CN120835336A - QoS guarantee method, function, storage medium and computer program product - Google Patents

Abstract

The present application provides a quality of service (QoS) assurance method, function, storage medium, and computer program product, the method comprising: a first function receiving first information sent by a second function, the first information being used to describe requirements related to artificial intelligence (AI) service data transmission and/or requirements related to AI service computing power; generating a QoS control policy for the AI service based on the first information and an AI service template corresponding to a contracted AI service; and sending the QoS control policy to a third function, the QoS control policy being used by the third function to generate QoS parameters corresponding to the AI service based on the QoS control policy.

Description

QoS guarantee method, function, storage medium, and computer program product

Technical Field

The present application relates to the field of wireless communications technologies, and in particular, to a quality of service (Quality of Service, qoS) guarantee method, a function, a storage medium, and a computer program product.

Background

In the related art, the wireless network can provide QoS guarantee for the artificial intelligence (ARTIFICIAL INTELLIGENCE, AI) service, but the adopted technical scheme in the related art is easily affected by various factors such as network resources, network environment, resource allocation, flow control, management difficulty and the like, so that the QoS requirement of the high-quality AI service is difficult to meet.

Disclosure of Invention

In view of this, it is desirable to provide a QoS securing method, a function, a storage medium, and a computer program product, which can satisfy the QoS requirement of AI services.

The technical scheme of the embodiment of the application is realized as follows:

in a first aspect, an embodiment of the present application provides a QoS securing method, applied to a first function, where the method includes:

receiving first information sent by a second function, wherein the first information is used for describing requirements related to AI service data transmission and/or requirements related to AI service computing power;

generating QoS control strategy of AI service based on the first information and AI service template corresponding to the subscribed AI service;

the QoS control policy is sent to the third function, the QoS control policy being for the third function to generate QoS parameters corresponding to the AI services based on the QoS control policy.

In a second aspect, an embodiment of the present application provides a QoS securing method, applied to a third function, the method including:

receiving a QoS control strategy sent by a first function, wherein the QoS control strategy is generated by the first function based on first information sent by a second function and an AI service template corresponding to a subscribed AI service;

generating QoS parameters corresponding to the AI service based on the QoS control strategy;

The QoS parameters are sent to the equipment related to data transmission and/or the equipment related to computing power, and the QoS parameters are used for the equipment related to data transmission and/or the equipment related to computing power to carry out QoS guarantee on the AI service based on the QoS parameters.

In a third aspect, an embodiment of the present application provides a first function, including:

the first receiving unit is used for receiving first information sent by the second function, wherein the first information is used for describing requirements related to AI service data transmission and/or requirements related to AI service computing power;

The first processing unit is used for generating a QoS control strategy of the AI service based on the first information and an AI service template corresponding to the subscribed AI service;

And the first sending unit is used for sending the QoS control strategy to the third function, and the QoS control strategy is used for the third function to generate QoS parameters corresponding to the AI service based on the QoS control strategy.

In a third aspect, embodiments of the present application provide a third function, the third function comprising:

the second receiving unit is used for receiving the QoS control strategy sent by the first function, wherein the QoS control strategy is generated by the first function based on the first information sent by the second function and an AI service template corresponding to the subscribed AI service;

A second processing unit, configured to generate QoS parameters corresponding to the AI service based on the QoS control policy;

And the second sending unit is used for sending the QoS parameters to the equipment related to data transmission and/or the equipment related to computing power, wherein the QoS parameters are used for guaranteeing the QoS of the AI service based on the QoS parameters by the equipment related to data transmission and/or the equipment related to computing power.

In a fifth aspect, an embodiment of the present application provides a first function, where the first function includes a first processor and a first memory, and when the first processor executes an operation program stored in the first memory, the QoS guarantee method on the first function side is implemented.

In a sixth aspect, an embodiment of the present application provides a third function, where the third function includes a second processor and a second memory, and the QoS guarantee method on the third function side is implemented when the second processor executes an operation program stored in the second memory.

In a seventh aspect, an embodiment of the present application provides a storage medium having stored thereon a computer program that, when executed by a processor, implements the QoS securing method on the first functional side described above, or that, when executed by a processor, implements the QoS securing method on the third functional side described above.

In an eighth aspect, an embodiment of the present application provides a computer program product, including a computer program that implements the QoS securing method of the first function side described above when executed by a processor, or implements the QoS securing method of the third function side described above when executed by a processor.

The embodiment of the application provides a QoS guarantee method, a function, a storage medium and a computer program product, wherein the method comprises the steps that a first function receives first information sent by a second function, the first information is used for describing requirements related to AI service data transmission and/or requirements related to AI service computing power, a QoS control strategy of the AI service is generated based on an AI service template corresponding to the first information and the subscribed AI service, the QoS control strategy is sent to a third function, the QoS control strategy is used for enabling the third function to generate QoS parameters corresponding to the AI service based on the QoS control strategy, the third function receives the QoS control strategy sent by the first function, the QoS control strategy is generated by the first function based on the first information sent by the second function and the AI service template corresponding to the subscribed AI service, the QoS parameters are generated based on the QoS control strategy, the QoS parameters are sent to equipment related to the data transmission and/or equipment related to the computing power, and the QoS parameters are used for guaranteeing the QoS of the service based on the QoS parameters. After the second function transmits the first information to the first function, the first information corresponds to the data transmission related requirement of the AI service and the calculation power related requirement of the AI service, so that the QoS parameters finally generated according to the first information and the AI service template also correspond to the specific requirement of the AI service, and when the related equipment for data transmission and/or the equipment related to the calculation power adopt the corresponding QoS parameters to execute the corresponding AI service, the specific requirements of different AI services on QoS can be met.

Drawings

Fig. 1 is a schematic flow chart of a QoS securing method according to an embodiment of the present application;

fig. 2 is a schematic diagram of an implementation process corresponding to a car networking-intelligent collision prediction according to an embodiment of the present application;

Fig. 3 is a schematic flow chart II of a QoS securing method according to an embodiment of the present application;

Fig. 4 is a schematic overall flow chart of a QoS securing method according to an embodiment of the present application;

Fig. 5 is an overall flowchart of AI service processing according to an embodiment of the present application;

Fig. 6 is a schematic diagram of a first functional component provided in an embodiment of the present application;

fig. 7 is a schematic diagram of a first functional component according to a second embodiment of the present application;

Fig. 8 is a schematic diagram of a third functional composition provided in an embodiment of the present application;

fig. 9 is a schematic diagram of a third functional component according to the second embodiment of the present application.

Detailed Description

In order to more fully understand the features and technical content of the embodiments of the present application, the following detailed description of the embodiments of the present application is provided with reference to the accompanying drawings, which are not intended to limit the embodiments of the present application.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the embodiments of the application is for the purpose of describing embodiments of the application only and is not intended to be limiting of the application.

In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is to be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with one another without conflict. It should also be noted that the term "first/second/third" is used merely to distinguish similar objects and does not represent a specific ordering of objects, it being understood that the "first/second/third" may be interchanged with a specific order or sequence, as permitted, to enable embodiments of the application described herein to be implemented in an order other than that illustrated or described herein.

In the related art, the method for providing QoS guarantee for AI services by the wireless network generally adopts the following key technical schemes:

(1) Resource management based on service level agreements (SERVICE LEVEL AGREEMENT, SLA) determines performance metrics and security requirements of AI services, including latency, bandwidth, availability, etc., by establishing SLAs. The resource management based on SLA can dynamically allocate and manage network resources according to different AI service demands, and ensure that the service quality requirements are met.

(2) Network slicing techniques-utilizing network slicing techniques, a wireless network is divided into multiple virtual slices, each of which may provide dedicated resources for a particular AI service. By providing independent network slices for different AI services, resource isolation can be achieved, avoiding resource contention, and thus guaranteeing QoS for AI services.

(3) Reserving resource strategy, namely reserving certain resource capacity for key AI service to ensure priority and performance of the key AI service. The reserved resource strategy can be used for preferentially scheduling data transmission and calculation tasks of the key AI service when the network is congested, so that the stability and timeliness of the key AI service are ensured.

(4) Dynamic resource allocation and adjustment, namely dynamically allocating and adjusting network resources according to real-time change of AI service requirements. And carrying out resource scheduling and optimization by monitoring the network condition and the load condition of the AI service in real time so as to meet the service quality requirement of the AI service.

(5) QoS-aware flow control, namely limiting the flow of each AI service in the network through a flow control mechanism, and avoiding excessive consumption or congestion of resources. Different flow scheduling algorithms are adopted for different AI services, so that the network resource utilization is optimized, and the QoS can meet the requirements.

In summary, through SLA-based resource management, network slicing technology, reserved resource policy, dynamic resource allocation and adjustment, and QoS-aware flow control, the wireless network can provide QoS guarantee for AI services. The technical methods can dynamically allocate and manage network resources according to service demands, and ensure the performance and reliability of each AI service.

However, although the method in the related art can implement the wireless network to provide QoS guarantee for AI services, the following disadvantages still exist, mainly including:

(1) Resource limitations-computing resources and bandwidth in wireless networks are limited, making it difficult to meet the high quality AI service requirements. When multiple AI services request resources at the same time, resource bottlenecks and performance degradation may result.

(2) The network environment is difficult to predict, the network environment of the wireless network is complex and changeable, and network delay and bandwidth capacity can be affected by other factors, such as network congestion, signal interference and the like. Making it more difficult to accurately predict and guarantee QoS of AI services.

(3) SLA management complexity-setting up and managing SLAs requires negotiating and agreeing on various metrics, including latency, bandwidth, availability, etc. Coordination and collaboration of multiple parties may be involved and there may be problems with ambiguity of interpretation, disputes, etc.

(4) Dynamic resource allocation is difficult, and even if a strategy of dynamic resource allocation and adjustment is adopted, network resources are adjusted in real time according to AI service requirements, so that the implementation of the method can face a certain challenge. The accuracy and efficiency of resource allocation is often limited by the performance of real-time monitoring and prediction.

(5) Complexity of flow control-achieving good flow control and scheduling is a complex task in high load situations. The traffic of each AI service is reasonably distributed, so that each service can meet QoS requirements, various network factors need to be comprehensively considered, and an intelligent scheduling algorithm is adopted.

In order to solve the above problems, the embodiments of the present application provide a QoS guarantee method, where a wireless network may provide required computing power, connection resources, etc. for an AI Service according to a QoS guarantee level of the AI Service, so as to solve the problem that in the related art, the wireless network does not have QoS guarantee for traffic transmission related to the AI Service and QoS guarantee for the computing power resources, so as to meet the requirements of a network edge AI Service (english may be expressed as Service) for low Service response delay and high Service accuracy.

The technical solution in the embodiment of the present application is implemented in such a way that the embodiment of the present application provides a QoS securing method, as shown in fig. 1, applied to a first function, where the method may include:

s101, receiving first information sent by a second function.

Wherein the first information is used to describe requirements related to AI-service data transfer and/or requirements related to AI-service computing power.

In the embodiment of the present application, the first function and the second function may be called a functional body (the first function is called a first functional body, the second function is called a second functional body), a communication node (the first function is called a first communication node, the second function is called a second communication node), a network element (the first function is called a first network element, the second function is called a second network element), or the like.

In an embodiment of the present application, the first function may include a policy control function (Policy Control Function, PCF).

In an embodiment of the application, the second function may comprise an application function (Application Function, AF).

In an embodiment of the present application, the first function is configured to generate a QoS control policy for an AI service, and the second function is configured to send a demand related to data transmission of the AI service and/or a demand related to computational effort of the AI service to the second function.

In the embodiment of the present application, the first information may be service information (english may be expressed as service information).

In the embodiment of the present application, after a session (english may be expressed as session) is established/changed between a User Equipment (UE) and an AF, the AF converts the SDI received at AF session signalling into service information and transmits the SDI to the PCF because an AF session signal (english may be expressed as session signalling) may carry a session description language (i.e., SDI).

In the embodiment of the present application, for the requirements related to AI service data transmission, the data service to be transmitted by AI service is used as a new service, the media component (english may be expressed as MediaComponent) in the session description protocol (Session Description Protocol, SDP) is added with the media type (english may be expressed as MediaType), and further the corresponding requirements correspondingly modify the QoS parameters and mapping relations in the SDI mapping function (english may be expressed as SDI MAPPING function), service information, authorized QoS parameters (english may be expressed as Authorized QoS parameters), and QoS configuration file (english may be expressed as QoS profile).

In the embodiment of the application, aiming at the requirements related to the calculation power of the AI service, the AI task requiring the calculation power of the AI service is added in the SDP as a new calculation component (English can be expressed as Computing Component), and corresponding requirements further modify QoS parameters and mapping relations in SDI MAPPING functions, service information, authorized QoS parameters and QoS profiles correspondingly.

In the embodiment of the present application, for the AI service, the AF needs to have a function of converting the requirements related to AI service data transmission and/or the requirements related to AI service computing power into service information, so that the function of converting the requirements related to AI service in SDP parameters (english may be expressed as SDP PARAMETERS) into service information needs to be added to the AF. Referring to the related content of the SDP parameters in the related art, as shown in table 1 below, table 1 lists the values of SDP parameters transmitted by an application server.

Wherein V represents a protocol version, o represents a session creator, s represents a session name, i represents session information, t represents time information, m represents a media description, c represents connection information, and a represents an attribute.

TABLE 1

In the embodiment of the present application, the function of converting the requirements related to AI service data transmission into service information in AF is added SDP PARAMETERS, and specifically, as shown in table 2 below, a rule for deriving service information in a media component description from an SDP media component (english may be expressed as Rules for derivation of service information within MediaComponent Description from SDP media component). when a session is started or modified, and AF should derive a media component description AVP of an Rx interface or a "media component" attribute of an N5 interface from SDP parameters) is described in table 2 with reference to related technologies, wherein SDP parameters are described in an internet function task group (THE INTERNETENGINEERING TASK Force, IETF) solicitation opinion document (Request For Comments, RFC) (english may be expressed as THE SDP PARAMETERS ARE described IN IETF RFC).

TABLE 2

In the embodiment of the present application, a type media component (english may be expressed as Definition of type MediaComponent) is defined, and reference is made to the related art, as shown in table 3 below.

TABLE 3 Table 3

Wherein medType denotes a media type parameter, maxSuppBwDl denotes a downlink support maximum bandwidth parameter, maxSuppBwUl denotes an uplink support maximum bandwidth parameter, and tsnQos denotes a transmission QoS parameter.

In The embodiment of The present application, the Enumeration media type (English may be expressed as Enumeration MediaType) defines The establishment: mediaType is that The Enumeration "MediaType" represents The media type of The media component (English may be expressed as The establishment "MediaType" REPRESENTS THE MEDIA TYPE of a media component), and reference may be made to The related art, as shown in The following table 4.

TABLE 4 Table 4

In summary, the following increases are mainly made with respect to the related art:

1) AISERVICEDATA and AIServiceModel, which are added to the MediaType Enumeration value in table 4 above, are added to the MediaType Enumeration value, which is shown in table 5 below.

TABLE 5

2) In Table 3 of Definition of type MediaComponent, i.e., for the related art, AISERVICEDATA/AIServiceModel is added at Applicability and QoS parameters corresponding to AISERVICEDATA or AIServiceModel are added at the Attribute name, and some of the additions to Table 3 are exemplified in Table 6, as shown in Table 6 below.

TABLE 6

It should be noted that, the added contents in table 6 are only for illustrating the positions to be modified and the newly added portions of the contents, and specifically, the contents may be added according to the actual situation, which is not specifically limited in the embodiment of the present application.

3) In Rules for derivation of service information within Media Component Description from SDP media component, i.e., for the related art, a rule for QoS parameter conversion to Service information is defined in table 2 above.

Illustratively, as shown in Table 7 below, the AImaxBwUl parameter is taken as Service information per Media, where the value of AI Max Requested Bandwidth-UL is converted from the corresponding derivative from SDP PARAMETER, i.e., service information may be converted using the corresponding derivative from SDP PARAMETER execution of the parameter.

It should be noted that, compared to the related art, the parameter Max Requested Bandwidth-UL may be a substitute for the original marBwUl parameter.

TABLE 7

In the embodiment of the present application, a function of converting the demand related to the calculation force of the AI service calculation task in SDP PARAMETERS into service information is also required to be added in the AF. Specifically, Rules for derivation of service information within Media Component Description from SDP media component, referred to in table 2 may be analogically referred to as the following increases:

1) A new type Computing Component is added to the SDP.

2) Various types of Computing tasks are filled in Enumeration value of the Computing Type, including training tasks for AI service Computing tasks, reasoning tasks, model selection tasks, model update tasks, and the like. As shown in Table 8 below, an added enumeration calculation type (English may be expressed as Enumeration ComputingType).

TABLE 8

3) In the calculation component type definition (english may be expressed as Definition of type Computing Component) table (refer to table 3 above), qoS parameters are formulated as attributename for different AI service calculation tasks as Applicability, so that QoS parameters required to be guaranteed by the AI service task (english may be expressed as AI SERVICE TASK) can be expressed. The table corresponding to the increment Definition of type Computing Component is shown in table 9 below.

TABLE 9

Based on the above embodiment, by adding the function of converting the requirements related to AI services and the requirements related to AI service computing task computing power in SDP PARAMETERS into service information in the AF, it is ensured that after the AF establishes session connection with the UE, the requirements related to AI services and the requirements related to AI service computing task computing power can be converted into service information, sent to the PCF, and further processed by the PCF.

S102, generating QoS control strategies of the AI service based on the first information and the AI service templates corresponding to the subscribed AI service.

In the embodiment of the application, the AI service template at least comprises one or more parameters including QoS parameters related to data transmission corresponding to QoS guarantee level of the AI service and/or QoS parameters related to computing power corresponding to the QoS guarantee level.

In the embodiment of the application, the QoS parameters related to the Computing power corresponding to the QoS guarantee level at least comprise one or more of Computing power resource type (Computing Resource type can be expressed in English), floating point operation times (computing_ Guaranteed Flops, GFPS) executed by Computing power guarantee per second, floating point operation times (computing_ Maximum Flops, MFPS) executed by Computing power maximum per second and Computing power priority (Computing Priority Level can be expressed in English).

In the embodiment of the application, an AI service template containing contracted AI service and QoS guarantee grade information is designed, and QoS parameters related to computing power and data transmission corresponding to QoS grades are provided in the AI service template. The AI service ID (AI service sub-scene ID or AI service scene ID) and QoS guarantee level can be managed and numbered according to scene layering, more scenes of AI services available by the network can be expanded, qoS level is subdivided aiming at the same AI service, and requirements of more kinds of users are met.

In the embodiment of the application, an AI service user subscribes an AI service which can be provided by a network and a QoS guarantee level related to AI service experience to an AI service platform, and the AI service platform inquires a corresponding AI service template according to the AI service subscribed by the user and the QoS guarantee level.

The specification of the AI service template is not limited to the standardized formulation, the network operator pre-configuration, and the like.

In an embodiment of the present application, the AI service templates may be as listed in tables 10-12 below.

Table 10 describes AI service instances to which a user may subscribe according to service scenarios and quality of service requirements.

Table 11 describes QoS parameters related to data transmission corresponding to AI service QoS guarantee levels.

Table 12 describes the QoS parameters associated with the computing resources corresponding to the a-service QoS security level.

Table 10

AI service case number

AI service scene ID

AI service sub-scene ID

QoS guarantee level

TABLE 11

Table 12

Taking the internet of vehicles-intelligent collision prediction as an example, fig. 2 is a schematic diagram illustrating the implementation process of the internet of vehicles-intelligent collision prediction, and the AI service templates corresponding to the internet of vehicles-intelligent collision prediction may be shown in table 13 below.

TABLE 13

For flammable and explosive vehicles, the QoS guarantee level of the vehicle is 1.1.1, for vehicles with dead zones such as buses and trucks, the QoS guarantee level of the vehicle is 1.1.2, and the anti-collision information of the vehicle is 1.1.3.

As shown in table 14 below, the network data transmission QoS requirements for the internet of vehicles-intelligent collision prediction are described in table 14.

TABLE 14

As shown in table 15 below, the network power QoS requirements for the internet of vehicles-intelligent collision prediction are described in table 15.

TABLE 15

In the embodiment of the present application, a part of contents listed in the content AI service templates contained in the above tables 10 to 12, specifically, the AI service templates include, but are not limited to, the following:

1) The AI service case number can uniquely represent the corresponding relation between the AI service and the QoS guarantee level.

2) AI service scene ID, network pre-numbering according to AI service application scene available.

3) The network can manage and number according to the scene layering, which is beneficial to expanding more scenes of the AI service which can be provided by the network.

4) QoS grade guarantee grade, the network classifies according to the available user experience guarantee, and the network further designs corresponding QoS parameters according to the QoS guarantee grade.

The AI service scene ID and QoS guarantee level can be managed and numbered according to the scene layering, and the method has the advantages that more scenes of AI services which can be provided by a network are expanded, and QoS guarantee levels are subdivided for the same AI service, so that the requirements of more kinds of users are met.

5) QoS parameters associated with AI-service data transmissions include, but are not limited to:

A. 5G QoS identifier (i.e., 5 QI).

B、ARP。

C. Notification control (english may be expressed as Notification control).

D. Traffic bit rate (english may be expressed as Flow Bit Rates) includes:

1) Guaranteed traffic bit rate (Guaranteed Flow Bit Rate, GFBR) -DL and UL;

2) Maximum traffic bit rate (Maximum Flow Bit Rate, MFBR) -DL and UL.

E. Aggregate bit rate (english may be expressed as AGGREGATE BIT RATES) includes:

1) Each UE aggregates a maximum bit rate (per UE Aggregate Maximum Bit Rate, UE-AMBR);

2) Maximum Bit Rate per Slice per UE (per UE per Slice-Maximum Bit Rate, UE-Slice-MBR).

Wherein the 5G QoS characteristics (characteristics) associated with 5QI include:

F. Resource type (Non-GBR, GBR, delay-critical) GBR;

G、Priority Level;

H. packet delay Budget (including core network packet delay Budget) (english may be expressed as PACKET DELAY budgets (including Core Network PACKET DELAY budgets));

I. packet error rate (Packet Error Rate);

J. Average window (applicable only to GBR and delay critical GBR resource types) (english may be expressed as AVERAGING WINDOW (for GBR and Delay-critical GBR resource type only));

K. The maximum data burst (applicable only to Delay critical GBR resource types) (english may be expressed as Maximum Data Burst Volume (for Delay-critical GBR resource type only)).

Based on the QoS parameters related to the data transmission of the AI service, the service guarantee QoS parameters related to the AI service computing power are correspondingly designed, including but not limited to the following:

a. calculating_5QI is a scalar, each value corresponding to a defined 5G calculating QoS feature parameter (e.g., d through f below);

b. ComputingARP (ARP English full name can be expressed as Allocation and Retention Priority, chinese is interpreted as allocation and retention priority) contains priority, preemption capability and preemption vulnerability;

c. Calculating_ Flops the number of floating point operations per second (English may be expressed as flowing-point operations per second) including:

1) The power guarantee number of floating point operations performed per second (computing_ Guaranteed Flops, GFPS);

2) The number of floating point operations performed per second (computing_ Maximum Flops, MFPS) with maximum Computing power;

5G Computing QoS characteristics associated with the calculation-5 QI therein includes:

d、Computing_Resource_type:(Non-Guaranteed,Guaranteed,Delay-critical Guaranteed);

e. The Priority of the same AI calculation task is expressed, so that different AI services of the same user can be distinguished in resource scheduling, and different users can be distinguished;

f. Calculating_ AVERAGING WINDOW (only applicable to Guaranteed resource types (english may be expressed as for Guaranteed, delay-critical Guaranteed resource type only)) that are critical to Guaranteed, delay;

g、Computing Packet Delay Budget。

In the embodiment of the application, because the AI service template comprises QoS guarantee grades corresponding to the AI service, the network pre-configures related QoS parameters (including QoS parameters related to data transmission and QoS parameters related to computing power) for the AI service according to the QOS grades subscribed by the user, establishes a detection mechanism of the AI service data business, and provides required connection resources and/or computing power resources for the AI service by the network equipment when detecting different AI services through the AI service distribution function of the UPF. QoS parameters associated with data transmission for the corresponding AI service are used when the data bearer is established (i.e., core network to UE), and computational effort related QoS parameters for the corresponding AI service are used when the computational tasks are deployed. Specifically, when the UPF data distribution function detects related data service, the IP quintuple or service detection algorithm can be used to distinguish the data of different services, the corresponding connection QoS parameter related to the data is used to ensure the service AI service, the detection mechanism of the AI task is established, when the AI task distribution function detects related AI task, the detection algorithm related to the AI task can be used to distinguish the different AI tasks, and the corresponding calculation related QoS parameter is used to ensure the AI service.

In the embodiment of the present application, after the network pre-configures relevant QoS parameters for AI services, when the PCF receives Service information sent by the AF, the PCF converts the received service information into Authorized QoS parameters/PER SERVICE DATA flows, and the PCF merges Authorized QoS parameters/per flows in each direction (i.e. after PER SERVICE DATA flows are merged) and delivers to the third function.

The third function may be called a third functional body, a third communication node, or a third network element, and the name of the third function is not specifically limited in the embodiment of the present application, and different modes may be selected according to the effect of the third function, so long as the corresponding function can be implemented.

In an embodiment of the application, the third function comprises a session management function (Session Management Function, SMF).

In the embodiment of the present application, when the PCF converts service information to Authorized QoS parameters, the PCF needs to have a corresponding function, and the function of converting QoS parameters related to AI services in SDP PARAMETERS to Authorized QoS parameters can be added to the PCF.

In the present embodiment, the PCF is typically configured to extract an Authorized 5G QoS Identifier (i.e., 5 QI), grant allocation and retention priority (Authorized Allocation and Retention Priority, ARP), and grant maximum/guaranteed data rate UL/DL (english may be expressed as Authorized Maximum/Guaranteed Data Rate UL/DL).

Thus, the function of converting QoS parameters related to AI-service data transmission in service information to Authorized QoS parameters can be added to PCF.

In the embodiment of the present application, referring to the related art, according to the maximum authorized data rate, the authorized guaranteed data rate, the maximum authorized QoS class, and the derivation rule of other authorized QoS parameters of each service data flow or the bidirectional combination of service data flows in the PCF (english may be expressed as Rules for derivation of the Maximum Authorized Data Rates,Authorized Guaranteed Data Rates,Maximum Authorized QoS Class and other authorized QoS parameters per service data flow or bidirectional combination of service data flows in the PCF),, for example, the information of QoS parameters related to AI service data transmission in the following table 16 may be added on the basis of the related art, as shown in the following table 16:

Table 16

Note that, in table 16 above, compared with Rules for derivation of the Maximum Authorized Data Rates,Authorized Guaranteed Data Rates,Maximum Authorized QoS Class and other authorized QoS parameters per service data flow or bidirectional combination of service data flows in the PCF in the related art, max_dr_dl/UL in the related art table may be replaced with ai_max_dr_dl/UL in table 16 above, and AImaxBwDl/UL is referred to in the formula of the corresponding column "Derivation from service information", so that Authorized QoS parameters is further obtained through conversion.

In the embodiment of the present application, referring to the related art, according to the calculation rules (english may be expressed as Rules for calculating the Maximum Authorized/supported DATA RATES,5QI and ARP in the PCF) of maximum grant/Guaranteed data rate, 5QI and ARP in PCF, for example, the information of QoS parameters related to AI service data transmission in the following table 17 may be added on the basis of the related art.

TABLE 17

Further, authorized QoS parameters of all SERVICE DATA flows defined in PCC rule, authorized QoS parameters of all SERVICE DATA flows in PDU session, or AF session (english may be expressed as ALL SERVICE DATA flows with corresponding AF session) corresponding to all service data flows may be further obtained according to table 17.

In the embodiment of the application, the function of converting QoS parameters related to the calculation power of the AI service calculation task in service information into Authorized QoS parameters can be added in the PCF.

In the embodiment of the present application, referring to the related art, according to the maximum authorized data rate, the authorized guaranteed data rate, the maximum authorized QoS class, and the derivation rule of other authorized QoS parameters of each service data flow or the bidirectional combination of service data flows in the PCF (english may be expressed as Rules for derivation of the Maximum Authorized Data Rates,Authorized Guaranteed Data Rates,Maximum Authorized QoS Class and other authorized QoS parameters per service data flow or bidirectional combination of service data flows in the PCF), information of QoS parameters related to the computing power of the AI service computing task may be added on the basis of the related art.

In the embodiment of the present application, referring to the related art, according to the maximum authorized/Guaranteed data rate in PCF, calculation rules of 5QI and ARP (english may be expressed as Rules for calculating the Maximum Authorized/supported DATA RATES,5QI and ARP in the PCF), the information of QoS parameters related to the calculation effort of AI service may be added on the basis of the related art. Specifically, reference may be made to the content related to AI-service data transmission enumerated in the foregoing embodiment.

Further, authorized QoS parameters of all AI SERVICE TASK defined in PCC rule, authorized QoS parameters of all SERVICE DATA flows in PDU session, or all service data flows and corresponding AF sessions (english may be expressed as ALL SERVICE DATA flows with corresponding AF session) may be further obtained.

It should be noted that, the added parameter related to the AI service computing force may refer to an adding manner of the parameter related to AI service data transmission, which is not described herein.

In the embodiment of the application, task Detection Rule is added to the possible AI TASK TYPE increase in the AI service in the PCC rule. QoS parameters are formulated for each AI TASK TYPE according to the QoS class, so that the QoS of the AI computation task required in the AI service can be satisfied as a whole. Possible AI TASK TYPE in the AI service include, but are not limited to, training tasks, reasoning tasks, model selection tasks, model update tasks, and the like.

S103, sending the QoS control strategy to a third function.

Wherein the QoS control policy is used for the third function to generate QoS parameters corresponding to the AI service based on the QoS control policy.

In an embodiment of the present application, the PCF sends Authorized QoS parameters to the SMF, which further generates QoS parameters corresponding to the AI services according to Authorized QoS parameters.

It can be understood that, in the QoS securing method provided in the embodiment of the present application, after the second function transmits the first information to the first function, the first information corresponds to the data transmission related requirement of the AI service and the calculation power related requirement of the AI service, so that the first function generates the QoS control policy of the AI service according to the first information and the AI service template to correspond to the corresponding AI service requirement, and further the QoS parameter finally generated by the third function corresponds to the specific requirement of the AI service, so that when the corresponding AI service is executed by the equipment related to the data transmission and/or the equipment related to the calculation power by adopting the corresponding QoS parameter, the specific requirement of different AI services on QoS can be satisfied.

In an embodiment of the application, after the PCF sends the QoS control strategy to the third function, the PCF monitors QoS information corresponding to the AI service, acquires network connection resources and computing resources at the current moment when the QoS information corresponding to the AI service is not satisfied, updates the QoS control strategy of the AI service based on the network connection resources and the computing resources, and sends the updated QoS control strategy to the third function, wherein the updated QoS control strategy is used for the third function to generate updated QoS parameters corresponding to the AI service based on the updated QoS control strategy.

In the embodiment of the present application, the current time may be understood as the time of monitoring QoS information.

In the embodiment of the application, a PCF network element in a network monitors QoS information of an AI service, when the QoS information of the AI service does not meet the service requirement (such as lower prediction accuracy), network connection resources and computing resources at the current moment are acquired, and according to the monitored QoS information and the network connection and computing resources, qoS parameters are selected again in a mapping relation table of QoS guarantee levels and QoS parameters (namely, the corresponding relation between one QoS guarantee level and one QoS parameter can be understood), and an AI service template corresponding to the QoS parameters of the AI service network resources is regenerated, so that the PCF can regenerate a QoS control strategy based on the regenerated QoS parameters.

It should be noted that, the process of regenerating the QoS control policy may refer to the implementation process of the foregoing embodiment, which is not described herein.

Illustratively, taking the internet of vehicles-intelligent collision prediction as an example, the QoS parameters in the AI service template are updated according to the network-monitored QoS parameters.

When the network detects the QoS of the AI service to be reduced, one reason is that the terminal computing power resource where the AI task 1 (predictive algorithm) is deployed is sufficient, and the base station computing power resource where the AI task 2 (decision algorithm) is deployed is sufficient, but the resource terminal and the base station uplink and downlink channel environments are poor, so that the AI service delay is too high.

In order to ensure better QoS of the AI service, the QoS parameters in the AI service template can be slightly adjusted, for example, in the case, the QoS parameters in the aspect of calculation force can be finely adjusted to reduce the priority of the AI task, the QoS parameters in the aspect of data transmission can be finely adjusted to improve the transmission priority of the data transmission service of the AI task, and the real-time network resource is more adapted.

Referring to tables 13 to 15, for AI service case No. 1.1.1-internet of vehicles-intelligent collision prediction-1.1.1, the corresponding adjustment is shown in tables 18 to 19 below, table 18 shows QoS requirements of internet of vehicles-intelligent collision prediction for network data transmission adjustment, and table 19 shows QoS requirements of internet of vehicles-intelligent collision prediction for network calculation adjustment.

TABLE 18

TABLE 19

The embodiment of the application also provides a QoS guarantee method, as shown in fig. 3, applied to a third function, the method may include:

S201, receiving a QoS control strategy sent by a first function.

The QoS control strategy is generated by the first function based on the first information sent by the second function and an AI service template corresponding to the subscribed AI service.

In the embodiment of the present application, the first function and the third function have been explained in the foregoing embodiment, and are not repeated here.

In the embodiment of the present application, the process of generating the QoS control policy by the PCF may refer to the implementation process of the foregoing embodiment, which is not described herein in detail.

S202, generating QoS parameters corresponding to the AI service based on the QoS control strategy.

In the embodiment of the application, the QoS parameters comprise QoS parameters related to data transmission corresponding to the QoS guarantee level of the AI service and/or QoS parameters related to computing power corresponding to the QoS guarantee level.

In the embodiment of the application, the QoS parameters related to the computing power corresponding to the QoS guarantee level at least comprise one or more of the following parameters of computing power resource type, floating point operation times GFPS executed by computing power guarantee per second, floating point operation times MFPS executed by computing power maximum per second and computing power priority.

In the embodiment of the present application, the SMF receives Authorized QoS parameters, and converts the received Authorized QoS parameters into access specific QOS parameters (english may be expressed as ACCESS SPECIFIC QOS parameters).

In the embodiment of the application, the function of converting QoS parameters related to AI service in SDP PARAMETERS into ACCESS SPECIFIC QoS parameters is added in SMF.

Specifically, the SMF is added with a function of converting QoS parameters related to AI service data transmission in authorized QoS parameters into ACCESS SPECIFIC QoS parameters.

In the embodiment of the present application, referring to the related art, according to a rule of deriving an authorized QoS coefficient for each QoS flow from authorized QoS parameters in SMF (english may be expressed as Rules for derivation of the Authorized QoS Parameters per QoS flow from the Authorized QoS Parameters in SMF),, for example, information of QoS parameters related to AI service data transmission may be added on the basis of the related art, as shown in the following table 20:

Table 20

In the embodiment of the present application, the SMF needs to further add a function of converting QoS parameters related to the AI service calculation task instance in authorized QoS parameters into ACCESS SPECIFIC QoSparameters.

In the embodiment of the application, referring to the related art, the information of QoS parameters related to the computing power of the AI service computing task is added on the basis of the related art according to Rules for derivation of theAuthorized QoS Parameters per QoS flow from the Authorized QoS Parameters inSMF,. The specific manner of adding the parameters may refer to the above table 20, and will not be described herein.

S203, the QoS parameter is sent to the device related to data transmission and/or the device related to computing power.

Wherein the QoS parameter is used for the equipment related to data transmission and/or the equipment related to computing power to carry out QoS guarantee on the AI service based on the QoS parameter.

In the embodiment of the present application, the device related to data transmission and/or the device related to computing power may be understood as a device for specifically performing data transmission of an AI service or a device for performing a computing task of an AI service, which may be a terminal, a base station, etc., specifically, may be selected according to an actual situation, and the embodiment of the present application is not limited specifically.

In the embodiment of the application, after receiving the corresponding QoS parameters, the related equipment of data transmission and/or the related equipment of computing power performs QoS guarantee on the AI service according to the specific QoS parameters.

It can be understood that, in the QoS securing method provided in the embodiment of the present application, since the first information transmitted by the second function to the first function corresponds to the data transmission related requirement of the AI service and the calculation power related requirement of the AI service, the QoS control policy sent by the first function received by the third function according to the first information and the AI service template also corresponds to the specific requirement of the AI service, and the QoS parameter finally generated corresponds to the specific requirement of the AI service, when the corresponding QoS parameter is adopted by the equipment related to the data transmission and/or the equipment related to the calculation power to execute the corresponding AI service, the specific requirement of different AI services on QoS can be satisfied.

In an embodiment of the present application, the third function may further receive an updated QoS control policy sent by the first function, generate an updated QoS parameter corresponding to the AI service based on the updated QoS control policy, and send the updated QoS parameter to the device related to data transmission and/or the device related to computing power, where the updated QoS parameter is used for QoS guarantee for the AI service by the device related to data transmission and/or the device related to computing power based on the updated QoS parameter.

In the embodiment of the application, after updating the QoS control strategy, the PCF sends the updated QoS control strategy to the SMF, and the SMF generates updated QoS parameters according to the obtained updated QoS control strategy.

It should be noted that, the manner of generating the updated QoS parameters according to the obtained updated QoS control policy may refer to the process of generating QoS parameters according to the QoS control policy in the foregoing embodiment, which is different from the QoS control policy, and the specific implementation process is not described herein.

In the embodiment of the application, after generating the updated QoS parameters, the SMF sends the QoS parameters to the execution device related to data transmission or the execution device related to AI computing task, and performs QoS guarantee on the execution device related to data transmission or the execution device related to AI computing task.

Based on the above embodiment, the embodiment of the present application further provides an overall flow diagram of QoS guarantee, as shown in fig. 4, mainly comprising the following steps:

1. After the Session is established/changed, the AF receives AF Session signaling (english may be expressed as AF Session signalling possibly with SID) possibly with SID, converts SDI received at AF Session signalling into service information by SDI mapping function (english may be expressed as SID MAPPING function) and delivers to PCF.

2. The PCF performs the Policy Engine (English may be expressed as Policy Engine) function, converts received service information to Authorized QoS parameters/PER SERVICE DATA flow, and the PCF merges Authorized QoS parameters/per flow for each direction and passes it to the SMF.

3. The SMF performs a streaming service management (english may be expressed as Flow SERVICE MANAGER) function, and the SMF converts the received Authorized QoS parameters into ACCESS SPECIFIC QoS parameters and transmits the parameters to the UE.

4. The SMF sends AISERVICE detection function rules to the UPF for classification of AI SERVICE for QoS flow marking and other operations. UPF performs the function of marking the QoS flow of a packet or performing other operations (english may be expressed as CLASSIFY PACKETS for QoS flow MARKING AND other actions).

5. QoS profile is assigned to the gNB by the SMF or predefined on the gNB.

6. QoS rules are assigned to UEs by SMF.

In fig. 4, SMF english is fully called Session Management Function, which is correspondingly interpreted as a session management function, PCF english is fully called Policy Control Function, which is correspondingly interpreted as a policy control function, AF english is fully called Application Function, which is correspondingly interpreted as an application function, UPF english is fully called User Plane Function, which is correspondingly interpreted as a user plane function, UE represents a user equipment, and gNB represents a next generation base station.

In fig. 4, the gNB performs scheduling resources (english may be expressed as Schedule resources for PDU session) for PDU sessions, and the UE performs mapping UL packets to QoS flows and applying QoS flow flags (english may be expressed as mapping UL packets to QoS flows and apply QoS flow marking).

The QoS profile (english may be expressed as profile) transmitted between SMF to gNB needs to be enhanced based on the QoS profile in the related art:

The QoS profile may be served by the SMF to the gNB or may be predefined on the gNB. 5QI and QoS parameter values related to connection resources associated with the corresponding scene of the AI service are preconfigured in AN AN network element of the network.

In the related art, 5QI corresponding to different scenes of the Internet of vehicles is defined in QoS profile, and can be used for guaranteeing V2X messages (such as advanced driving: crashproof (English can be expressed as ADVANCED DRIVING: collision Avoidance)). As shown in table 21 below.

Table 21

In the embodiment of the present application, 5QI related to computing power resources associated with a scenario corresponding to AI services and QoS parameter values related to the computing power resources are preconfigured in AN network element of a network, as shown in table 22 below, table 22 shows QoS profiles after enhancement.

Table 22

In the embodiment of the application, qoS guarantee is needed for the AI service, and the determination and deployment of the AI task or AI subtask deployment scheme corresponding to the AI service are needed.

In the embodiment of the present application, an overall flow diagram for AI service processing is provided, as shown in fig. 5, and mainly includes the following steps:

1. AI services subscription and translation, among others:

1) The user subscribes to the AI service;

2) Making an AI service template;

3) Querying an AI service template;

4) The AI service template-AI service number is transmitted.

2. An AI service orchestration process, comprising:

1) AI task disassembly scheme set;

2) Making an AI subtask template;

3) And transmitting the solution set corresponding to the AI service.

3. The optimal solution selection and AI task deployment scheme generation comprise the following steps:

1) Selecting an optimal solution;

2) Generating an AI task deployment scheme;

3) And transmitting an AI task deployment scheme.

4. AI task execution control, which includes the following:

1) AI task deployment;

2) AI task lifecycle management;

3) QoS guarantee.

5. QoS information monitoring and dynamic adjustment of optimal solutions.

In the embodiment of the application, the AI service subscription and translation can be carried out on the AI service platform, after the AI service template is formulated, the AI service user sends an AI service subscription request to the AI service platform at the network side, and after the AI service platform receives the AI service subscription request, the AI service platform can translate according to the AI service template to obtain the AI service case number corresponding to the AI service subscription request. Furthermore, the AI service platform completes AI service subscription and sends a subscription success request to the subscriber. Wherein the subscriber corresponds to the user who sends the AI service subscription request.

In the embodiment of the present application, the AI service templates may include a traffic template for AI service and an AI service template for AI service QoS, which may refer to tables 10 to 12 in the foregoing embodiments.

In the embodiment of the application, the AI service business template comprises, but is not limited to, the following:

1) And the service triggering mode is to start the related business flow of the AI service when the triggering condition is met.

A. The triggering mode can be event triggering, including but not limited to positioning detection, weather detection, accident detection and the like.

B. The triggering mode may be a switch triggering, for example, the UE triggers to start a service flow by clicking an AI service interface.

2) Service network scope-after the AI service business flow is triggered, the network scope of the service is provided for the user.

3) AI task ID/description, functional module required to complete AI services.

4) The input data required for the service.

5) Model selection and model source.

6) Model tasks (model selection, model training, model verification optimization, model reasoning, etc.).

7) Conclusions/decisions of model output, etc.

8) The AI service case number can uniquely represent the AI service subscribed by the user, and can correspond to the requirement of the AI service on the first network resource, thereby playing a role in translating the AI service subscribed by the user into the requirement on the network resource. For example, the AI service case number corresponds to a specific AI service, and the network resource requirement of the specific AI service may be correspondingly determined through the AI service case number, where each AI service case number corresponds to the first network resource requirement.

9) AI service scenario ID the network (e.g., AI service platform) pre-numbers according to the AI service application scenarios available.

10 AI service sub-scene ID the network (e.g., AI service platform) can manage and number according to scene layering, which is beneficial to expanding more scenes of AI service available by the network.

For example, if the internet of vehicles-intelligent collision detection in table 23 is taken as an example, after determining that the AI service template ID is 1, the network resource requirements required for the corresponding AI service when the ID is 1 may be correspondingly determined, as shown in the following table 24.

Table 23

Table 24

In the embodiment of the application, the AI service arrangement processing may specifically be splitting processing of one or more AI tasks in AI tasks included in the AI service, for example, splitting between AI function modules (i.e. data acquisition, model reasoning, model training, model selection, model updating, etc.), or splitting within an AI function module (i.e. the same AI model may be split and then executed by different network elements), so as to obtain multiple combined AI subtasks, and determining a set composed of the multiple combined AI subtasks and the multiple AI tasks as the first set.

In the embodiment of the application, based on the AI subtask template, the processing mode corresponding to the AI subtask and/or the AI task in the first set is determined, and the processing mode corresponding to the AI subtask and/or the AI task is determined as the second set.

In the embodiment of the application, a third set is obtained by using the first set and the second set, and the third set is determined as a solution set corresponding to the AI service.

In the embodiment of the application, the SMO can carry out service arrangement processing on the AI service to obtain a solution set for processing the AI service.

In the embodiment of the application, the first set is all possible combination schemes obtained after the AI task is split.

In the embodiment of the present application, the second set is a processing manner corresponding to an AI task or an AI subtask.

In the embodiment of the application, when the AI task is split to obtain different AI subtasks, the split processing method includes but is not limited to the following split processing modes:

1) AI function module splitting, such as data collection, model reasoning, model training, model selection, model updating, etc.

2) The splitting within the AI function module, for example, the same AI model may be performed by different network elements after splitting.

In the embodiment of the application, the SMO splits one or more AI tasks in a plurality of AI tasks in an arbitrary splitting manner to obtain a plurality of combined AI subtasks, for example, splits AI task 1 to obtain AI task 1_ai subtask 1 and AI task 1_ai subtask 2, and does not split AI task 2, so AI task 1_ai subtask 1, AI task 1_ai subtask 2 and AI task 2 are referred to as a combination.

In the embodiment of the present application, the first set includes a plurality of combined AI subtasks and a plurality of AI tasks, as shown in table 25 below.

A first set of different splits of AI tasks is shown in table 25, where only three combinations are shown in the first set, but the manner of splitting AI tasks is not limited to the following three.

Table 25

In the embodiment of the application, AI subtask templates of various AI tasks are designed. For example, the data acquisition subtask template of the AI task comprises a data source, a data consumer, a data type, a data task, a data granularity, a data volume, a data processing method and the like and specific description, wherein the Near-RT RIC used on the network side extracts effective information from the data acquisition subtask template according to the data acquisition task for deployment, and the Near-RT RIC performs communication connection resource allocation and guarantee according to the data type, the data task (application) and the like.

The embodiment of the application provides a data acquisition subtask template, which is shown in the following table 26:

Table 26

In the embodiment of the application, the AI subtask templates of the various AI tasks also comprise AI model task templates, and the AI model task templates comprise one or more of AI task types, AI models and AI model management related configuration information.

The AI model task template comprises one or more of AI task types, AI models and AI model management related configuration information, wherein the configuration information is used for extracting effective information from the AI model task template to deploy AI tasks and manage life cycles, and the Near-RT RIC is used for distributing and guaranteeing communication resources and computing resources according to the AI task types, the AI models and other related information.

In an embodiment of the present application, the AI model task templates are shown in table 27 below:

Table 27

In the embodiment of the present application, the third set obtained by combining the above tables 25 to 27 is determined as a solution set for solving the AI service.

Illustratively, the set of solutions corresponding to AI services for internet of vehicles-intelligent collision prediction may be:

The AI service of the vehicle networking-intelligent collision prediction is completed by a decision algorithm of AI task 1-service vehicle track prediction and AI task 2-vehicle collision early warning, and each task comprises processes of data acquisition, model reasoning and the like.

The network element SMO in the network may generate the AI service solution set by performing a service orchestration process on the AI services. A number of solutions may be included in the solution set, two of which may be:

In the solution 1, the AI task 1 is divided into 2 subtasks, and the AI service is completed by the AI task 1-subtask 1, the AI task 1-subtask 2 and the AI task 2.

The AI task 1-subtask 1 is a part of the service vehicle track prediction, for example, 3 layers HIDDEN LAYER are arranged in CNN, and the reasoning task of the previous 2 layers belongs to the AI task 1-subtask 1.

AI task 1-subtask 2 is another part of the service vehicle trajectory prediction, e.g., HIDDEN LAYER in CNN has 3 layers, and the reasoning task of the last 1 layer belongs to AI task 1-subtask 2.

And the AI task 2 predicts the collision early warning decision algorithm for the vehicle.

Solution 1 also contains a corresponding AI model task template, as shown in table 28 below:

table 28

It should be noted that, if data acquisition is involved in the solution 1, reference may be made to the data acquisition subtask templates in the foregoing table 26, which is not exemplified here.

Solution 2 in solution 2 AI services are completed by AI task 1 together with AI task 2.

Solution 2 also contains a corresponding AI model task template, as shown in table 29 below:

Table 29

It should be noted that, in the case of the solution 2, if data acquisition is involved, reference may be made to the data acquisition subtask templates in the foregoing table 26, which is not illustrated here.

In the embodiment of the application, the optimal solution selection and AI task deployment scheme generation specifically comprise:

The SMO at the network side sends the generated solution set for solving the AI service to the Near-RT RIC to realize the management arrangement of the network resources. When selecting the optimal one from the solution set, the demands on network resources will also be different due to the different AI service solutions. For example, when the AI task 1-subtask 1 and the AI task 1-subtask 2 are not in the same network element, the model middle layer data and the model parameters need to be transmitted between the network elements. When the AI task 1 and the AI task 2 are not in the same network element, the output data of the predicted track of the vehicle needs to be transmitted between the network elements. Thus, different AI service solutions correspond to different network resource requirements (e.g., computing resources, connection resources, etc.), which may be computing resources, connection resources, etc.

Thus, after the SMO sends the solution set to the Near-RT RIC, the Near-RT RIC may determine network resource requirements corresponding to AI service solutions from among different AI service solutions, thereby further determining an optimal one of the solution sets to solve AI services from among the solution set according to the network resource requirements corresponding to each AI service solution.

In the embodiment of the application, when the optimal solution is selected from the solution set, aiming at each solution in the solution set, according to the information in the AI model task template contained in each solution, the second network resource required by each solution is estimated, the third network resource at the current moment is acquired, the solution with the second network resource being most similar to the third network resource is determined from the solution set, and the solution with the second network resource being most similar to the third network resource is determined as the optimal solution.

In the embodiment of the application, as it is possible that the third network resource is not satisfied for all solutions, for example, in the case of extremely resource-intensive situations, a closest/most suitable/most matching/most similar solution is selected.

In an embodiment of the application, the second network resource and the third network resource comprise one or more of a computing resource and a connection resource.

In the embodiment of the application, the AI service platform sends the formulated AI service solution set to the intelligent network controller (namely Near-RT RIC) to realize resource management arrangement, namely, according to network perceived network resources (namely the acquired third network resources at the current moment), the solution which is most suitable for the current real-time resource condition is selected through an optimal algorithm, and the deployment scheme of AI tasks is further generated.

Wherein the network aware third network resource or the second network resource may comprise:

1) Connection resources of AI service corresponding coverage:

a cell PRB utilization rate situation;

b cell radio channel conditions;

c cell connection load situation.

2) AI services correspond to the computational power resources of the coverage:

a hardware structure, such as GPU, CPU and the like;

b, calculating the force;

c data storage space.

The solution which best accords with the current real-time resource condition is selected based on cross-domain collaborative management arrangement of multi-objective optimization algorithms such as data, calculation power, connection, AI model and the like. The solution set can obtain the requirements of different solutions on network resources in advance through information such as AI task types, AI models and the like, including limited computing power and connection resource requirements on a network, and can map complex AI service arrangement into matching and selection of the network resource requirements and real-time perception resources.

In the embodiment of the application, the deployment scheme of the AI task refers to network elements distributed with specific deployment such as data acquisition, AI model tasks and the like.

By way of example, taking the selection of an AI service optimal solution for vehicle networking-intelligent collision prediction and the generation of an AI task deployment scheme as an example, a resource management arrangement layer of a network determines that a service vehicle has the reasoning data and calculation power of an AI task 1 according to perceived network resources, the transmission channel quality of the service vehicle and a main service cell can meet the real-time performance of AI task 1 reasoning result transmission and task 2 decision algorithm issuing, and the main service cell has enough calculation power, storage and other environments to support the AI task 2. Thus, solution 3 in table 21 is selected and a deployment scenario for the AI task is generated, as shown in table 30 below.

Table 30

In the embodiment of the application, after the resource management arrangement layer (which may be Near-RT RIC) of the network formulates the deployment scheme of the AI task, the lifecycle flow of the AI service and the lifecycle of the AI task in the AI service solution can be executed, and the parameters for QoS guarantee are determined for the AI service, so that the QoS guarantee is performed for the AI service. The QoS securing method may refer to the foregoing embodiments, and will not be described herein.

In the embodiment of the application, qoS guarantee is carried out on the AI service, meanwhile, qoS parameters of the AI service are required to be monitored, when the QoS parameters do not meet the requirements of the AI service, the determination of the QoS parameters is carried out again or when the QoS information does not meet the requirements of the AI service, the optimal solution corresponding to the third network resource is reselected from the solution set, and the updated AI task deployment scheme is determined based on the optimal solution corresponding to the third network resource.

Based on the above embodiment, in another embodiment of the present application, there is provided a first function 1, as shown in fig. 6, the first function 1 including:

The first receiving unit 10 is configured to receive first information sent by the second function, where the first information is used to describe a requirement related to AI-service data transmission and/or a requirement related to AI-service computing power.

The first processing unit 11 is configured to generate a QoS control policy of the AI service based on the first information and an AI service template corresponding to the subscribed AI service.

A first transmitting unit 12, configured to transmit a QoS control policy to the third function, where the QoS control policy is used by the third function to generate QoS parameters corresponding to the AI service based on the QoS control policy.

In an embodiment, the first function 1 may further comprise a monitoring unit, an acquisition unit.

And the monitoring unit is used for monitoring QoS information corresponding to the AI service.

And the acquisition unit is used for acquiring network connection resources and computing power resources at the current moment when the QoS information corresponding to the AI service is monitored to not meet the requirement.

The first sending unit 12 is further configured to update a QoS control policy of the AI service based on the network connection resource and the computing resource, and send the updated QoS control policy to the third function, where the updated QoS control policy is used for the third function to generate an updated QoS parameter corresponding to the AI service based on the updated QoS control policy.

In one embodiment, the AI service template includes at least one or more parameters including QoS parameters related to data transmission corresponding to QoS guarantee levels of the AI service and/or QoS parameters related to computational power corresponding to QoS guarantee levels.

In one embodiment, the QoS parameters associated with the computing power corresponding to the QoS guarantee level include at least one or more of computing power resource type, floating point number of operations performed per second GFPS for computing power guarantee, floating point number of operations performed per second MFPS for computing power maximum, computing power priority.

The embodiment of the application provides a first function, which receives first information sent by a second function, wherein the first information is used for describing requirements related to AI service data transmission and/or requirements related to AI service computing power; generating QoS control strategy of AI service based on the first information and AI service template corresponding to the subscribed AI service; the QoS control policy is sent to the third function, where the QoS control policy is used for the third function to generate QoS parameters corresponding to the AI service based on the QoS control policy, so that it can be seen that after the second function transmits the first information to the first function, the first function generates the QoS control policy of the AI service corresponding to the corresponding AI service requirement according to the first information and the AI service template, and the QoS parameters finally generated by the third function also correspond to the specific requirement of the AI service, so that when the corresponding AI service is executed by the equipment related to data transmission and/or the equipment related to the computing power by adopting the corresponding QoS parameters, the specific requirement of different AI services on QoS can be met.

Fig. 7 is a schematic diagram of the composition structure of a first function 1 according to an embodiment of the present application, and in practical application, based on the same disclosure concept of the above embodiment, as shown in fig. 7, the first function 1 according to an embodiment of the present application includes a first processor 13, a first memory 14, and a first communication bus 15.

In a specific embodiment, the first receiving unit 10, the first processing unit 11, the first transmitting unit 12, the monitoring unit, and the acquiring unit may be implemented by a first Processor 13 located on the first function 1, where the first Processor 13 may be at least one of an ASIC (Application SPECIFIC INTEGRATED Circuit), a digital signal Processor (DSP, digital Signal Processor), a digital signal processing image processing device (DSPD, digital Signal Processing Device), a programmable logic image processing device (PLD, programmable Logic Device), a field programmable gate array (FPGA, field Programmable GATE ARRAY), a CPU, a controller, a microcontroller, and a microprocessor. It will be appreciated that the electronics for implementing the above-described processor functions may be other for different devices, and embodiments of the present application are not particularly limited.

In the embodiment of the present application, the first communication bus 15 is used to implement connection communication between the first processor 13 and the first memory 14, and the first processor 13 implements the following QoS securing method when executing the running program stored in the first memory 14:

The method comprises the steps of receiving first information sent by a second function, wherein the first information is used for describing requirements related to AI service data transmission and/or requirements related to AI service computing power, generating QoS control strategies of AI services based on the first information and AI service templates corresponding to subscribed AI services, sending the QoS control strategies to a third function, and enabling the third function to generate QoS parameters corresponding to the AI services based on the QoS control strategies.

In an embodiment, the first processor 13 is further configured to monitor QoS information corresponding to an AI service, acquire network connection resources and computing resources at a current moment when it is monitored that the QoS information corresponding to the AI service does not meet a requirement, update a QoS control policy of the AI service based on the network connection resources and the computing resources, and send the updated QoS control policy to the third function, where the updated QoS control policy is used by the third function to generate an updated QoS parameter corresponding to the AI service based on the updated QoS control policy.

In one embodiment, the AI service template includes at least one or more parameters including QoS parameters related to data transmission corresponding to QoS guarantee levels of the AI service and/or QoS parameters related to computational power corresponding to QoS guarantee levels.

In one embodiment, the QoS parameters associated with the computing power corresponding to the QoS guarantee level include at least one or more of computing power resource type, floating point number of operations performed per second GFPS for computing power guarantee, floating point number of operations performed per second MFPS for computing power maximum, computing power priority.

Based on the above embodiment, in another embodiment of the present application, a third function 2 is provided, as shown in fig. 8, the third function 2 includes:

The second receiving unit 20 is configured to receive a QoS control policy sent by the first function, where the QoS control policy is generated by the first function based on the first information sent by the second function and an AI service template corresponding to the subscribed AI service.

The second processing unit 21 is configured to generate QoS parameters corresponding to the AI service based on the QoS control policy.

A second sending unit 22, configured to send QoS parameters to a device related to data transmission and/or a device related to computing power, where the QoS parameters are used by the device related to data transmission and/or the device related to computing power to QoS guarantee the AI service based on the QoS parameters.

In an embodiment, the second receiving unit 20 is further configured to receive the updated QoS control policy sent by the first function.

The second processing unit 21 is further configured to generate updated QoS parameters corresponding to the AI service based on the updated QoS control policy.

The second sending unit 22 is further configured to send the updated QoS parameter to the device related to data transmission and/or the device related to computing power, where the updated QoS parameter is used for QoS guarantee of the AI service by the device related to data transmission and/or the device related to computing power based on the updated QoS parameter.

In an embodiment, the QoS parameters include QoS parameters related to data transmission corresponding to QoS security levels of AI services and/or QoS parameters related to computing power corresponding to QoS security levels.

In one embodiment, the QoS parameters related to the computing power corresponding to the QoS guarantee level at least comprise one or more of computing power resource type, floating point operation times GFPS executed by computing power guarantee per second, floating point operation times MFPS executed by computing power maximum per second and computing power priority.

The third function is provided by the embodiment of the application, the QoS control strategy sent by the first function is received, the QoS control strategy is generated by the first function based on the first information sent by the second function and the AI service template corresponding to the contracted AI service, the QoS parameter corresponding to the AI service is generated based on the QoS control strategy, the QoS parameter is sent to equipment related to data transmission and/or equipment related to computing power, and the QoS parameter is used for carrying out QoS guarantee on the AI service based on the QoS parameter, so that the third function provided by the embodiment of the application can realize different QoS requirements on the AI service when the equipment related to data transmission and/or equipment related to computing power adopt the same QoS parameters.

Fig. 9 is a schematic diagram of the composition structure of a third function 2 according to an embodiment of the present application, and in practical application, based on the same disclosure concept of the above embodiment, as shown in fig. 9, the third function 2 according to an embodiment of the present application includes a second processor 23, a second memory 24, and a second communication bus 25.

In a specific embodiment, the second receiving unit 20, the second processing unit 21, and the second transmitting unit 22 may be implemented by a second processor 23 located on the third function 2, where the second processor 23 may be at least one of ASIC, DSP, DSPD, PLD, FPGA, CPU, a controller, a microcontroller, and a microprocessor. It will be appreciated that the electronics for implementing the above-described processor functions may be other for different devices, and embodiments of the present application are not particularly limited.

In the embodiment of the present application, the second communication bus 25 is used to implement connection communication between the second processor 23 and the second memory 24, and the second processor 23 implements the following QoS securing method when executing the running program stored in the second memory 24:

the method comprises the steps of receiving a QoS control strategy sent by a first function, wherein the QoS control strategy is generated by the first function based on first information sent by a second function and an AI service template corresponding to a subscribed AI service, generating a QoS parameter corresponding to the AI service based on the QoS control strategy, sending the QoS parameter to equipment related to data transmission and/or equipment related to computing power, and ensuring QoS of the AI service by the QoS parameter for the equipment related to the data transmission and/or the equipment related to computing power based on the QoS parameter.

In an embodiment, the second processor 23 is further configured to receive an updated QoS control policy sent by the first function, generate an updated QoS parameter corresponding to the AI service based on the updated QoS control policy, and send the updated QoS parameter to the device related to data transmission and/or the device related to computing power, where the updated QoS parameter is used for QoS guarantee of the AI service by the device related to data transmission and/or the device related to computing power based on the updated QoS parameter.

In an embodiment, the QoS parameters include QoS parameters related to data transmission corresponding to QoS security levels of AI services and/or QoS parameters related to computing power corresponding to QoS security levels.

In one embodiment, the QoS parameters related to the computing power corresponding to the QoS guarantee level at least comprise one or more of computing power resource type, floating point operation times GFPS executed by computing power guarantee per second, floating point operation times MFPS executed by computing power maximum per second and computing power priority.

Based on the above embodiments, the embodiments of the present application provide a storage medium having stored thereon a computer program, the computer readable storage medium storing one or more programs executable by one or more processors and applied to a first function/a third function, the computer program implementing the QoS securing method as described above.

Based on the above embodiments, the embodiments of the present application provide a computer program product, including a computer program, which can be executed by one or more processors and applied to the first function/third function, where the computer program implements the QoS securing method as described above.

It should be noted that, in embodiments of the present application, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the embodiments of the present application may be embodied essentially or in a part contributing to the related art in the form of a software product stored in a storage medium (such as ROM/RAM, a magnetic disk, an optical disk), including several instructions for causing an image display device (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to perform the method described in the embodiments of the present application.

The foregoing is merely illustrative of embodiments of the present application, and the scope of the present application is not limited thereto, and any person skilled in the art can easily think about variations or substitutions within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims (14)

1. A method for ensuring quality of service (QoS), characterized in that it is applied to a first function, the method comprising: receiving first information sent by the second function, where the first information is used to describe requirements related to artificial intelligence (AI) service data transmission and/or requirements related to AI service computing power; Generate a QoS control policy for the AI service based on the first information and an AI service template corresponding to the contracted AI service; The QoS control policy is sent to a third function, where the third function generates QoS parameters corresponding to the AI service based on the QoS control policy. 2. The method according to claim 1, wherein after sending the QoS control policy to the third function, the method further comprises: Monitor QoS information corresponding to the AI service; When it is detected that the QoS information corresponding to the AI service does not meet the requirements, obtain the current network connection resources and computing power resources; Based on the network connection resources and the computing power resources, the QoS control policy of the AI service is updated, and the updated QoS control policy is sent to the third function, and the updated QoS control policy is used for the third function to generate updated QoS parameters corresponding to the AI service based on the updated QoS control policy. 3. The method according to claim 1 is characterized in that the AI service template includes at least one or more of the following parameters: QoS parameters related to data transmission corresponding to the QoS guarantee level of the AI service and/or QoS parameters related to computing power corresponding to the QoS guarantee level. 4. The method according to claim 3 is characterized in that the QoS parameters related to the computing power corresponding to the QoS guarantee level include at least one or more of the following parameters: computing power resource type, the number of floating-point operations GFPS performed per second by the computing power, the maximum number of floating-point operations MFPS performed per second by the computing power, and computing power priority. 5. A QoS guarantee method, characterized in that it is applied to the third function, the method comprising: receiving a QoS control policy sent by the first function, where the QoS control policy is generated by the first function based on the first information sent by the second function and an AI service template corresponding to the subscribed AI service; Based on the QoS control policy, generate QoS parameters corresponding to the AI service; The QoS parameters are sent to devices related to data transmission and/or devices related to computing power. The QoS parameters are used by the devices related to data transmission and/or devices related to computing power to provide QoS guarantee for AI services based on the QoS parameters. 6. The method according to claim 5, further comprising: receiving an updated QoS control policy sent by the first function; Based on the updated QoS control policy, generate updated QoS parameters corresponding to the AI service; The updated QoS parameters are sent to the devices related to the data transmission and/or the devices related to the computing power. The updated QoS parameters are used by the devices related to the data transmission and/or the devices related to the computing power to provide QoS guarantee for the AI service based on the updated QoS parameters. 7. The method according to claim 5 or 6 is characterized in that the QoS parameters include QoS parameters related to data transmission corresponding to the QoS guarantee level of the AI service and/or QoS parameters related to computing power corresponding to the QoS guarantee level. 8. The method according to claim 7 is characterized in that the QoS parameters related to the computing power corresponding to the QoS guarantee level include at least one or more of the following parameters: computing power resource type, the number of floating-point operations GFPS performed per second by the computing power, the maximum number of floating-point operations MFPS performed per second by the computing power, and computing power priority. 9. A first function, characterized in that the first function includes: A first receiving unit, configured to receive first information sent by the second function, where the first information is used to describe requirements related to AI service data transmission and/or requirements related to AI service computing power; A first processing unit, configured to generate a QoS control policy for the AI service based on the first information and an AI service template corresponding to the subscribed AI service; The first sending unit is configured to send the QoS control policy to a third function, where the QoS control policy is used by the third function to generate QoS parameters corresponding to the AI service based on the QoS control policy. 10. A third function, characterized in that the third function includes: A second receiving unit is configured to receive a QoS control policy sent by the first function, where the QoS control policy is generated by the first function based on the first information sent by the second function and an AI service template corresponding to the subscribed AI service; A second processing unit, configured to generate QoS parameters corresponding to the AI service based on the QoS control policy; The second sending unit is used to send the QoS parameters to devices related to data transmission and/or devices related to computing power, and the QoS parameters are used by the devices related to data transmission and/or devices related to computing power to provide QoS guarantee for AI services based on the QoS parameters. 11. A first function, characterized in that the first function comprises: a first processor and a first memory; when the first processor executes the running program stored in the first memory, the method according to any one of claims 1 to 4 is implemented. 12. A second function, characterized in that the second function comprises: a second processor and a second memory; when the second processor executes the running program stored in the second memory, the method according to any one of claims 5 to 8 is implemented. 13. A storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented, or when the computer program is executed by a processor, the method according to any one of claims 5 to 8 is implemented. 14. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements the method according to any one of claims 1 to 4, or the computer program implements the method according to any one of claims 5 to 8 when the computer program is executed by a processor.