collecting local usage data of the network function comprises collecting resource consumption information.

[0097] The network function may be configured to calculate, using the estimation function, an estimation of the energy consumption of the network function based on further usage data. The further usage data may comprise resource consumption information collected at a later runtime. This information may then be mapped according to the estimation function to obtain an estimation of the energy consumption of the network function.

[0098] The network function may be configured to calculate the estimation of the energy consumption, the network function configured to receive a request for energy consumption information, collect resource consumption information, and calculate the estimation of the energy consumption using the estimation function based on the collected resource consumption information.

[0099] The network function may be configured to calculate the estimated energy consumption of a further network function, where the request for energy consumption information comes from the further network function and the estimated energy consumption is sent to the further network function.

[0100] The network function may be configured to collect resource consumption information comprising information corresponding to any combination of processing IDs including PDU session, user equipment identifier, flow identifier, service identifier, or all traffic of the network function.

[0101] The network function may be configured to collect resource consumption information comprising any combination of request characteristics including request type, request length, request duration, local memory access size, response type, or response length.

[0102] The network function may be a docker container. The network container may be a User Plane Function.

[0103] The energy management entity may be configured to instruct a network function to collect resource consumption information. The resource consumption information may be in the form of usage data. The energy management entity may be an Energy Consumption Management Function, ECMF.

[0104] The energy management entity may be configured to receive energy consumption data from an Operations, Administrations, and Management, OAM, entity.

[0105] The energy management entity may be configured to derive the energy consumption estimation function based on the network function usage data and energy consumption data by calculating estimation factors based on the received data and forming the estimation function using the calculated estimation factors.

[0106] The energy management function may be configured to form the estimation function as any one or more of a linear function, a polynomial function, a neural network, and a configuration policy.

[0107] The energy management entity may be configured to receive the energy consumption data comprising information corresponding to any combination of processing IDs including PDU session, user equipment identifier, flow identifier, service identifier, or all traffic of the network function.

[0108] The energy management entity may be configured to receive the resource consumption information comprising any combination of request characteristics including request type, request length, request duration, local memory access size, response type, or response length. The OAM may be configured to retrieve energy consumption data comprising receiving an energy consumption measurement from a further network entity comprising any one of a virtualisation infrastructure entity, a physical infrastructure entity, or another management plan network function.

[0109] The OAM entity may be configured to collect and transmit the energy consumption information comprising information corresponding to any combination of processing IDs including PDU session, user equipment identifier, flow identifier, service identifier, or all traffic of the network function. The OAM may be configured to retrieve energy consumption data for a User Plane Function through API access.

[0110] Thus, there is presented herein a method and associated devices for providing an estimated energy consumption of a network function in a wireless network. With this idea it is possible to estimate NF energy consumption and determine efficiency calculations. It is possible to determine a runtime calculation of energy consumption independent of the OAM. The energy consumption may be estimated per service or per flow request. The estimation function can be an add on module. The add on module may be sold by a vendor.

[0111] The benefits of implementing the proposed approach include that no OAM is required during runtime. The computational overheads may be reduced. Access is provided to local information. Strict isolation properties of certain network components, e.g. containerization, can be respected and is able to be maintained.

[0112] The applicant hereby discloses in isolation each individual feature described herein and any combination of two or more such features, to the extent that such features or combinations are capable of being carried out based on the specification as a whole in the light of the common general knowledge of a person skilled in the art, irrespective of whether such features or combinations of features solve any problems disclosed herein, and without limitation to the scope of the claims. The applicant indicates that aspects of the invention may consist of any such individual feature or combination of features. In view of the foregoing description it will be evident to a person skilled in the art that various modifications may be made within the scope of the invention.

Claims

CLAIMS1. A method for deriving an energy consumption estimation of a contained network function (106) in a wireless network, the method comprising: collecting energy consumption data of the network function; collecting local usage data of the network function; forming an energy consumption estimation function based on the local usage data and energy consumption data; and calculating using the estimation function, an estimation of the energy consumption of the network function based on further usage data.

2. The method of claim 1 , wherein collecting energy consumption data comprises: receiving instructions at an Operations, Administrations, and Management, OAM, entity (104) to collect energy consumption data from a first timestamp to a second timestamp; retrieving the energy consumption data by the OAM from a network element level (108); and sending the energy consumption data to an Energy Consumption Management Function, ECMF (102).

3. The method of claim 1 , wherein retrieving energy consumption data comprises receiving an energy consumption measurement from a further entity (108) comprising any one of a virtualisation infrastructure entity, a physical infrastructure entity, or another management plan network function.

4. The method of any of claims 1 to 3, wherein collecting usage data comprises: instructing the network function to collect resource consumption information by tracking all requests and the requests’ characteristics; collecting the resource consumption information by the network function; and sending the resource consumption information to the Energy Consumption Management Function, ECMF.

5. The method of any preceding claim, wherein forming the estimation function comprises: receiving at the ECMF, energy consumption data and resource consumption information; calculating estimation factors by the ECMF based on the received data; and forming the estimation function using the calculated estimation factors.

6. The method of any preceding claim, wherein calculating the estimation of the energy consumption comprises: receiving at the network function a request for energy consumption information; collecting resource consumption information for the network function; and calculating the estimation of the energy consumption of the network function using the estimation function based on the collected resource consumption information.

7. The method of claim 6, wherein the energy consumption information comprises information corresponding to any combination of processing IDs including PDU session, user equipment identifier, flow identifier, service identifier, or all traffic of the network function.

8. The method of claim 6 or 7, wherein the request for energy consumption information comes from a further network function (202) and the estimated energy consumption is sent to the further network function.

9. The method of any of claims 1 to 8, wherein the estimation function is a configuration policy in a PCF of the network.

10. The method of any preceding claim, wherein the resource consumption information comprises information corresponding to any combination of processing IDs including PDU session, user equipment identifier, flow identifier, service identifier, or all traffic of the network function.11 . The method of any preceding claim, wherein the resource consumption information comprises any combination of request characteristics (304) including request type, request length, request duration, local memory access size, response type, or response length.

12. The method of any preceding claims, wherein the wireless network is a communication network of the future.

13. The method of any of claims 1 to 1 1 , wherein the wireless network is a cloud network with isolation at the cloud control plane.

14. The method of claim 13, wherein the network function (604) and ECMF (606) entities are docker containers.

15. The method of any of claims 2 to 12, wherein the network function is a UPF (702) and the OAM retrieves energy consumption data for the UPF through API access.

16. A network function, NF, (106) the NF configured to: receive a request from an Energy Consumption Management Function, ECMF, (102) instructing the network function to collect resource consumption information; collect the resource consumption information of the network function by tracking all requests and the requests’ characteristics; and send the resource consumption information to the ECMF for use in determining an energy consumption estimation function.

17. A network energy management entity (102) for deriving an energy consumption estimation of a contained network function in a wireless network, the entity configured to: receive a calibration initiation indication; receive energy consumption data of the network function (106) from an Operations, Administrations, and Management, OAM, entity (104); receive network function usage data from the network function; derive an energy consumption estimation function based on the network function usage data and energy consumption data; and send the estimation function to the network function.

18. An Operations, Administrations, and Management, OAM, entity (104) for providing calibration measurements for deriving an energy consumption estimation of a contained network function (106) in a wireless network, the OAM configured to: receive a request from an Energy Consumption Management Function, ECMF, (102) to collect energy consumption data for a network function; request said information from a network element level entity (108) of the network; and send the energy consumption data to the ECMF.

19. A method of calibrating an energy consumption estimation function, the calibration comprising deriving a model for mapping network function parameters onto energy consumption of the network function (106), the method comprising: initiating calibration by an energy consumption management entity (102) of the wireless network; receiving, at the energy management entity, usage data of the network function; receiving, at the energy management entity, energy consumption data of the network function; and forming an energy consumption estimation function based on the received network function usage data and received energy consumption data, whereby the calculated estimated energy consumption is proportionate to the energy consumption data and to the usage data.

20. A system for deriving an energy consumption estimation of a contained network function in a wireless network, the system comprising: a network function (106) configured to collect usage data about itself and send the usage data to a network energy management entity (102); an Operations, Administrations, and Management, OAM, entity (104) configured to collect energy consumption data of the network function and send it to the network energy management entity; and a network energy management entity configured to receive the usage data and the energy consumption data, form an estimation function for estimating the energy consumption of the network19function based on the received data, and based on received resource consumption information estimate the energy consumption of the network function using the estimation function.

Images