Apparatus, method and computer program

Abstract

Embodiments of the present disclosure relate to an apparatus, a method, and a computer program. There is provided an apparatus comprising: means for receiving, at a first entity, a request from a second entity, the request to offload at least one inference step of a machine learning model for an application to a host node of a network, the network comprising a plurality of host nodes; means for acquiring first information related to the application; means for acquiring second information related to the machine learning model; means for obtaining third information relating to at least one of the plurality of host nodes, wherein the third information comprises at least an indication of a carbon emission footprint associated with the at least one of the plurality of host nodes; means for determining, based on the first information, the second information and the third information, at least one deployment option for offloading at least one inference step on at least one of the plurality of host nodes; and means for providing an indication of the determined at least one deployment option from the first entity to the second entity.

Description

Background Art

[0001] The present application relates to methods, apparatus, systems and computer programs, and in particular but not limited to joint reasoning orchestration for context awareness. Background Art

[0003] A communication system can be viewed as a facility that enables a communication session between two or more entities, such as user terminals, base stations and / or other nodes, by providing a bearer between the various entities involved in the communication path. A communication system can be provided, for example, by a communication network and one or more compatible communication devices. A communication session may include, for example, the communication of data for carrying the communication, such as voice, video, electronic mail (email), text messaging, multimedia and / or content data, etc. Non-limiting examples of the services provided include two-way or multi-way calls, data communications or multimedia services, and access to a data network system (such as the Internet).

[0004] In a wireless communication system, at least a portion of a communication session between at least two stations occurs over a wireless link. Examples of wireless systems include public land mobile networks (PLMNs), satellite-based communication systems, and different wireless local networks, such as wireless local area networks (WLANs). Some wireless systems can be divided into cells and are therefore often referred to as cellular systems.

[0005] A user can access the communication system through an appropriate communication device or terminal. The user's communication device may be referred to as a user equipment (UE) or user device. The communication device is provided with appropriate signal receiving and sending means for implementing communications, for example, implementing access to a communication network or communicating directly with other users. The communication device can access a carrier provided by a site (e.g., a base station of a cell) and send and / or receive communications on the carrier.

[0006] Communication systems and associated equipment typically operate according to a given standard or specification that specifies what the various entities associated with the system are allowed to do and how it should be implemented. Communication protocols and / or parameters that should be used for the connection are also typically defined. An example of a communication system is the Universal Mobile Telecommunications System (UMTS) Terrestrial Radio Access Network (UTRAN) (3G wireless). Other examples of communication systems are the Long Term Evolution (LTE) of the Universal Mobile Telecommunications System (UMTS) radio access technology and the so-called 5G or New Radio (NR) networks. NR is being standardized by the Third Generation Partnership Project (3GPP). Other examples of communication systems include 5G-Advanced (NR Rel-18 and above) and 6G. Summary of the invention

[0007] In a first aspect, a device is provided, comprising: a component for receiving a request from a second entity at a first entity, the request being used to offload at least one inference step of a machine learning model for an application to a host node of a network, the network comprising a plurality of host nodes; a component for obtaining first information related to the application; a component for obtaining second information related to the machine learning model; a component for obtaining third information related to at least one host node of the plurality of host nodes, wherein the third information includes at least an indication of a carbon emission footprint associated with at least one host node of the plurality of host nodes; a component for determining at least one deployment option for offloading at least one inference step on at least one host node of the plurality of host nodes based on the first information, the second information and the third information; and a component for providing an indication of the determined at least one deployment option from the first entity to the second entity.

[0008] The means for obtaining the first information may include means for receiving the first information at the first entity from the second entity.

[0009] The means for obtaining the second information may include means for requesting the second information from the network function and means for receiving the second information from the network function.

[0010] The means for acquiring the third information may include means for requesting the third information from another network function and means for receiving the third information from the another network function.

[0011] The first information may include an indication of at least one inference step of the machine learning model to be offloaded to at least one host node among the plurality of host nodes.

[0012] The first information may include a quality of service indicator.

[0013] The first information may include at least one of time sensitivity information, hardware requirements of at least one reasoning step, and data privacy requirements of at least one reasoning step.

[0014] The request may include first information.

[0015] The third information may also include an indication of a computing capability of at least one of the plurality of host nodes.

[0016] The second information may include at least one of the following: input data size, output data size, computational complexity, and parameter size of at least one inference step of the machine learning model.

[0017] The first entity may include a managed service producer hosted on the network function.

[0018] The second entity may include at least one of: a user device, a management service consumer hosted on a network function, or a machine learning entity.

[0019] The multiple host nodes may include at least one of the following: an edge cloud server, a central cloud server, and a network function server.

[0020] In a second aspect, an apparatus is provided, comprising: a component for providing a request from a second entity to a first entity, the request being used to offload at least one inference step of a machine learning model for an application to a network, the network comprising a plurality of host nodes, and a component for receiving, at the second entity, from the first entity an indication of at least one determined deployment option for offloading the at least one inference step on at least one of the plurality of host nodes, the at least one deployment option being determined by the first entity based on first information related to the application, second information related to the machine learning model, and third information related to at least one of the plurality of host nodes, wherein the third information comprises at least an indication of a carbon emission footprint associated with at least one of the plurality of host nodes.

[0021] The apparatus may include means for providing first information from the second entity to the first entity.

[0022] The first information may include an indication of at least one inference step of the machine learning model to be offloaded to at least one host node among the plurality of host nodes.

[0023] The first information may include a quality of service indicator.

[0024] The first information may include at least one of: time sensitivity information, hardware requirements of at least one reasoning step, and data privacy requirements of at least one reasoning step.

[0025] The request may include first information.

[0026] The third information may also include an indication of a computing capability of at least one of the plurality of host nodes.

[0027] The second information may include at least one of the following: input data size, output data size, computational complexity, and parameter size of at least one inference step of the machine learning model.

[0028] The first entity may include a managed service producer hosted on the network function.

[0029] The second entity may include at least one of: a user device, a management service consumer hosted on a network function, or a machine learning entity.

[0030] The multiple host nodes may include at least one of the following: an edge cloud server, a central cloud server, and a network function server.

[0031] The apparatus may include means for receiving an indication of a plurality of deployment options from a first entity and determining to use one of the deployment options to offload at least one reasoning step to at least one host node.

[0032] In a third aspect, a method is provided, comprising: receiving a request at a first entity from a second entity to offload at least one inference step of a machine learning model for an application to a host node of a network, the network comprising multiple host nodes, obtaining first information related to the application, obtaining second information related to the machine learning model, obtaining third information related to at least one host node of the multiple host nodes, wherein the third information includes at least an indication of a carbon emission footprint associated with at least one host node of the multiple host nodes, determining at least one deployment option for offloading at least one inference step on at least one host node of the multiple host nodes based on the first information, the second information and the third information, and providing an indication of the determined at least one deployment option from the first entity to the second entity.

[0033] Obtaining the first information may include receiving the first information at the first entity from the second entity.

[0034] Obtaining the second information may include requesting the second information from the network function and receiving the second information from the network function.

[0035] Obtaining the third information may include requesting the third information from another network function and receiving the third information from the another network function.

[0036] The first information may include an indication of at least one inference step of the machine learning model to be offloaded to at least one host node among the plurality of host nodes.

[0037] The first information may include a quality of service indicator.

[0038] The first information may include at least one of: time sensitivity information, hardware requirements of at least one reasoning step, and data privacy requirements of at least one reasoning step.

[0039] The request may include first information.

[0040] The third information may also include an indication of a computing capability of at least one of the plurality of host nodes.

[0041] The second information may include at least one of the following: input data size, output data size, computational complexity, and parameter size of at least one inference step of the machine learning model.

[0042] The first entity may include a managed service producer hosted on the network function.

[0043] The second entity may include at least one of: a user device, a management service consumer hosted on a network function, or a machine learning entity.

[0044] The multiple host nodes may include at least one of the following: an edge cloud server, a central cloud server, and a network function server.

[0045] In a fourth aspect, a method is provided, comprising providing a request from a second entity to a first entity, the request being used to offload at least one inference step of a machine learning model for an application to a network comprising a plurality of host nodes, and receiving at the second entity from the first entity an indication of at least one determined deployment option for offloading the at least one inference step on at least one of the plurality of host nodes, the at least one deployment option being determined by the first entity based on first information related to the application, second information related to the machine learning model, and third information related to at least one of the plurality of host nodes, wherein the third information comprises at least an indication of a carbon emission footprint associated with at least one of the plurality of host nodes.

[0046] The method may include providing first information from a second entity to a first entity.

[0047] The first information may include an indication of at least one inference step of the machine learning model to be offloaded to at least one host node among the plurality of host nodes.

[0048] The first information may include a quality of service indicator.

[0049] The first information may include at least one of: time sensitivity information, hardware requirements of at least one reasoning step, and data privacy requirements of at least one reasoning step.

[0050] The request may include first information.

[0051] The third information may also include an indication of a computing capability of at least one of the plurality of host nodes.

[0052] The second information may include at least one of the following: input data size, output data size, computational complexity, and parameter size of at least one inference step of the machine learning model.

[0053] The first entity may include a managed service producer hosted on the network function.

[0054] The second entity may include at least one of: a user device, a management service consumer hosted on a network function, or a machine learning entity.

[0055] The multiple host nodes may include at least one of the following: an edge cloud server, a central cloud server, and a network function server.

[0056] The method may include receiving an indication of a plurality of deployment options from a first entity and determining to use one of the deployment options to offload at least one reasoning step to at least one host node.

[0057] In a fifth aspect, a device is provided, comprising at least one processor and at least one memory, the at least one memory comprising computer program code, wherein the at least one memory and the computer program code are configured to, together with the at least one processor, enable the device to receive a request from a second entity at least at a first entity, the request being used to offload at least one inference step of a machine learning model for an application to a host node of a network, the network comprising multiple host nodes, obtain first information related to the application, obtain second information related to the machine learning model, obtain third information related to at least one host node of the multiple host nodes, wherein the third information includes at least an indication of a carbon emission footprint associated with at least one host node of the multiple host nodes, determine at least one deployment option for offloading at least one inference step on at least one host node of the multiple host nodes based on the first information, the second information and the third information, and provide an indication of the determined at least one deployment option from the first entity to the second entity.

[0058] The apparatus may be caused to receive, at a first entity, first information from a second entity.

[0059] The apparatus may be caused to request the second information from the network function and to receive the second information from the network function.

[0060] The apparatus may be caused to request third information from another network function and to receive third information from the another network function.

[0061] The first information may include an indication of at least one inference step of the machine learning model to be offloaded to at least one host node among the plurality of host nodes.

[0062] The first information may include a quality of service indicator.

[0063] The first information may include at least one of: time sensitivity information, hardware requirements of at least one reasoning step, and data privacy requirements of at least one reasoning step.

[0064] The request may include first information.

[0065] The third information may also include an indication of a computing capability of at least one of the plurality of host nodes.

[0066] The second information may include at least one of the following: input data size, output data size, computational complexity, and parameter size of at least one inference step of the machine learning model.

[0067] The first entity may include a managed service producer hosted on the network function.

[0068] The second entity may include at least one of: a user device, a management service consumer hosted on a network function, or a machine learning entity.

[0069] The multiple host nodes may include at least one of the following: an edge cloud server, a central cloud server, and a network function server.

[0070] In a sixth aspect, there is provided an apparatus comprising at least one processor and at least one memory, the at least one memory comprising computer program code, wherein the at least one memory and the computer program code are configured to, together with the at least one processor, cause the apparatus to at least provide a request from a second entity to a first entity, the request being used to offload at least one inference step of a machine learning model for an application to a network comprising a plurality of host nodes, and receiving at the second entity from the first entity an indication of at least one determined deployment option for offloading the at least one inference step on at least one of the plurality of host nodes, the at least one deployment option being determined by the first entity based on first information related to the application, second information related to the machine learning model, and third information related to at least one of the plurality of host nodes, wherein the third information comprises at least an indication of a carbon emission footprint associated with at least one of the plurality of host nodes.

[0071] The apparatus may be caused to provide first information from the second entity to the first entity.

[0072] The first information may include an indication of at least one inference step of the machine learning model to be offloaded to at least one host node among the plurality of host nodes.

[0073] The first information may include a quality of service indicator.

[0074] The first information may include at least one of: time sensitivity information, hardware requirements of at least one reasoning step, and data privacy requirements of at least one reasoning step.

[0075] The request may include first information.

[0076] The third information may also include an indication of a computing capability of at least one of the plurality of host nodes.

[0077] The second information may include at least one of the following: input data size, output data size, computational complexity, and parameter size of at least one inference step of the machine learning model.

[0078] The first entity may include a managed service producer hosted on the network function.

[0079] The second entity may include at least one of: a user device, a management service consumer hosted on a network function, or a machine learning entity.

[0080] The multiple host nodes may include at least one of the following: an edge cloud server, a central cloud server, and a network function server.

[0081] The apparatus may be caused to receive an indication of a plurality of deployment options from a first entity and determine to use one of the deployment options to offload at least one reasoning step to at least one host node.

[0082] In a seventh aspect, a computer-readable medium is provided, comprising instructions that, when executed by an apparatus, cause the apparatus to perform at least the following operations: receiving a request from a second entity at a first entity to offload at least one inference step of a machine learning model for an application to a host node of a network, the network comprising a plurality of host nodes, obtaining first information related to the application, obtaining second information related to the machine learning model, obtaining third information related to at least one of the plurality of host nodes, wherein the third information includes at least an indication of a carbon emission footprint associated with at least one of the plurality of host nodes, determining at least one deployment option for offloading at least one inference step on at least one of the plurality of host nodes based on the first information, the second information, and the third information, and providing an indication of the determined at least one deployment option from the first entity to the second entity.

[0083] The apparatus may be caused to perform, at the first entity, receiving first information from the second entity.

[0084] The apparatus may be caused to perform requesting second information from the network function and receiving second information from the network function.

[0085] The apparatus may be caused to perform requesting third information from another network function and receiving third information from the another network function.

[0086] The first information may include an indication of at least one inference step of the machine learning model to be offloaded to at least one host node among the plurality of host nodes.

[0087] The first information may include a quality of service indicator.

[0088] The first information may include at least one of: time sensitivity information, hardware requirements of at least one reasoning step, and data privacy requirements of at least one reasoning step.

[0089] The request may include first information.

[0090] The third information may also include an indication of a computing capability of at least one of the plurality of host nodes.

[0091] The second information may include at least one of the following: input data size, output data size, computational complexity, and parameter size of at least one inference step of the machine learning model.

[0092] The first entity may include a managed service producer hosted on the network function.

[0093] The second entity may include at least one of: a user device, a management service consumer hosted on a network function, or a machine learning entity.

[0094] The multiple host nodes may include at least one of the following: an edge cloud server, a central cloud server, and a network function server.

[0095] In an eighth aspect, a computer-readable medium is provided, comprising instructions that, when executed by an apparatus, cause the apparatus to perform at least the following operations: providing a request from a second entity to a first entity, the request being used to offload at least one inference step of a machine learning model for an application to a network comprising a plurality of host nodes, and receiving at the second entity from the first entity an indication of at least one determined deployment option for offloading at least one inference step on at least one of the plurality of host nodes, the at least one deployment option being determined by the first entity based on first information related to the application, second information related to the machine learning model, and third information related to at least one of the plurality of host nodes, wherein the third information comprises at least an indication of a carbon emission footprint associated with at least one of the plurality of host nodes.

[0096] The apparatus may be caused to perform providing the first information from the second entity to the first entity.

[0097] The first information may include an indication of at least one inference step of the machine learning model to be offloaded to at least one host node among the plurality of host nodes.

[0098] The first information may include a quality of service indicator.

[0099] The first information may include at least one of: time sensitivity information, hardware requirements of at least one reasoning step, and data privacy requirements of at least one reasoning step.

[0100] The request may include first information.

[0101] The third information may also include an indication of a computing capability of at least one of the plurality of host nodes.

[0102] The second information may include at least one of the following: input data size, output data size, computational complexity, and parameter size of at least one inference step of the machine learning model.

[0103] The first entity may include a managed service producer hosted on the network function.

[0104] The second entity may include at least one of: a user device, a management service consumer hosted on a network function, or a machine learning entity.

[0105] The multiple host nodes may include at least one of the following: an edge cloud server, a central cloud server, and a network function server.

[0106] The apparatus may be caused to perform receiving an indication of a plurality of deployment options from a first entity and determining to use one of the deployment options to offload at least one reasoning step to at least one host node.

[0107] In a ninth aspect, a non-transitory computer readable medium is provided, comprising program instructions for causing an apparatus to perform a method according to at least the third or fourth aspect.

[0108] In the above, many different embodiments have been described. It should be appreciated that further embodiments may be provided by combining any two or more of the above embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0109] Embodiments will now be described, by way of example only, with reference to the accompanying drawings, in which:

[0110] Figure 1 A schematic diagram of an example 5GS communication system is shown;

[0111] Figure 2 A schematic diagram of an example mobile communication device is shown;

[0112] Figure 3 A schematic diagram of an example control device is shown;

[0113] Figure 4 A schematic diagram of the AI / ML model split is shown;

[0114] Figure 5 A schematic diagram of the AI / ML model split is shown;

[0115] Figure 6 A schematic diagram of the AI / ML model split is shown;

[0116] Figure 7 A flow chart showing a method according to an example embodiment is shown;

[0117] Figure 8 A flow chart showing a method according to an example embodiment is shown;

[0118] Fig. 9 A schematic diagram showing a joint deployment option according to an example is shown;

[0119] Fig.10 shows a signaling flow according to an example embodiment;

[0120] Fig.11 A signaling flow according to an example embodiment is shown. DETAILED DESCRIPTION

[0121] Before explaining the examples in detail, refer to Figure 1 , Figure 2 and Figure 3 Certain general principles of wireless communication systems and mobile communication devices are briefly explained to aid understanding of the technology on which the described examples are based.

[0122] An example of a suitable communication system is the 5G or NR concept. The network architecture in NR may be similar to that of LTE-Advanced. The base stations of the NR system may be referred to as next generation Node Bs (gNBs). Changes in the network architecture may depend on the need to support various wireless technologies and more refined quality of service (QoS), as well as some on-demand requirements for QoS levels such as supporting quality of experience (QoE) for users. In addition, network-aware services and applications, as well as service- and application-aware networks, may bring changes to the architecture. These are related to information-centric network (ICN) and user-centric content delivery network (UC-CDN) approaches. NR may use multiple-input-multiple-output (MIMO) antennas, many more base stations or nodes than LTE (the so-called small cell concept), including macro sites operating in cooperation with smaller sites, and perhaps various wireless technologies for better coverage and enhanced data rates.

[0123] Future networks may utilize network function virtualization (NFV), a network architecture concept that proposes virtualizing network node functions into "building blocks" or entities that can be operatively connected or linked together to provide services. A virtualized network function (VNF) may include one or more virtual machines that use standard or general-purpose servers rather than custom hardware to run computer program code. Cloud computing or data storage may also be utilized. In wireless communications, this may mean that node operations are performed at least in part in a server, host, or node that is operatively coupled to a remote radio head. Node operations will also likely be distributed among multiple servers, nodes, or hosts. It should also be understood that the labor distribution between core network operations and base station operations may be different from that of LTE or even non-existent.

[0124] Figure 1 A schematic representation of a 5G system (5GS) 100 is shown. The 5GS may include a user equipment (UE) 102 (which may also be referred to as a communication device or terminal), a 5G radio access network (5GRAN) 104, a 5G core network (5GCN) 106, one or more internal or external application functions (AFs) 108, and one or more data networks (DNs) 110.

[0125] An example 5G core network (CN) includes functional entities. The 5GCN 106 may include one or more access and mobility management functions (AMFs) 112, one or more session management functions (SMFs) 114, an authentication server function (AUSF) 116, a unified data management (UDM) 118, one or more user plane functions (UPFs) 120, a unified data repository (UDR) 122, and / or a network exposure function (NEF) 124. The UPF is controlled by the SMF (session management function) which receives policies from the PCF (policy control function).

[0126] The CN is connected to the UE via a radio access network (RAN). The 5G RAN may include one or more gNodeB (gNB) distributed unit (DU) functions connected to one or more gNodeB (gNB) centralized unit (CU) functions. The RAN may include one or more access nodes.

[0127] A user plane function (UPF) called a PDU Session Anchor (PSA) may be responsible for forwarding frames back and forth between the DN and the tunnels established over 5G towards the UE(s) exchanging traffic with the DN.

[0128] Now refer to Figure 2 describing in more detail possible mobile communication devices, Figure 2A schematic partial cross-sectional view of a communication device 200 is shown. Such a communication device is generally referred to as a user equipment (UE) or a terminal. Any device capable of sending and receiving wireless signals can provide a suitable mobile communication device. Non-limiting examples include a mobile station (MS) or a mobile device, such as a mobile phone or a so-called "smart phone", a computer provided with a wireless interface card or other wireless interface facilities (such as a USB dongle), a personal data assistant (PDA) or a tablet computer provided with wireless communication capabilities, an IP voice (VoIP) phone, a portable computer, a desktop computer, an image capture terminal device such as a digital camera, a game terminal device, a music storage and playback device, a vehicle-mounted wireless terminal device, a wireless endpoint, a mobile station, a laptop embedded device (LEE), a laptop mounted device (LME), an intelligent device, a wireless client device (CPE), or any combination of these or the like. A mobile communication device can provide, for example, a communication for carrying data such as voice, email (email), text messages, multimedia, etc. Therefore, many services can be supplied and provided to users via the user's communication device. Non-limiting examples of these services include two-way or multi-way calls, data communications or multimedia services, or simply access to a data communications network system such as the Internet. Broadcast or multicast data may also be provided to the user. Non-limiting examples of content include downloads, television and radio programs, videos, advertisements, various alerts, and other information.

[0129] The mobile device is typically provided with at least one data processing entity 201, at least one memory 202 and other possible components 203 for use in software and hardware assisted execution of the tasks it is designed to perform, including controlling access to and communication with access systems and other communication devices. The data processing, storage and other related components may be provided on appropriate circuit boards and / or in chipsets. This feature is represented by reference numeral 204. The user can control the operation of the mobile device through an appropriate user interface (such as a keypad 205, voice commands, a touch screen or touchpad, a combination thereof, etc.). A display 208, a speaker and a microphone may also be provided. In addition, the mobile communication device may include appropriate connectors (wired or wireless) to other devices and / or for connecting external accessories (such as a hands-free device) to it.

[0130] The mobile device 200 may receive signals over the air or wireless interface 207 via appropriate means for receiving, and may transmit signals via appropriate means for transmitting wireless signals. Figure 2 In the mobile device, the transceiver means is schematically represented by block 206. The transceiver means 206 may be provided, for example, by a radio part and associated antenna means. The antenna means may be arranged internally or externally to the mobile device.

[0131] Figure 3 An example of a control device 300 for a communication system is shown, for example, the control device is to be coupled to a site of an access system and / or a site for controlling an access system, the site being such as a RAN node, such as a base station, eNB or gNB, a relay node, or a core network node, such as an MME or a serving gateway (S-GW) or a packet data network gateway (P-GW), or a core network function, such as an AMF / SMF, or a server or a host. The method can be implemented in a single control device or across more than one control device. The control device can be integrated with a node or module of a core network or RAN or located outside a node or module of a core network or RAN. In some embodiments, the base station includes a separate control device unit or module. In other embodiments, the control device can be another network element, such as a radio network controller or a spectrum controller. In some embodiments, each base station can have such a control device and a control device provided in a radio network controller. The control device 300 can be arranged to provide control of communications in a service area of ​​the system. The control device 300 includes at least one memory 301, at least one data processing unit 302, 303 and an input / output interface 304. Via the interface, the control device may be coupled to a receiver and a transmitter of the base station.The receiver and / or the transmitter may be implemented as a radio front end or a remote radio head.

[0132] With the increasing popularity of AI / ML applications and growing concerns about sustainability, two contextual perspectives of AI / ML management in mobile networks are considered below, namely, AI / ML models and sustainability.

[0133] A solution of splitting the AI / ML model has been introduced to reduce the energy consumption on the UE side. For example, the AI / ML model is split into two parts, one part to be deployed on the UE and the other part to be deployed in the operator network (ON). The split is done in such a way that the workload of AI / ML reasoning can be offloaded from the UE to the ON. This can lead to a reduction in energy consumption on the UE side.

[0134] Depend on Figure 4 and Figure 5 Two additional scenarios for AI / ML model splitting for media services are illustrated. Figure 4 and Figure 5 In the scenario shown in FIG, a given AI / ML model is split into two parts: the UE part is to be deployed on the UE, and the ON part is to be deployed in the ON.

[0135] exist Figure 4In the scenario 1 illustrated in FIG, UE 400 performs reasoning up to the partial UE, and then sends intermediate data to the network to terminate the reasoning of the partial ON. After the reasoning for the partial ON in the network (at the edge server 401 or the cloud server 402), the result is fed to the receiver. The receiver can be the original UE 400, another UE in the same / different RAN, or a receiver in the DN. The carbon footprint of each server is represented by the dotted circle.

[0136] exist Figure 5 In scenario 2 illustrated in FIG, UE 500 sends input data of the AI / ML model to the network for reasoning until partial ON (at edge server 501 or cloud server 502), and then sends intermediate data back to the UE for reasoning for partial UE. After reasoning for the partial UE on the UE, the result is available on the UE. The carbon footprint of each server is represented by a dotted circle.

[0137] The sustainability characteristics of a network node may vary due to different energy consumption levels and energy sources. The carbon footprint of a network node is a characteristic that can be used to measure the environmental impact of a node. In the context of AI / ML inference, the carbon footprint indicates the amount of CO2 emissions generated by a network node performing AI / ML inference tasks.

[0138] Let's think of carbon emissions as an indicator of environmental impact. The carbon emissions of a network node vary depending on the type of energy source of the network node, the time of energy consumption, geographical location, etc. Therefore, it may not always be optimal for a single network node to perform all inference tasks offloaded to the network. In addition, data privacy, service latency, and other QoS requirements must also be considered. Therefore, the host node needs to be carefully selected. Finally, due to the offloading of inference tasks, it may be necessary to send intermediate results of inference, which costs additional energy and latency. Therefore, the part of the model to be deployed on the selected host node should be carefully exported.

[0139] In the face of increasing climate change, carbon footprint is becoming increasingly important as a measure of environmental impact. For network operators, considering that more and more AI / ML inference tasks will be offloaded to ON, the overall footprint of ON should be optimized. Therefore, it may be necessary to orchestrate the network's context-aware model inference to minimize the network's overall footprint.

[0140] By leveraging AI / ML model splitting, AI / ML model reasoning tasks can be partially offloaded from the UE to the network node. In this case, the environmental impact (e.g., carbon footprint) can be optimized on the UE side. However, offloading tasks to the network will consume resources / energy and thus increase the network environmental impact (e.g., increase carbon emissions), such as Figure 6As shown in , the energy consumption of inference is offloaded from UE 600 to edge server 601 and cloud server 602. Therefore, the problem of how to optimize the environmental impact / footprint of AI inference in the network part is exposed.

[0141] Faced with the challenge that huge amounts of data generation and large deep learning model inference cannot coexist on a single network node, a method for end-to-end cloud-edge collaborative intelligent deployment of deep learning models (called "auto-splitting") has been proposed to optimize the overall latency by considering the deep learning model structure and edge / cloud device constraints. This approach focuses on optimizing latency rather than sustainability. With the proposed "auto-splitting", the offloaded model is split into two units for deep learning model inference, namely for the edge and the cloud. The cost of sending the feature map output (i.e., intermediate data) from the first device (i.e., edge) to the second device (i.e., cloud) is not considered, which may affect the carbon footprint of the system.

[0142] A proposal to save energy on the network side for XR applications and other similar high-demand applications involves selecting devices that host AI / ML models by considering energy efficiency, and model splitting for a given host device on the network side is an optional feature. In this proposal, the number of split AI / ML models is limited by the number of selected host devices. This proposal aims to optimize energy consumption rather than environmental impact (e.g., carbon footprint). Given these two differences, more carbon emission information and available host resource information can be expected.

[0143] Three model splitting principles have been proposed to reduce the workload of AI / ML model inference in the network and guarantee service time for host devices by considering the model structure, computing power of the host devices, and intermediate data transmission. However, environmental impact is not part of the consideration of these proposals.

[0144] Figure 7 A flow chart of a method according to an example embodiment is shown.The method may be performed at a network function, for example at a MnS producer hosted on the network function.

[0145] In 701, the method includes receiving, at a first entity, a request from a second entity to offload at least one inference step of a machine learning model for an application to a host node of a network, the network comprising a plurality of host nodes.

[0146] At 702 , the method includes obtaining first information related to an application.

[0147] At 703, the method includes obtaining second information related to the machine learning model.

[0148] At 704, the method includes obtaining third information related to at least one host node of the plurality of host nodes, wherein the third information includes at least an indication of a carbon emissions footprint associated with the at least one host node of the plurality of host nodes.

[0149] At 705 , the method includes determining at least one deployment option to offload at least one inference step on at least one host node of the plurality of host nodes based on the first information, the second information, and the third information.

[0150] At 706, the method includes providing an indication of the determined at least one deployment option to the second entity.

[0151] Obtaining the first information may include receiving the first information at the first entity from the second entity.

[0152] Obtaining the second information may include requesting the second information from the network function and receiving the second information from the network function.

[0153] Obtaining the third information may include requesting the third information from another network function and receiving the third information from the another network function.

[0154] Figure 8 A flow chart of a method according to an example embodiment is shown. The method may be performed at a UE, an MnS consumer hosted on a network function, or a machine learning entity or application.

[0155] In 801, the method includes providing a request from a second entity to a first entity, the request to offload at least one inference step of a machine learning model for an application to a host node of a network, the network comprising a plurality of host nodes.

[0156] Optionally, in 802, the method includes providing first information related to the application from the second entity to the first entity.

[0157] In 803, the method includes receiving, at a second entity, from a first entity an indication of at least one determined deployment option for offloading at least one inference step on at least one of a plurality of host nodes, the at least one deployment option being determined by the first entity based on first information, second information related to a machine learning model, and third information related to at least one of the plurality of host nodes, wherein the third information includes at least an indication of a carbon emission footprint associated with at least one of the plurality of host nodes.

[0158] The first entity may include a managed service producer hosted on a network function. The second entity may include at least one of: a user device, a managed service consumer hosted on a network function, or a machine learning entity. The plurality of host nodes may include at least one of: an edge cloud server, a central cloud server, and a network function server.

[0159] An inference step can be called a model unit. A model unit is a step of inference in a machine learning model consisting of one or more neurons / layers. A model unit is the smallest unit to be allocated. The inference of the UE part and the inference of all model units in the network constitute the inference of the entire AI / ML model. A model unit can be identified by ModelUnitID.

[0160] The minimum number of available host nodes is one, and the theoretical maximum is the total number of network nodes. Available host nodes can be edge cloud, central cloud, available NF processing resources from underutilized NF (in-network computing concept). Each host node can be identified by NodeID.

[0161] The minimum number of selected host nodes is one and the maximum number is the total number of module units or network nodes.

[0162] The parameters of the model units, the minimum and maximum number of available and selected host nodes can be configured by the MnS producer at the beginning of the orchestration or derived from the received information related to the AI / ML models and host nodes. These parameters are the boundaries of the deployment options that will be derived. The configuration of these parameters is optional.

[0163] At least one deployment option may be determined such that the carbon footprint of the inference of the AI / ML model is minimized. The carbon footprint may be minimized subject to other requirements of the AI / ML model. The method may provide a function to orchestrate the host node selection and AI / ML model splitting of the inference AI / ML model in the network to achieve an overall optimal carbon footprint relative to a given QoS of the AI / ML application. Using the proposed method, the offloaded AI / ML inference task is split into several units, which are deployed on several selected distributed host nodes in the network. These selected host nodes perform the inference task in a joint manner.

[0164] Three information types may be defined to be used as input information to the proposed orchestration function to obtain one or several deployment options using the proposed method. It may be the case that more than one deployment option is obtained by the orchestration method using given information.

[0165] The first information may include a quality of service (QoS) indicator. The QoS indicator may be mandatory. The first information may also include at least one of the following: time sensitivity information, hardware requirements of at least one reasoning step, and data privacy requirements of at least one reasoning step. The first information may be referred to as application related information.

[0166] The first information, such as application related information, may be provided by, for example, a UE, an MnS consumer, or an ML entity / ML application. The quality of service indicator may be a QoS characteristic (e.g., latency, bitrate). The first information may include requirements regarding host node selection and AI / ML model splitting, such as privacy of data processing (this is an example of data privacy requirements for the reasoning step), dedicated hardware (this is an example of hardware requirements for the reasoning step), and time sensitivity.

[0167] Privacy of data processing may lead to exclusion of host nodes below a certain acceptable level. Dedicated hardware requirements may force deployment of a certain AI / ML model unit on selected host nodes. QoS of AI / ML model inference should be guaranteed.

[0168] The first information may include an indication of at least one inference step of the machine learning model to be offloaded to at least one host node among the plurality of host nodes.

[0169] For example, a second entity (e.g., a UE or MnS consumer) may define the portion of the AI / ML inference tasks to be offloaded in the network, e.g., Figure 4 and Figure 5 This can ensure the non-sustainability requirements of the application.

[0170] Alternatively, the second entity may not define what part of the AI / ML model to offload in the network, and the network will default to offloading the entire AI / ML model.

[0171] The request may include first information.

[0172] The second information may be at least one of the following: the input data size, output data size, computational complexity, and parameter size of at least one inference step of the machine learning model. The second information (e.g., AI / ML related information) may be provided by a network function (e.g., NWDAF, ADRF, or AIML training function / model producer). The second information may indicate the structure of the ML model without exposing proprietary details and computational complexity of the AI / ML model to be offloaded. With this information, the size of computational dependencies and intermediate results (the size of the intermediate results may be an example of the input or output data size of the inference step) may be obtained, which may play a role in model splitting (i.e., determining deployment options). For example, split points with larger sized intermediate results may be avoided.

[0173] The third information may also include computing capabilities (e.g., CPU / GPU capabilities, storage capabilities) of at least one host node summarized by the host node. The third information may be referred to as host node related information. The third information is provided by another network function (e.g., service exposure function).

[0174] The host node related information can show how much computing power (e.g., computing power, storage) can be provided by each available host node. It can also describe how much carbon will be generated by consuming a unit of energy on each available host node using a "carbon footprint index". The carbon footprint index is an example of an indication of a carbon footprint associated with at least one host node. The carbon footprint can be used as an environmental impact indicator for each available host node. Using this information, it can be ensured that the selected host node can provide sufficient computing power at a reduced carbon footprint.

[0175] Available network nodes can share the information "Carbon footprint index" as a new attribute to the NRF via existing interfaces. The "Carbon footprint index" varies depending on the region, energy source and time. This value can be collected by the network operator from its energy provider, or generated based on indices published by government and public organizations such as the European Environment Agency of the EU, the US Environmental Protection Agency or other third-party organizations.

[0176] Three information types are used as input, such as application related information, AI / ML related information, and host node related information. The attributes of the three information types are defined in Table 1.

[0177]

[0178]

[0179] Table 1

[0180] Based on the privacy of data processing and dedicated hardware, some available network nodes may be excluded for offloading. Model units with high workloads may prefer to be deployed on network nodes with low carbon footprint index. Model units can be deployed on different network nodes due to their computational dependencies. Some model units can be co-located on the same network node due to intermediate results.

[0181] In an example embodiment, the call flow may have the following steps.

[0182] The UE, any MnS consumer (e.g., AF, NF), or ML entity / ML application initiates a request to offload the model to the network in a distributed manner using a given application requirement. The UE or MnS consumer may choose to specify a portion of the inference task to be offloaded to the network, or not specify any portion and offload the entire inference task to the network. This is an example of providing a request from a second entity to a first entity to offload at least one inference step of a machine learning model for an application to a network comprising a plurality of host nodes.

[0183] The MnS producer hosted in the MDAF, AF or equivalent network management function will derive which AI / ML model unit should be deployed on which host node after collecting AI / ML and host node related information from the NWDAF or ADRF or AI / ML training function / model producer and or service exposure function (e.g., NRF). This is an example of determining at least one deployment option for offloading at least one inference step on at least one of the plurality of host nodes based on the first information, the second information, and the third information, wherein the third information includes at least an indication of a carbon emission footprint associated with at least one of the plurality of host nodes.

[0184] The obtained deployment options are sent to the offloading function, i.e. the MnS consumer or the UE. This is an example of providing an indication of at least one determined deployment option from a first entity to a second entity. The model units to be offloaded are then deployed on several host nodes in a joint manner. If needed, the intermediate results are shared between the host nodes. If they are needed for inference, the intermediate results are sent between the host devices.

[0185] In case more than one deployment option is obtained, the MnS consumer or UE selects one option to be deployed. This is an example of a second entity receiving an indication of multiple deployment options and determining to use one of the deployment options to offload at least one reasoning step to at least one host node. The selection may be random, based on criteria set by the UE (e.g., but not limited to, a minimum number of reasoning steps to be offloaded on the UE or a minimum amount of data forwarded by the UE), or any other suitable method for determining which of the multiple deployment options to use.

[0186] By considering the footprint of the intermediate data transmission 910, the footprint and computational complexity of the AL / ML model unit 920, and the computational dependencies of these units (arrow lines between the UE and the model unit 920), Fig. 9 A diagram showing two different example deployment (or “joint inference orchestration”) options (ie, deployment option 1 and deployment option 2 ) of a model unit 920 from a UE 900 to host nodes 901 and 902 .

[0187] Fig.10 An example of a call flow and signaling in an example embodiment is shown, where the second entity is a UE or an MnS consumer.

[0188] Step 1 includes initiating an orchestration request by a UE or an MnS consumer. The UE or the MnS consumer indicates requirements for orchestration defined in application related information in Table 1. These requirements include QoS that must be satisfied by inference. This is an example of receiving a request from a second entity at a first entity to offload at least one inference step of a machine learning model for an application to a host node of a network, the network including a plurality of host nodes, and receiving first information related to the application from the first entity.

[0189] If the UE defines the IE "Offload AI / ML model", these requirements may include model units of the AI / ML model to be offloaded to the network. Otherwise, the entire AI / ML model will be considered as a task to be offloaded during orchestration. The UE's requirements may include privacy and hardware requirements. In this example, the first information includes a QoS indicator and may include privacy requirements for at least one reasoning step, hardware requirements for at least one reasoning step, and an indication of at least one reasoning step of a machine learning model to be offloaded to at least one of a plurality of host nodes.

[0190] There are two example options to achieve initiation.

[0191] In option A, the UE or MnS consumer initiates and requests orchestration to the pointed MnS producer by sharing the application related information defined in Table 1.

[0192] Option B involves two steps.

[0193] In step 1a of option B, the UE initiates orchestration to the MnS consumer by sharing the application related information defined in Table 1.

[0194] In step 1b of option B, the MnS consumer requests the orchestration service from the MnS producer by forwarding the application related information from the UE.

[0195] The difference between Option A and Option B is which entity will enforce the derived deployment options after orchestration. In the case of Option A, the UE will receive the derived deployment options and select one to start enforcement. In the case of Option B, the MnS consumer entity will receive the derived deployment options and perform enforcement, and the UE will only get a response from the MnS consumer. These processes are described with reference to step 7 Option A and Option B, respectively.

[0196] In step 2, if the IE "Offload AI / ML model" is given by the UE, the MnS producer takes action to collect information about the computation and transmission resources required by the AI / ML model unit indicated by the UE. Otherwise, the MnS producer takes action to collect information about the computation and transmission resources required by all model units of the entire AI / ML model. The MnS producer requests the AI / ML related information defined in Table 1 from the NWDAF, ADRF or AIML training function / model producer.

[0197] In step 3, the NWDAF, ADRF or AIML training function / model producer provides the requested AI / ML related information and responds to the MnS producer. Steps 2 and 3 are examples of requesting second information from a network function and receiving second information from the network function.

[0198] In step 4, the MnS producer takes actions to collect information about computing and transmission resources that can be provided by all available network nodes. The MnS producer requests the host node related information defined in Table 1 from the service exposure function.

[0199] In step 5, the service exposure function provides the requested host node related information and responds to the MnS producer. Steps 4 and 5 are examples of requesting third information from another network function and receiving the third information from the another network function.

[0200] In step 6, after collecting all relevant information, the MnS producer derives deployment options for context-aware federated reasoning. Several network nodes are selected as host nodes, and model units deployed on these host nodes are determined. The host node selection and model unit deployment constitute the final deployment options. The deployment options can be in the form of the following tuple: OptionID–[String], list{NodeID–[String], ModelUnitID–[String]}. This is an example of determining at least one deployment option for offloading at least one reasoning step on at least one host node among the plurality of host nodes based on the first information, the second information, and the third information.

[0201] In step 7, after the MnS producer has derived the final deployment option(s), the final deployment option(s) are provided in response to the UE. There are two options to respond to the UE, corresponding to the two options of step 1. This step is an example of providing an indication of the determined at least one deployment option from the first entity to the second entity.

[0202] In option A, the MnS producer responds to the UE or MnS consumer with a deployment option.

[0203] Option B involves two steps. In step 7a, the MnS Producer responds to the MnS Consumer with the deployment options. In step 7b, the MnS Consumer responds to the UE with the completion of the request.

[0204] Fig.11 A call flow and signaling according to an example embodiment are shown, where the second entity is an AI / ML entity (“ML entity”) or an AI / ML application (“ML application”).

[0205] Step 1 includes initiating an orchestration request by the ML entity / ML application. The ML entity / ML application indicates a requirement for an orchestration defined in the application-related information in Table 1. This is an example of receiving, at a first entity, a request from a second entity to offload at least one inference step of a machine learning model for an application to a host node of a network, the network including a plurality of host nodes, and receiving first information related to the application from the first entity.

[0206] The requirements include the QoS that must be met for inference. If the ML entity / ML application defines the IE "Offload AI / ML model", the requirements may include the model units of the AI / ML model to be offloaded to the network. Otherwise, the entire AI / ML model will be considered as a task to be offloaded during orchestration. The requirements of the ML entity / ML application may include privacy and hardware. The ML entity / ML application requests orchestration services from the MnS producer. In this example, the first information includes a QoS indicator and may include a privacy requirement for at least one inference step, a hardware requirement for at least one inference step, and an indication of at least one inference step of the machine learning model to be offloaded to at least one of a plurality of host nodes.

[0207] In step 2, if IE "Offload AI / ML model" is given by the ML entity / ML application, the MnS producer takes action to collect information about the computing and transmission resources required by the AI / ML model unit indicated by the ML entity / ML application. Otherwise, the MnS producer takes action to collect information about the computing and transmission resources required by all model units of the entire AI / ML model. The MnS producer requests the AI / ML related information defined in Table 1 from the NWDAF, ADRF or AIML training function / model producer.

[0208] In step 3, the NWDAF, ADRF or AIML training function / model producer provides the requested AI / ML related information and responds to the MnS producer.

[0209] Step 2 and step 3 are examples of requesting the second information from the network function and receiving the second information from the network function.

[0210] In step 4, the MnS producer takes action to collect information about the computing and transmission resources that can be provided by all available network nodes. The MnS producer requests the host node related information defined in Table 1 from the service exposure function.

[0211] In step 5, the service exposure function provides the requested host node related information and responds to the MnS producer.

[0212] Step 4 and step 5 are examples of requesting the third information from the other network function and receiving the third information from the other network function.

[0213] In step 6, after collecting all relevant information, the MnS producer derives deployment options for context-aware federated reasoning. Several network nodes are selected as host nodes, and model units deployed on these host nodes are determined. The host node selection and model unit deployment constitute the final deployment options. The deployment options can be in the form of the following tuple: OptionID–[String], list{NodeID–[String], ModelUnitID–[String]}. This is an example of determining at least one deployment option for offloading at least one reasoning step on at least one host node among the plurality of host nodes based on the first information, the second information, and the third information.

[0214] In step 7, after the MnS producer exports the deployment option(s), the final deployment option(s) are provided to the ML entity / ML application. The ML entity / ML application will receive the exported deployment options and select one deployment option to initiate enforcement. This step is an example of providing an indication of at least one determined deployment option from the first entity to the second entity.

[0215] An apparatus may include: a component for receiving a request from a second entity at a first entity, the request being used to offload at least one inference step of a machine learning model for an application to a host node of a network, the network comprising a plurality of host nodes; a component for obtaining first information related to the application; a component for obtaining second information related to the machine learning model; a component for obtaining third information related to at least one host node among the plurality of host nodes, wherein the third information includes at least an indication of a carbon emission footprint associated with at least one host node among the plurality of host nodes; a component for determining at least one deployment option for offloading at least one inference step to at least one host node among the plurality of host nodes based on the first information, the second information and the third information; and a component for providing an indication of the determined at least one deployment option from the first entity to the second entity.

[0216] The apparatus may comprise a network function, may be a network function, or be comprised in a network function or a chipset for performing at least some actions of / for a network function.

[0217] An apparatus may include: a component for providing a request from a second entity to a first entity, the request being used to offload at least one inference step of a machine learning model for an application to a network, the network comprising a plurality of host nodes; and a component for receiving, at the second entity, from the first entity an indication of at least one determined deployment option for offloading the at least one inference step to at least one of the plurality of host nodes, the at least one deployment option being determined by the first entity based on first information related to the application, second information related to the machine learning model, and third information related to at least one of the plurality of host nodes, wherein the third information includes at least an indication of a carbon emission footprint associated with at least one of the plurality of host nodes.

[0218] The apparatus may comprise a user equipment such as a mobile phone, may be a user equipment or be comprised in a user equipment or a chipset for performing at least some operations of / for the user equipment.

[0219] Alternatively or additionally, the apparatus may comprise a network function, may be a network function or be comprised in a network function or a chipset for performing at least some actions of / for a network function.

[0220] It should be understood that the apparatus may include or be coupled to other units or modules for transmission and / or reception, such as a radio part or a radio head. Although the apparatus has been described as one entity, the different modules and memories may be implemented in one or more physical or logical entities.

[0221] It is worth noting that although some embodiments have been described with respect to 5G networks, similar principles may apply with respect to other networks and communication systems, such as 6G networks or 5G-advanced networks. Therefore, although certain embodiments are described above by way of example with reference to certain example architectures of wireless networks, technologies and standards, the embodiments may be applied to any other suitable form of communication system in addition to the communication system illustrated and described herein.

[0222] It is also noted herein that while the above describes example embodiments, there are several variations and modifications that may be made to the disclosed solution without departing from the scope of the present invention.

[0223] As used herein, “at least one of: ” and “at least one of ” and similar expressions, where a list of two or more elements is connected by “and” or “or”, mean at least any one of the elements, or at least any two or more of the elements, or at least all of the elements.

[0224] In general, various embodiments may be implemented in hardware or dedicated circuits, software, logic, or any combination thereof. Some aspects of the disclosure may be implemented in hardware, while other aspects may be implemented in firmware or software, which may be executed by a controller, microprocessor, or other computing device, but the disclosure is not limited thereto. Although various aspects of the disclosure may be illustrated and described as block diagrams, flow charts, or illustrated and described using some other graphical representation, it is fully understood that, as non-limiting examples, the blocks, devices, systems, techniques, or methods described herein may be implemented in hardware, software, firmware, dedicated circuits or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.

[0225] As used in this application, the term "circuitry" may refer to one or more or all of the following:

[0226] (a) pure hardware circuit implementation (e.g., implementation in analog and / or digital circuitry only) and (b) a combination of hardware circuitry and software, such as (where applicable):

[0227] (i) a combination of analog and / or digital hardware circuits and software / firmware and

[0228] (ii) any portion of hardware processor(s) with software (including digital signal processor(s), software and memory(s) that work together to enable a device, such as a mobile phone or server, to perform various functions) and (c) hardware circuit(s) and or processor(s), such as microprocessor(s) or portion(s) of microprocessor(s), which requires software (e.g. firmware) for operation, but where software is not required for operation, the software may be absent.”

[0229] This definition of circuitry applies to all uses of the term in this application, including in any claims. As another example, as used in this application, the term "circuitry" also covers an implementation of only a hardware circuit or processor (or multiple processors) or a portion of a hardware circuit or processor and its (their) accompanying software and / or firmware. For example and when applicable to a particular claim element, the term "circuitry" also covers a baseband integrated circuit or processor integrated circuit for a mobile device or a similar integrated circuit in a server, cellular network device, or other computing or network device.

[0230] Embodiments of the present disclosure may be implemented by computer software executable by a data processor of a mobile device (such as a processor entity), or by hardware, or by a combination of software and hardware. Computer software or programs (also referred to as program products) including software routines, applets and / or macros may be stored in any device-readable data storage medium, and they include program instructions for performing specific tasks. A computer program product may include one or more computer executable components that are configured to perform embodiments when the program is run. One or more computer executable components may be at least one software code or a portion thereof.

[0231] In addition, in this regard, it should be noted that any block of the logic flow shown in the figure can represent a program step, or an interconnected logic circuit, block and function, or a combination of a program step and a logic circuit, block and function. Software can be stored on a physical medium (such as a memory chip or a memory block implemented in a processor), a magnetic medium (such as a hard disk or a floppy disk), and an optical medium (such as, for example, a DVD and its data variant CD). Physical media is a non-transient medium. The term "non-transient" as used herein is a limitation of the medium itself (i.e., tangible, rather than a signal), rather than a limitation on data storage persistence (such as RAM and ROM).

[0232] The memory may be of any type suitable for the local technical environment and may be implemented using any suitable data storage technology, such as semiconductor-based memory devices, magnetic memory devices and systems, optical memory devices and systems, fixed memory, and removable memory. The data processor may be of any type suitable for the local technical environment and may include, as non-limiting examples, one or more of a general-purpose computer, a special-purpose computer, a microprocessor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an FPGA, a gate-level circuit, and a processor based on a multi-core processor architecture.

[0233] Embodiments of the present disclosure may be practiced in various components such as integrated circuit modules. The design of integrated circuits is generally a highly automated process. Complex and powerful software tools are available to convert a logic level design into a semiconductor circuit design ready to be etched and formed on a semiconductor substrate.

[0234] The scope of protection sought by various embodiments of the present disclosure is defined by the independent claims. Embodiments and features described in this specification that do not fall under the scope of the independent claims (if any) should be interpreted as examples that help understand the various embodiments of the present disclosure.

[0235] The foregoing description has provided a complete and informative description of exemplary embodiments of the present disclosure by way of non-limiting examples. However, when read in conjunction with the accompanying drawings and the appended claims, various modifications and adjustments may become apparent to those skilled in the relevant art in view of the foregoing description. However, all such and similar modifications of the teachings of the present disclosure will still fall within the scope of the present invention as defined in the appended claims. In fact, there are additional embodiments that include a combination of one or more embodiments with any other embodiments previously discussed.

Claims

1. A device for use in a communication network, comprising: means for receiving, at a first entity, a request from a second entity to offload at least one inference step of a machine learning model for an application to a host node of a network, the network comprising a plurality of host nodes; A component for obtaining first information related to the application; A component for obtaining second information related to the machine learning model; means for obtaining third information related to at least one of the plurality of host nodes, wherein the third information comprises at least an indication of a carbon emissions footprint associated with at least one of the plurality of host nodes; means for determining, based on the first information, the second information, and the third information, at least one deployment option to offload at least one inference step on at least one host node of the plurality of host nodes; as well as Means for providing, from the first entity to the second entity, an indication of the determined at least one deployment option.

2. The apparatus according to claim 1, wherein the means for obtaining the first information comprises: Means for receiving, at the first entity, the first information from the second entity.

3. The apparatus according to claim 1 or claim 2, wherein the means for obtaining the second information comprises: Means for requesting the second information from a network function and means for receiving the second information from the network function.

4. The device according to any one of claims 1 to 3, wherein the component for obtaining the third information comprises: Means for requesting the third information from another network function, and means for receiving the third information from the another network function.

5. The apparatus according to any one of claims 1 to 4, wherein the first information comprises: An indication of at least one inference step of the machine learning model to be offloaded to at least one host node of the plurality of host nodes.

6. The apparatus according to any one of claims 1 to 5, wherein the first information comprises a quality of service indicator. 7 . The apparatus of claim 6 , wherein the first information comprises at least one of: time sensitivity information, hardware requirements of at least one reasoning step, and data privacy requirements of at least one reasoning step.

8. The apparatus according to any one of claims 1 to 7, wherein the request comprises the first information.

9. The device according to any one of claims 1 to 8, wherein the third information further includes: An indication of a computing capability of at least one host node of the plurality of host nodes.

10. An apparatus according to any one of claims 1 to 9, wherein the second information comprises at least one of the following: input data size, output data size, computational complexity and parameter size of at least one inference step of the machine learning model.

11. The apparatus according to any one of claims 1 to 10, wherein the first entity comprises a managed service producer hosted on a network function.

12. An apparatus according to any one of claims 1 to 11, wherein the second entity comprises at least one of: a user equipment, a management service consumer hosted on a network function, or a machine learning entity.

13. The apparatus according to any one of claims 1 to 12, wherein the plurality of host nodes comprises at least one of the following: an edge cloud server, a central cloud server, and a network function server.

14. An apparatus for use in a communication network, comprising: means for providing a request from the second entity to the first entity to offload at least one inference step of a machine learning model for an application to a network, the network comprising a plurality of host nodes; as well as A component for receiving, at the second entity, from the first entity, an indication of at least one determined deployment option for offloading at least one inference step on at least one of the plurality of host nodes, the at least one deployment option being determined by the first entity based on first information related to the application, second information related to the machine learning model, and third information related to at least one of the plurality of host nodes, wherein the third information includes at least an indication of a carbon emission footprint associated with at least one of the plurality of host nodes.

15. The apparatus according to claim 14, comprising: Means for providing, from the second entity to the first entity, the first information.

16. The apparatus according to claim 14 or claim 15, wherein the first information comprises: An indication of at least one inference step of the machine learning model to be offloaded to at least one host node of the plurality of host nodes.

17. The apparatus according to any one of claims 14 to 16, wherein the first information comprises a quality of service indicator.

18. The apparatus of claim 17, wherein the first information comprises at least one of: time sensitivity information, hardware requirements of at least one reasoning step, and data privacy requirements of at least one reasoning step.

19. The apparatus of any one of claims 14 to 18, wherein the request comprises the first information.

20. The apparatus according to any one of claims 14 to 19, wherein the third information further comprises: An indication of a computing capability of at least one host node of the plurality of host nodes.

21. An apparatus according to any one of claims 14 to 20, wherein the second information comprises at least one of the following: input data size, output data size, computational complexity and parameter size of at least one inference step of the machine learning model.

22. An apparatus according to any one of claims 14 to 21, wherein the first entity comprises a managed service producer hosted on a network function.

23. An apparatus according to any one of claims 14 to 22, wherein the second entity includes at least one of the following: a user equipment, a management service consumer hosted on a network function, or a machine learning entity.

24. The apparatus according to any one of claims 14 to 23, wherein the plurality of host nodes include at least one of the following: an edge cloud server, a central cloud server, and a network function server.

25. The device according to any one of claims 14 to 24, comprising: Means for receiving an indication of a plurality of deployment options from the first entity and determining to use one of the deployment options to offload at least one inference step to at least one host node.

26. A method for use in a communication network, comprising: receiving, at a first entity, a request from a second entity to offload at least one inference step of a machine learning model for an application to a host node of a network, the network comprising a plurality of host nodes; Acquire first information related to the application; Acquire second information related to the machine learning model; obtaining third information associated with at least one of the plurality of host nodes, wherein the third information includes at least an indication of a carbon emissions footprint associated with at least one of the plurality of host nodes; determining, based on the first information, the second information, and the third information, at least one deployment option to offload at least one inference step on at least one host node among the plurality of host nodes; as well as An indication of the determined at least one deployment option is provided from the first entity to the second entity.

27. A method for use in a communication network, comprising: providing, from the second entity to the first entity, a request to offload at least one inference step of a machine learning model for an application to a network, the network comprising a plurality of host nodes; as well as Receiving, at the second entity, from the first entity, an indication of at least one determined deployment option for offloading at least one inference step on at least one of the plurality of host nodes, the at least one deployment option being determined by the first entity based on first information related to the application, second information related to the machine learning model, and third information related to at least one of the plurality of host nodes, wherein the third information includes at least an indication of a carbon emission footprint associated with at least one of the plurality of host nodes.

28. A computer-readable medium comprising instructions, which, when executed by an apparatus, cause the apparatus to at least perform the following: receiving, at a first entity, a request from a second entity to offload at least one inference step of a machine learning model for an application to a host node of a network, the network comprising a plurality of host nodes; Acquire first information related to the application; Acquire second information related to the machine learning model; obtaining third information associated with at least one of the plurality of host nodes, wherein the third information includes at least an indication of a carbon emissions footprint associated with at least one of the plurality of host nodes; determining, based on the first information, the second information, and the third information, at least one deployment option to offload at least one inference step on at least one host node among the plurality of host nodes; as well as An indication of the determined at least one deployment option is provided from the first entity to the second entity.

29. A computer-readable medium comprising instructions, which, when executed by an apparatus, cause the apparatus to at least perform the following: providing, from the second entity to the first entity, a request to offload at least one inference step of a machine learning model for an application to a network comprising a plurality of host nodes; and Receiving, at the second entity, from the first entity, an indication of at least one determined deployment option for offloading at least one inference step on at least one of the plurality of host nodes, the at least one deployment option being determined by the first entity based on first information related to the application, second information related to the machine learning model, and third information related to at least one of the plurality of host nodes, wherein the third information includes at least an indication of a carbon emission footprint associated with at least one of the plurality of host nodes.

Images