Heterogeneous resource cooperative scheduling method for training AI reasoning service

Abstract

The invention discloses a heterogeneous resource collaborative scheduling method for training an AI inference service, which comprises the following steps of: establishing a mixed integer linear programming model MILP based on an AI inference collaborative heterogeneous resource scheduling problem trained by a distributed model, and maximizing the number of successfully sent AI inference sub-tasks by the final goal of the proposed MILP, the utilization rate of heterogeneous resources in the NG-RAN and the backlog of the AI reasoning service are minimized; constructing an edge computing resource request, computing processing capability, a task backlog threshold value of an AI reasoning sub-task on an edge server and a complaint rate threshold value of the AI reasoning sub-task as constraint conditions; the deployment of the DU-CU in the network architecture is standardized; adjusting wavelength allocation in the network; and the capacity requirement of the network is specified. According to the method, the number of successfully sent AI inference service subtasks can be maximized, and meanwhile, the heterogeneous resource occupancy rate and the AI inference service subtask backlog amount are minimized.

Description

Technical field:

[0001] The present invention relates to the field of mobile communication technologies, and in particular to a heterogeneous resource collaborative scheduling method for training AI reasoning services. Background technology:

[0002] The next generation (NG) of mobile communications is coming, which will lead us into an unprecedented era with higher data transmission rates, shorter round-trip latency, more diverse service scenarios, and combined with advanced technologies such as artificial intelligence (AI) and machine learning (ML). These heterogeneous requirements have posed new challenges to the capacity, latency, and network flexibility of the radio access network (RAN), accelerating the evolution of RAN to the next generation radio access network (NG-RAN). In the NG-RAN architecture, the baseband unit (BBU)-active antenna unit (AAU) is redefined as a three-layer network, including a distributed processing unit (DU), a central processing unit (CU), and an AAU. Delay-sensitive functions below the packet data convergence protocol (PDCP) layer are handled by the DU, while functions above the PDCP layer are managed by the CU. The connection between the DU and the CU is called the mid-haul network, which is usually mediated by an optical network.

[0003] AI reasoning is the process of applying trained AI models to new data for analysis and decision-making, and it has become a core technology in multiple industries. With the advancement of computing power and big data technology, the application scope of AI reasoning has expanded to medical diagnosis, financial risk assessment, autonomous driving, intelligent manufacturing and other fields. At present, the development direction of AI reasoning is mainly focused on real-time processing and low latency, model compression and optimization, multimodal reasoning and other aspects. Despite significant progress, AI reasoning still faces challenges such as high computing resource requirements and model adaptability. AI reasoning based on distributed model training has significant performance advantages, including faster processing speed and higher scalability, thanks to parallel computing and resource optimization across multiple nodes.

[0004] NG-RAN has higher spectrum efficiency and scalability, can handle more data traffic, and support a wider range of AI inference services. In addition, optical networks have excellent performance with high bandwidth and low latency. RAN networks have abundant edge computing resources that can be used for AI inference model training. The converged optical network and NG-RAN are considered to be efficient networks for carrying AI inference services. Therefore, we propose a new network architecture based on edge computing network (ECFN) for distributed model training for AI inference.

[0005] However, optimal utilization and quality of service (QoS) of heterogeneous resources for AI inference services are crucial. Therefore, we focus on efficient heterogeneous resource collaborative scheduling strategies to support the widespread application of AI inference based on distributed model training while meeting the high quality of service for AI inference.

[0006] With the widespread application of AI reasoning in various fields, the demand for computing resources and bandwidth resources continues to increase. Especially in NG-RAN supported by ECFN, as the network architecture evolves into DU and CU, how to effectively manage and schedule these heterogeneous resources becomes a key challenge. At present, many resource scheduling methods cannot achieve the coordinated allocation of heterogeneous resources, resulting in improper resource allocation, which in turn causes waste of computing and transmission resources. In addition, whether the AI ​​reasoning subtask is successfully sent will also affect the user's QoS, resulting in task backlogs and affecting the quality of service. Therefore, the number of successfully sent AI reasoning subtasks should also be considered. Therefore, it is urgent to explore the heterogeneous resource collaborative scheduling strategy based on NG-RAN supported by ECFN to improve the heterogeneous resource efficiency of AI reasoning services. Summary of the invention:

[0007] The purpose of the present invention is to provide a heterogeneous resource collaborative scheduling method for AI reasoning business based on distributed model training to address the deficiencies of the prior art.

[0008] The present invention is implemented by the following technical solution: a method for coordinating heterogeneous resources for training AI reasoning services, comprising the following steps:

[0009] A mixed integer linear programming model MILP is established for the AI ​​inference collaborative heterogeneous resource scheduling problem based on distributed model training. The ultimate goal of the proposed MILP is to maximize the number of successfully sent AI inference subtasks and minimize the utilization rate of heterogeneous resources and the backlog of AI inference services in NG-RAN.

[0010] Construct edge computing resource requests, computing processing capabilities, task backlog thresholds for AI reasoning subtasks on edge servers, and complaint rate thresholds for AI reasoning subtasks as constraints;

[0011] Standardize the deployment of DU-CU in the network architecture, ensure that the baseband processing function of each user service is deployed on only one GPP node, and specify the locations of RU, DU-CU and MEC;

[0012] Adjust the wavelength allocation in the network, establish the necessary optical links from the access AAU to the MEC through the selected GPP nodes, and establish constraints to ensure the continuity of each service request in the service flow and prevent the formation of loop paths in the network;

[0013] Specify the capacity requirements of the network, including limiting the sum of processing resources required for all services mapped to a GPP node to not exceed its available capacity, specifying that the links between required endpoints must have sufficient bandwidth to satisfy service requests, specifying that the sum of processing resources required for all services mapped to the rth node must not exceed its available capacity, specifying that the edge data center m has sufficient capacity to satisfy all service requests, and specifying that each AI reasoning subtask should meet the total latency requirements.

[0014] Furthermore, it also includes a cooperative heterogeneous resource scheduling algorithm based on a greedy strategy for resource allocation, specifically:

[0015] Sort the computing power requirements in descending order to get the first list;

[0016] Sort the remaining edge computing resources of the edge data center in ascending order to obtain a second list;

[0017] Build the remaining edge computing resources of the edge data center;

[0018] Allocate computing power requests to edge data centers;

[0019] Calculate the usage of computing resources, the number of successfully matched AI reasoning subtasks, the backlog of AI reasoning subtasks, and the joint optimization results of the three.

[0020] Furthermore, the computing power request is allocated to the edge data center, specifically:

[0021] Search all users in descending order of computing power requirements. If the computing resource request is positive, and the computing capacity of each edge data center meets the computing resources requested by each user, and the data processing capacity complaint rate of all users of the jth edge data center allocated to the i-th computing power provider meets D i,j , then the f-th AI reasoning service of the k-th user is assigned to the j-th edge data center of the ith computing power provider;

[0022] C i,j ′-B k,f Assign the value of C i,j ′, if C i,j There is not enough edge computing power in B k,f The data processing capacity complaint rate of all users of the jth edge data center assigned to the i-th computing power provider satisfies D i,j , then C on the jth edge data center of the i-th computing power provider i,j ′The fth AI reasoning service assigned to the kth user;

[0023] 0 is assigned to C i,j ′,(B k,f-C i,j ′) is assigned to B k,f , until all B k,f All computing power requests are allocated to the edge data center, and the algorithm ends.

[0024] Furthermore, the ultimate goal of the MILP is to maximize the number of successfully sent AI reasoning subtasks and minimize the utilization of heterogeneous resources in NG-RAN and the backlog of AI reasoning services. The specific formula is expressed as follows:

[0025]

[0026] To minimize the resource usage of PP nodes, links and edge data centers in NG-RAN; V k,f,s,r is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user passes through the r-th node after being processed by the s-th baseband function, it is 1; The computation and processing requirements of the sth segment of baseband processing function serving the fth AI inference service for the kth user, U r is the computing power processing capacity of the rth node; is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user is processed by the s-th baseband function and occupies the w-th wavelength of the link (m, n), then it is 1; τ k,f,s The bandwidth requirement for the f-th AI reasoning service of the k-th user after being processed by the s-th baseband function; U l is the bandwidth capacity of the lth link; m k,f is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user uses the edge data center of the m-th center, it is 1; η k,f The computing power required to serve the fth AI reasoning service for the kth user; U m is the computing power processing capacity of the mth edge data center;

[0027] To maximize the number of successfully sent AI reasoning subtasks; where S i,j The computing power matching success rate of the jth edge data center of the i-th computing power provider; The computing power requirement (FLOPS / s) of the k-th user’s f-th AI reasoning service request allocated to the j-th edge data center of the i-th computing power provider;

[0028] To minimize the backlog of AI reasoning services; where σ i,j The AI ​​inference service backlog of the jth edge data center of the i-th edge computing provider (FLOPS / s); T is a task computing processing cycle;

[0029] α, β, and γ are weight coefficients.

[0030] Furthermore, the edge computing resource request, computing processing capacity, task backlog threshold of AI reasoning subtasks on edge servers, and complaint rate threshold of AI reasoning subtasks are constructed as constraints, specifically:

[0031]

[0032] Among them, in formula (2), The computing power requirement (FLOPS / s) allocated to the jth edge data center of the ith computing power provider for the fth AI reasoning service request of the kth user, B k,f is the f-th AI reasoning service request of the k-th user (FLOPS / s); in formula (3), C i,j is the edge computing processing capacity of the jth edge data center of the i-th computing power provider (FLOPS / s); in formula (4), ε is the backlog threshold of the AI ​​reasoning subtask of the edge data center (FLOPS); in formula (5), D i,j is the data processing capacity complaint rate (statistical value) of the jth edge data center of the i-th computing power provider, J i,j The remaining computing power resources (FLOPS) that can be claimed by the j-th edge data center of the i-th computing power provider.

[0033] Furthermore, the deployment of the DU-CU in the network architecture is specified to ensure that the baseband processing function of each user service is deployed on only one GPP node, and to specify the locations of the RU, DU-CU and MEC, specifically:

[0034]

[0035]

[0036] ∑ k,f,r Y k,f,r ≤∑ r,p G r,p ≤∑xF r ≤N r (8)

[0037]

[0038] These constraints also ensure the feasibility of photoelectric conversion and GPP node activation; where, in formula (6), O k,f,s,r,p is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user is deployed on the GPPp on the r-th node after the s-th baseband processing, it is 1; in formula (7), λ k,f,ris a Boolean variable, indicating that if the source node of the fth AI service of the kth user is deployed on r, it is 1; in formula (8), Y k,f,r is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user performs OEO photoelectric conversion at the r-th node, it is 1; G r,p is a Boolean variable, which means that if the pth GPP in the rth node is activated, it is 1; F r is a Boolean variable, indicating that if the rth node is activated, it is 1; N r is the total number of nodes.

[0039] Furthermore, the wavelength allocation in the adjustment network is specifically:

[0040]

[0041] if λ k,f,r =0,k,f,j <sp,r (12)

[0042]

[0043] Equations (11) and (12) ensure the coordination between DU-CU arrangement and optical path allocation, and establish the necessary optical link from access AAU to MEC through the selected GPP node; where O k,f,s,r,p is a Boolean variable, which means that if the f-th AI reasoning service of the k-th user is deployed on the GPPp on the r-th node after being processed by the s-th baseband, it is 1; is a Boolean variable, indicating that if the fth AI reasoning service of the kth user is processed by the sth baseband function and occupies the wth wavelength of the link (m, n), then it is 1; N m,n is a Boolean variable, which means that if the mth node is connected to the nth node, it is 1; k,f,r is a Boolean variable, which means that if the source node of the fth AI service of the kth user is deployed on r, it is 1; Formula (13) ensures the continuity of each service request in the service flow and avoids any interruption in the transmission process; where M r is a Boolean variable, which means that if MEC is deployed at the rth node, it is 1; Formula (14) can prevent the formation of a loop path in the network.

[0044] Furthermore, the capacity requirements of the specified network are specifically:

[0045]

[0046] Formula (15) ensures that the sum of the processing resources required by all services mapped to a GPP node does not exceed its available capacity, where O k,f,s,r,pis a Boolean variable, which means that if the f-th AI reasoning service of the k-th user is deployed on the GPPp on the r-th node after being processed by the s-th baseband, it is 1; The computation and processing requirements of the sth segment of baseband processing function serving the fth AI inference service for the kth user; C p is the GPP computing power processing capacity of the rth node (FLOPS / s); Formula (16) stipulates that the link between the required endpoints must have sufficient bandwidth to satisfy the service request, where is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user is processed by the s-th baseband function and occupies the w-th wavelength of the link (m, n), then it is 1; τ k,f,s The bandwidth requirement for the f-th AI reasoning service of the k-th user after being processed by the s-th baseband function; U l is the bandwidth capacity of the lth link; Formula (17) stipulates that the sum of the processing resources required for all services mapped to the rth node shall not exceed its available capacity, where V k,f,s,r is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user passes through the r-th node after being processed by the s-th baseband function, it is 1; The computation and processing requirements of the sth segment of baseband processing function serving the fth AI inference service for the kth user; U r is the computing power of the rth node; finally, formula (18) ensures that the edge data center m has sufficient capacity to meet all service requests; where m k,f is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user uses the edge data center of the m-th center, it is 1; η k,f The computing power required to serve the fth AI reasoning service for the kth user; U m is the computing power of the mth edge data center; Formula (19) stipulates that each AI reasoning subtask should meet the total delay requirement; where, is a Boolean variable, indicating that if the fth AI reasoning service of the kth user is processed by the sth baseband function and occupies the wth wavelength of the link (m, n), then it is 1; L m,n is the link distance (m,n); φ is the optical fiber propagation delay per kilometer (i.e. 5μs / km); Y k,f,r is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user performs OEO photoelectric conversion at the r-th node, it is 1; T oeo is the delay between OEO and electronic switch; G r,p is a Boolean variable, which means that if the pth GPP in the rth node is activated, it is 1; T v Handle delays for virtual functions; t bv_oxcF is the transmission delay of the variable bandwidth optical cross-connector; r is a Boolean variable, indicating that if the rth node is activated, it is 1; m k,f is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user uses the edge data center of the m-th center, then it is 1; k,f,r is a Boolean variable, which means that if the source node of the fth AI service of the kth user is deployed on r, it is 1; T MEC The latency of processing the fth type of AI inference service for the edge data center; T k,f The total latency requirement for computing power to serve the f-th AI inference class for the k-th user.

[0047] Advantages of the present invention:

[0048] This paper explores a heterogeneous resource collaborative scheduling strategy based on ECFN to improve the processing efficiency of AI reasoning services in NG-RAN, and proposes a mixed integer linear programming mathematical model to maximize the number of successfully sent AI reasoning service subtasks while minimizing the heterogeneous resource occupancy rate and the backlog of AI reasoning service subtasks. In addition, an efficient heuristic algorithm named CHRS-GS is proposed, which adopts a greedy strategy for resource allocation. Description of the drawings:

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0050] Figure 1 This is a diagram of NG-RAN supported by ECFN in an embodiment of the present invention; Specific implementation method:

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] like Figure 1 As shown, the present invention establishes a mixed integer linear programming (MILP) model for the AI ​​reasoning collaborative heterogeneous resource scheduling problem based on distributed model training in NG-RAN supported by ECFN.

[0053] First, the parameter definition of the MILP model is introduced in detail. For the constant parameters shown in Table I, the edge supplier set and edge data center set are denoted as S and E, respectively. The user set and different types of AI reasoning sets are denoted as U and R, respectively, where k∈U and f∈R. In addition, ε is defined as the backlog threshold of the distributed model training tasks in the edge data center.

[0054] For the variable parameters shown in Table II, the computing processing capacity (GOPS) of the jth edge data center of the i-th computing power provider is represented by C i,j , where i∈S, j∈E. B k,f P represents the computing resources (GOPS) required for the f-th AI reasoning service of the k-th user, where k∈U and f∈R. k,f Represents the complaint rate of distributed model training tasks, which is a statistical value. i,j and S i,j They respectively represent the remaining computing resources and computing power matching success rate of the j-th edge data center of the ith computing power provider.

[0055] For the decision variables shown in Table III, define a Boolean variable as If the computing resources requested by the k-th user for the f-th AI reasoning service are allocated to the j-th edge data center of the i-th edge computing provider within the t-th time frame, then Equal to 1. Defined as a Boolean variable, equal to 1 if the jth edge data center of the i-th edge computing provider is used in the t-th time frame. Defined as a Boolean variable, equal to 1 if the jth edge data center of the i-th edge computing provider is used in the t-th time frame.

[0056] Table I

[0057]

[0058] Table II

[0059]

[0060]

[0061] Table III

[0062]

[0063]

[0064] The ultimate goal of the MILP proposed in the present invention is to maximize the number of successfully sent AI reasoning subtasks and minimize the utilization rate of heterogeneous resources in NG-RAN and the backlog of AI reasoning services, as shown in formula (1).

[0065]

[0066] To minimize the resource usage of PP nodes, links and edge data centers in NG-RAN; V k,f,s,r is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user passes through the r-th node after being processed by the s-th baseband function, it is 1; The computation and processing requirements of the sth segment of baseband processing function serving the fth AI inference service for the kth user, U r is the computing power processing capacity of the rth node; is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user is processed by the s-th baseband function and occupies the w-th wavelength of the link (m, n), then it is 1; τ k,f,s The bandwidth requirement for the f-th AI reasoning service of the k-th user after being processed by the s-th baseband function; U l is the bandwidth capacity of the lth link; m k,f is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user uses the edge data center of the m-th center, it is 1; η k,f The computing power required to serve the fth AI reasoning service for the kth user; U m is the computing power processing capacity of the mth edge data center;

[0067] To maximize the number of successfully sent AI reasoning subtasks; where S i,j The computing power matching success rate of the jth edge data center of the i-th computing power provider; The computing power requirement (FLOPS / s) of the k-th user’s f-th AI reasoning service request allocated to the j-th edge data center of the i-th computing power provider;

[0068] To minimize the backlog of AI reasoning services; where σ i,j The AI ​​inference service backlog (FLOPS / s) of the jth edge data center of the i-th edge computing provider; T is a task computing processing cycle; α, β, γ are weight coefficients.

[0069] In addition, several aspects such as computing power service distribution, DU-CU deployment, optical wavelength allocation, capacity and delay constraints should also be considered.

[0070] It is necessary to consider constraints such as edge computing resource requests, computing processing power, task backlog threshold of AI reasoning subtasks on edge servers, and complaint rate threshold of AI reasoning subtasks, as shown in formulas (2)-(5) respectively.

[0071]

[0072] Among them, in formula (2), The computing power requirement (FLOPS / s) allocated to the jth edge data center of the ith computing power provider for the fth AI reasoning service request of the kth user, B k,f is the f-th AI reasoning service request of the k-th user (FLOPS / s); in formula (3), C i,j is the edge computing processing capacity of the jth edge data center of the i-th computing power provider (FLOPS / s); in formula (4), ε is the backlog threshold of the AI ​​reasoning subtask of the edge data center (FLOPS); in formula (5), D i,j is the data processing capacity complaint rate (statistical value) of the jth edge data center of the i-th computing power provider, J i,j The remaining computing power resources (FLOPS) that can be claimed by the j-th edge data center of the i-th computing power provider.

[0073] Formulas (6)-(10) comprehensively regulate the deployment of DU-CU in the network architecture. Specifically, they ensure that the baseband processing function of each user service is deployed on only one GPP node and specify the locations of RU, DU-CU and MEC. In addition, these constraints also guarantee the feasibility of optoelectronic conversion and GPP node activation, which is crucial for the deployment of DU-CU. In summary, these constraints effectively limit and optimize the deployment of DU-CU, meet system requirements, and improve the overall performance of the network.

[0074]

[0075] ∑ k,f,r Y k,f,r ≤∑ r,p G r,p ≤∑ r F r ≤N r (8)

[0076]

[0077] These constraints also ensure the feasibility of photoelectric conversion and GPP node activation; where, in formula (6), O k,f,s,r,pis a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user is deployed on the GPPp on the r-th node after the s-th baseband processing, it is 1; in formula (7), λ k,f,r is a Boolean variable, indicating that if the source node of the fth AI service of the kth user is deployed on r, it is 1; in formula (8), Y k,f,R is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user performs OEO photoelectric conversion at the r-th node, it is 1; G r,p is a Boolean variable, which means that if the pth GPP in the rth node is activated, it is 1; F r is a Boolean variable, indicating that if the rth node is activated, it is 1; N r is the total number of nodes.

[0078] Formulas (11)-(14) regulate the wavelength allocation process in the network. Formulas (11) and (12) ensure the coordination between DU-CU placement and optical path allocation, establishing the necessary optical links from the access AAU to the MEC through the selected GPP nodes. This integration of DU-CU placement and optical path allocation is crucial for the seamless operation of the network. Then, formula (13) guarantees the continuity of each service request in the service flow, avoiding any interruption in the transmission process. Finally, formula (14) prevents the formation of loop paths in the network.

[0079]

[0080] if λ k,f,r =0,k,f,j <sp,r (12)

[0081]

[0082] Equations (11) and (12) ensure the coordination between DU-CU arrangement and optical path allocation, and establish the necessary optical link from access AAU to MEC through the selected GPP node; where O k,f,s,r,p is a Boolean variable, which means that if the f-th AI reasoning service of the k-th user is deployed on the GPPp on the r-th node after being processed by the s-th baseband, it is 1; is a Boolean variable, indicating that if the fth AI reasoning service of the kth user is processed by the sth baseband function and occupies the wth wavelength of the link (m, n), then it is 1; N m,n is a Boolean variable, which means that if the mth node is connected to the nth node, it is 1; k,f,r is a Boolean variable, which means that if the source node of the fth AI service of the kth user is deployed on r, it is 1; Formula (13) ensures the continuity of each service request in the service flow and avoids any interruption in the transmission process; where M ris a Boolean variable, which means that if MEC is deployed at the rth node, it is 1; Formula (14) can prevent the formation of a loop path in the network.

[0083] Equations (15)-(18) specify the capacity requirements of the network. Equation (15) ensures that the sum of the processing resources required for all services mapped to a GPP node does not exceed its available capacity. Equation (16) states that the links between the required endpoints must have sufficient bandwidth to satisfy service requests. Equation (17) states that the sum of the processing resources required for all services mapped to the rth node must not exceed its available capacity. Finally, Equation (18) guarantees that edge data center m has sufficient capacity to satisfy all service requests.

[0084]

[0085] Formula (19) stipulates that each AI reasoning subtask should meet the total delay requirement.

[0086]

[0087]

[0088] Formula (15) ensures that the sum of the processing resources required by all services mapped to a GPP node does not exceed its available capacity, where O k,f,s,r,p is a Boolean variable, which means that if the f-th AI reasoning service of the k-th user is deployed on the GPPp on the r-th node after being processed by the s-th baseband, it is 1; The computation and processing requirements of the sth segment of baseband processing function serving the fth AI inference service for the kth user; C p is the GPP computing power processing capacity of the rth node (FLOPS / s); Formula (16) stipulates that the link between the required endpoints must have sufficient bandwidth to satisfy the service request, where is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user is processed by the s-th baseband function and occupies the w-th wavelength of the link (m, n), then it is 1; τ k,f,s The bandwidth requirement for the f-th AI reasoning service of the k-th user after being processed by the s-th baseband function; U l is the bandwidth capacity of the lth link; Formula (17) stipulates that the sum of the processing resources required for all services mapped to the rth node shall not exceed its available capacity, where V k,f,s,r is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user passes through the r-th node after being processed by the s-th baseband function, it is 1; The computation and processing requirements of the sth segment of baseband processing function serving the fth AI inference service for the kth user; U ris the computing power of the rth node; finally, formula (18) ensures that the edge data center m has sufficient capacity to meet all service requests; where m k,f is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user uses the edge data center of the m-th center, it is 1; η k,f The computing power required to serve the fth AI reasoning service for the kth user; U m is the computing power of the mth edge data center; Formula (19) stipulates that each AI reasoning subtask should meet the total delay requirement; where, is a Boolean variable, indicating that if the fth AI reasoning service of the kth user is processed by the sth baseband function and occupies the wth wavelength of the link (m, n), then it is 1; L m,n is the link distance (m,n); φ is the optical fiber propagation delay per kilometer (i.e. 5μs / km); Y k,f,r is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user performs OEO photoelectric conversion at the r-th node, it is 1; T oeo is the delay between OEO and electronic switch; G r,p is a Boolean variable, which means that if the pth GPP in the rth node is activated, it is 1; T v Handle delays for virtual functions; t bv_oxc F is the transmission delay of the variable bandwidth optical cross-connector; r is a Boolean variable, indicating that if the rth node is activated, it is 1; m k,f is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user uses the edge data center of the m-th center, then it is 1; k,f,r is a Boolean variable, which means that if the source node of the fth AI service of the kth user is deployed on r, it is 1; T MEC The latency of processing the fth type of AI inference service for the edge data center; T k,f The total latency requirement for computing power to serve the f-th AI inference class for the k-th user.

[0089] The present invention also allocates resources based on the cooperative heterogeneous resource scheduling algorithm based on the greedy strategy. The common optimization goal of the heuristic algorithm is to maximize the number of successfully sent AI reasoning service subtasks and minimize the utilization rate of heterogeneous resources in NG-RAN and the backlog of AI reasoning subtasks. The proposed heuristic algorithm uses the greedy thinking strategy to allocate the AI ​​reasoning service with the greatest computing demand to the edge data center with the least computing resources. The pseudo code of the cooperative heterogeneous resource scheduling algorithm based on the greedy strategy (CHRS-GS) is detailed in Algorithm Table IV.

[0090] Edge computing resource requests for all AI inference services M k,fSort by the number of computing resources required in descending order to get a list X. According to the number of remaining computing resources N in the edge data center i,j Sort in ascending order to get the list Y. Consider the remaining computing resources of the edge data center as (C i,j +σ i,j )(Table IV, rows 1-3). Search all users in descending order of computing power requirements. If the computing resource request is positive, and the computing capacity of each edge data center meets the computing resources requested by each user, and the data processing capacity complaint rate of all users of the jth edge data center allocated to the i-th computing power provider meets D i,j , then the fth AI reasoning service of the kth user is assigned to the jth edge data center of the ith computing power provider. i,j ′-B k,f Assign the value of C i,j '(Table IV, rows 4-11). If C i,j There is not enough edge computing power in B k,f The data processing capacity complaint rate of all users of the jth edge data center assigned to the i-th computing power provider satisfies D i,j , then C on the jth edge data center of the i-th computing power provider i,j ′ is assigned to the fth AI reasoning service of the kth user. Then, 0 is assigned to C i,j ′,(B k,f -C i,j ′) is assigned to B k,f Until all B k,f All computing power requests are assigned to the edge data center, and the algorithm ends (Table IV, lines 12-21). Third, the computing power resource usage, the number of successfully matched AI reasoning subtasks, the backlog of AI reasoning subtasks, and the joint optimization results of the three are calculated (Table IV, line 22).

[0091] Table IV

[0092]

[0093]

[0094] In summary, this paper explores a heterogeneous resource collaborative scheduling strategy based on ECFN to improve the processing efficiency of AI reasoning services in NG-RAN, and proposes a mixed integer linear programming mathematical model to maximize the number of successfully sent AI reasoning service subtasks while minimizing the heterogeneous resource occupancy rate and the backlog of AI reasoning service subtasks. In addition, an efficient heuristic algorithm named CHRS-GS is proposed, which adopts a greedy strategy for resource allocation.

[0095] This paper shows that ECFN has great potential in improving the efficiency of AI reasoning service execution, and also highlights the importance of optimizing resource scheduling in achieving high-performance network services. Future research can further explore how to apply these scheduling strategies to more complex network environments and consider more dynamic factors and user needs to promote the development of intelligent networks.

[0096] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for coordinating heterogeneous resources for training AI reasoning services, characterized in that: The following steps are involved: A mixed integer linear programming model MILP is established for the AI ​​inference collaborative heterogeneous resource scheduling problem based on distributed model training. The ultimate goal of the proposed MILP is to maximize the number of successfully sent AI inference subtasks and minimize the utilization rate of heterogeneous resources and the backlog of AI inference services in NG-RAN. Construct edge computing resource requests, computing processing capabilities, task backlog thresholds for AI reasoning subtasks on edge servers, and complaint rate thresholds for AI reasoning subtasks as constraints; Standardize the deployment of DU-CU in the network architecture, ensure that the baseband processing function of each user service is deployed on only one GPP node, and specify the locations of RU, DU-CU and MEC; Adjust the wavelength allocation in the network, establish the necessary optical links from the access AAU to the MEC through the selected GPP nodes, and establish constraints to ensure the continuity of each service request in the service flow and prevent the formation of loop paths in the network; Specify the capacity requirements of the network, including limiting the sum of processing resources required for all services mapped to the GPP node to not exceed its available capacity, specifying that the links between the required endpoints must have sufficient bandwidth to meet service requests, specifying that the sum of processing resources required for all services mapped to the rth node must not exceed its available capacity, specifying that the edge data center m has sufficient capacity to meet all service requests, and specifying that each AI reasoning subtask should meet the total latency requirements.

2. According to claim 1, a heterogeneous resource collaborative scheduling method for training AI reasoning services is characterized in that: It also includes a cooperative heterogeneous resource scheduling algorithm based on a greedy strategy for resource allocation, specifically: Sort the computing power requirements in descending order to obtain the first list; Sort the remaining edge computing resources of the edge data center in ascending order to obtain a second list; Build the remaining edge computing resources of the edge data center; Allocate computing power requests to edge data centers; Calculate the usage of computing resources, the number of successfully matched AI reasoning subtasks, the backlog of AI reasoning subtasks, and the joint optimization results of the three.

3. According to claim 2, a method for coordinating heterogeneous resources for training AI reasoning services is characterized in that: The computing power request is allocated to the edge data center, specifically: Search all users in descending order of computing power requirements. If the computing resource request is positive, and the computing capacity of each edge data center meets the computing resources requested by each user, and the data processing capacity complaint rate of all users of the jth edge data center allocated to the i-th computing power provider meets D i,j , then the f-th AI reasoning service of the k-th user is assigned to the j-th edge data center of the ith computing power provider; C i,j ′-B k,f Assign the value of C i,j ′, if C i,j There is not enough edge computing power in B k,f The data processing capacity complaint rate of all users of the jth edge data center assigned to the i-th computing power provider satisfies D i,j , then C on the jth edge data center of the i-th computing power provider i,j ′The fth AI reasoning service assigned to the kth user; 0 is assigned to C i,j ′,(B k,f -C i,j ′) is assigned to B k,f , until all B k,f All computing power requests are allocated to the edge data center, and the algorithm ends.

4. According to claim 1, a heterogeneous resource collaborative scheduling method for training AI reasoning services is characterized in that: The ultimate goal of the MILP is to maximize the number of successfully sent AI reasoning subtasks and minimize the utilization of heterogeneous resources in NG-RAN and the backlog of AI reasoning services. The specific formula is expressed as follows: To minimize the resource usage of PP nodes, links and edge data centers in NG-RAN; V k,f,s,r is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user passes through the r-th node after being processed by the s-th baseband function, it is 1; The computation and processing requirements of the sth segment of baseband processing function serving the fth AI inference service for the kth user, U r is the computing power processing capacity of the rth node; is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user is processed by the s-th baseband function and occupies the w-th wavelength of the link (m, n), then it is 1; τ k,f,s The bandwidth requirement for the f-th AI reasoning service of the k-th user after being processed by the s-th baseband function; U l is the bandwidth capacity of the lth link; m k,f is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user uses the edge data center of the m-th center, it is 1; η k,f The computing power required to serve the fth AI reasoning service for the kth user; U m is the computing power processing capacity of the mth edge data center; To maximize the number of successfully sent AI reasoning subtasks; where S i,j The computing power matching success rate of the jth edge data center of the i-th computing power provider; The computing power demand of the f-th AI reasoning service request of the k-th user allocated to the j-th edge data center of the i-th computing power provider; To minimize the backlog of AI reasoning services; where σ i,j The backlog of AI inference services in the jth edge data center of the i-th edge computing provider; T is a task computing processing cycle; α, β, and γ are weight coefficients.

5. According to claim 1, a method for coordinating heterogeneous resources for training AI reasoning services is characterized in that: The edge computing resource request, computing processing capacity, task backlog threshold of AI reasoning subtasks on edge servers, and complaint rate threshold of AI reasoning subtasks are constructed as constraints, specifically: Among them, in formula (2), The computing power requirement of the jth edge data center allocated to the i-th computing power provider for the f-th AI reasoning service request of the k-th user, B k,f is the f-th AI reasoning service request of the k-th user; in formula (3), C i,j is the edge computing processing capacity of the jth edge data center of the i-th computing power provider; in formula (4), ε is the backlog threshold of the AI ​​reasoning subtask of the edge data center; in formula (5), D i,j is the data processing capacity complaint rate of the jth edge data center of the i-th computing power provider, J i,j The remaining computing power resources that can be appealed by the j-th edge data center of the i-th computing power provider.

6. According to claim 1, a heterogeneous resource collaborative scheduling method for training AI reasoning services is characterized in that: The specification specifies the deployment of DU-CU in the network architecture, ensuring that the baseband processing function for each user service is deployed on only one GPP node, and specifies the locations of RU, DU-CU and MEC, specifically: These constraints also ensure the feasibility of photoelectric conversion and GPP node activation; where, in formula (6), O k,f,s,r,p is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user is deployed on the GPPp on the r-th node after the s-th baseband processing, it is 1; in formula (7), λ k,f,r is a Boolean variable, indicating that if the source node of the fth AI service of the kth user is deployed on r, it is 1; in formula (8), Y k,f,r is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user performs OEO photoelectric conversion at the r-th node, it is 1; G r,p is a Boolean variable, which means that if the pth GPP in the rth node is activated, it is 1; F r is a Boolean variable, indicating that if the rth node is activated, it is 1; N r is the total number of nodes.

7. The method for coordinating heterogeneous resources for training AI reasoning services according to claim 1, characterized in that: The wavelength allocation in the adjustment network is specifically: if λ k,f，r =0,k,f,j<sp,r (12) Equations (11) and (12) ensure the coordination between DU-CU arrangement and optical path allocation, and establish the necessary optical link from access AAU to MEC through the selected GPP node; where O k,f,s,r,p is a Boolean variable, which means that if the f-th AI reasoning service of the k-th user is deployed on the GPPp on the r-th node after being processed by the s-th baseband, it is 1; is a Boolean variable, indicating that if the fth AI reasoning service of the kth user is processed by the sth baseband function and occupies the wth wavelength of the link (m, n), then it is 1; N m,n is a Boolean variable, which means that if the mth node is connected to the nth node, it is 1; k,f,r is a Boolean variable, which means that if the source node of the fth AI service of the kth user is deployed on r, it is 1; Formula (13) ensures the continuity of each service request in the service flow and avoids any interruption in the transmission process; where M r is a Boolean variable, which means that if MEC is deployed at the rth node, it is 1; Formula (14) can prevent the formation of a loop path in the network.

8. The method for coordinating heterogeneous resources for training AI reasoning services according to claim 1, characterized in that: The capacity requirements of the specified network are specifically: Formula (15) ensures that the sum of the processing resources required by all services mapped to a GPP node does not exceed its available capacity, where O k,f,s,r,p is a Boolean variable, which means that if the f-th AI reasoning service of the k-th user is deployed on the GPPp on the r-th node after being processed by the s-th baseband, it is 1; The computation and processing requirements of the sth segment of baseband processing function serving the fth AI inference service for the kth user; C p is the GPP computing power processing capacity of the rth node (FLOPS / s); Formula (16) stipulates that the link between the required endpoints must have sufficient bandwidth to satisfy the service request, where is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user is processed by the s-th baseband function and occupies the w-th wavelength of the link (m, n), then it is 1; τ k,f,s The bandwidth requirement for the f-th AI reasoning service of the k-th user after being processed by the s-th baseband function; U l is the bandwidth capacity of the lth link; Formula (17) stipulates that the sum of the processing resources required for all services mapped to the rth node shall not exceed its available capacity, where V k,f,s,r is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user passes through the r-th node after being processed by the s-th baseband function, it is 1; The computation and processing requirements of the sth segment of baseband processing function serving the fth AI inference service for the kth user; U r is the computing power of the rth node; finally, formula (18) ensures that the edge data center m has sufficient capacity to meet all service requests; where m k,f is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user uses the edge data center of the m-th center, it is 1; η k,f The computing power required to serve the fth AI reasoning service for the kth user; U m is the computing power of the mth edge data center; Formula (19) stipulates that each AI reasoning subtask should meet the total delay requirement; where, is a Boolean variable, indicating that if the fth AI reasoning service of the kth user is processed by the sth baseband function and occupies the wth wavelength of the link (m, n), then it is 1; L m,n is the link distance (m,n); φ is the optical fiber propagation delay per kilometer (i.e. 5μs / km); Y k,f,r is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user performs OEO photoelectric conversion at the r-th node, it is 1; T oeo is the delay between OEO and electronic switch; G r,p is a Boolean variable, which means that if the pth GPP in the rth node is activated, it is 1; T v Handle delays for virtual functions; t bv_oxc F is the transmission delay of the variable bandwidth optical cross-connector; r is a Boolean variable, indicating that if the rth node is activated, it is 1; m k,f is a Boolean variable, indicating that if the f-th AI reasoning service of the k-th user uses the edge data center of the m-th center, then it is 1; k,f,r is a Boolean variable, which means that if the source node of the fth AI service of the kth user is deployed on r, it is 1; T MEC The latency of processing the fth type of AI inference service for the edge data center; T k,f The total latency requirement for computing power to serve the f-th AI inference class for the k-th user.

Images