A cloud-network converged service computing resource allocation method based on a two-layer game model

Abstract

The application discloses a cloud-network fusion service computing resource allocation method. Through the hierarchical resource allocation framework, the cooperation of three levels of CSP, ESP and terminal users is included, the data processing demand of the terminal user is estimated, the optimal resource request of each edge server is determined, and the resource request strategy is formulated to maximize the utility. The CSP determines the appropriate computing resource pricing to maximize its utility function, and realizes the optimization of resource allocation. In the two-layer Stackelberg Game model, the participants reach their own benefit maximization through mutual game and decision, the game model is solved by using the backward induction method, the optimal resource allocation scheme is determined, and the collaborative allocation of resources and the maximization of benefits are realized. The application has the advantages that the utility optimization of the terminal user can be realized, the service quality and satisfaction are improved, and the data processing delay and computing performance in the edge computing scene are optimized.

Description

Technical Field

[0001] This invention relates to the field of cloud computing technology, and in particular to a cloud-network converged service computing resource allocation method based on a two-layer game model. Background Technology

[0002] As cloud computing technology matures and becomes more widespread, traditional centralized data processing models can no longer meet the ever-increasing demand for data processing. Cloud computing provides a flexible and efficient way to process data, but in some cases, data processing needs require being closer to the source of the data.

[0003] Edge computing, as an emerging computing paradigm, emphasizes providing computing and storage services near the source of data generation to reduce data transmission latency and network congestion. Edge computing architectures are better suited to applications with high real-time requirements, such as the Internet of Things (IoT) and smart cities.

[0004] Stackelberg Game theory is widely used in multi-party game decision-making scenarios, including multi-level resource allocation problems. In this game model, there are leader-follower relationships among the participants, who formulate strategies to maximize their own utility.

[0005] In cloud and network converged architectures, the rational allocation and management of computing resources is a key issue. Traditional resource allocation methods often struggle to adapt to dynamically changing demands, thus requiring new algorithms and strategies to optimize resource utilization and performance.

[0006] Considering the interaction between cloud computing, edge computing, and end users, a multi-master, multi-slave game model can better describe the competition and cooperation among the parties, thereby finding the optimal resource allocation scheme.

[0007] A sound pricing strategy for computing resources is crucial for both cloud and edge service providers. By developing appropriate pricing strategies, both profit maximization and resource utilization optimization can be achieved.

[0008] To address dynamically changing computing resource demands, efficient resource allocation algorithms need to be designed. These algorithms should be able to adjust resource allocation in a timely manner to adapt to changes in demand over different times and in different scenarios.

[0009] In edge computing scenarios, optimizing end-to-end data processing latency is crucial. By performing some computational tasks at edge nodes closer to the data generation source, data transmission latency can be significantly reduced, and real-time performance improved.

[0010] With the continuous improvement of the computing power of terminal devices, more and more computing tasks can be executed on terminal devices, thereby reducing the pressure on centralized data processing and promoting the development of edge computing.

[0011] The cloud-network converged architecture combines cloud computing and edge computing to provide users with more flexible and efficient data processing services. By enabling dynamic scheduling and collaborative work of resources between cloud and edge nodes, better resource utilization and service quality can be achieved. Summary of the Invention

[0012] This invention addresses the shortcomings of existing technologies by providing a cloud-network converged service computing resource allocation method. By combining cloud computing and edge computing, it offers users more flexible and efficient data processing services. Through dynamic scheduling and collaborative work of resources between cloud and edge nodes, better resource utilization and service quality can be achieved.

[0013] To achieve the above-mentioned objectives, the technical solution adopted by the present invention is as follows:

[0014] A cloud-network converged service computing resource allocation method includes: a hierarchical resource allocation framework, a two-layer Stackelberg Game model, and a resource allocation algorithm based on the two-layer Stackelberg Game model;

[0015] The hierarchical resource allocation framework includes collaboration among three levels: cloud service providers (CSP), edge service providers (ESP), and end users.

[0016] In the hierarchical resource allocation framework, the data processing needs of end users are first estimated to determine the optimal resource requests for each edge server. Factors such as the distance from the terminal device to the edge server, the size of the task data, and the computational requirements of the task are considered to determine the range of data processing needs for each end user. Then, based on the estimation of end-user data processing needs, the edge servers formulate resource request strategies to meet user needs and maximize their own utility.

[0017] CSPs need to determine appropriate pricing for computing resources to maximize their utility function. This involves considering resource requests from edge servers and market supply and demand to develop suitable resource pricing strategies for optimal resource allocation.

[0018] The two-layer Stackelberg Game model aims to maximize the interests of CSP, ESP, and end users through mutual game theory and decision-making. Backward induction is employed to solve the game theory model, determine the optimal resource allocation scheme, and achieve coordinated resource allocation and profit maximization by optimizing end-user data processing needs and resource pricing strategies.

[0019] The resource allocation algorithm includes: using a gradient descent algorithm to determine the optimal pricing strategy for resources; and using an iterative update method to optimize the resource allocation strategy.

[0020] Furthermore, the two-layer Stackelberg Game model is divided into two subgames, specifically:

[0021] End-Edge Layer Subgame: In this subgame, the terminal device acts as a follower, and the ESP (Electronic Programming Entity) acts as the leader. The terminal device decides the quantity of computing resources to purchase by considering the ESP's pricing strategy. The ESP formulates a pricing strategy for computing resources to maximize its profit. Its pricing strategy takes into account the pricing strategies of other ESPs and the terminal device's resource requirements to achieve profit maximization.

[0022] Edge-Cloud Layer Subgame: In this subgame, the ESP acts as a follower, and the CSP acts as the leader. The ESP decides to purchase appropriate computing resources based on the CSP's data processing resource pricing and allocates suitable computing resources to tasks. The CSP formulates an appropriate computing resource pricing strategy to maximize its utility function. The pricing strategy considers the ESP's resource needs and other market factors to maximize profits.

[0023] Furthermore, the resource allocation algorithm includes:

[0024] Algorithm 1: Terminal device task resource request and edge server resource allocation strategy:

[0025] 1) Initialize the iteration count t to 0;

[0026] 2) Initialization precision;

[0027] 3) Begin the iterative process until the predetermined accuracy is reached;

[0028] 4) In each iteration, each edge node observes the pricing of other edge nodes;

[0029] 5) Then, each edge node updates its pricing strategy using an update strategy based on the pricing behavior of other nodes it observes;

[0030] 6) Increase the iteration number t by 1;

[0031] 7) Continue iterating until the predetermined accuracy is achieved.

[0032] Algorithm 2: Adjustment of CSP's resource pricing strategy:

[0033] 1) Initialization: Set the iteration count k = 0, and initialize the CSP pricing p. c .

[0034] 2) Conveying pricing information: CSP will deliver pricing p c Send to all ESPs.

[0035] 3) Calculate resource purchase strategy: Each ESP calculates its initial resource purchase strategy.

[0036] 4) Update Pricing Strategy: Once the CSP receives all resource purchase requests from the ESPs, it will update its pricing strategy. The update formula is:

[0037]

[0038] Where θ is the update step size, The CSP utility function represents the price p. c The partial derivatives;

[0039] 5) Iterative Update: Set k to 1 and repeat the above process until the required accuracy is achieved, at which point the algorithm stops iterating. Find the approximate optimal solution for CSP and ESP, and obtain the optimal utility.

[0040] Compared with the prior art, the advantages of the present invention are as follows:

[0041] This invention considers the interactions between cloud service providers, edge service providers, and end users, enabling collaborative decision-making among the game participants. By taking into account the interests and behaviors of each party, resource allocation and pricing can be better balanced, maximizing overall benefits.

[0042] This invention proposes a dynamic resource pricing strategy that can adjust resource pricing in real time based on market demand and supply. This strategy can help cloud service providers maximize profits and optimize resource utilization while maintaining reasonable and competitive pricing.

[0043] This invention employs a gradient-based dynamic pricing algorithm to optimize resource pricing strategies, enabling iterative adjustments and optimization of resource prices. This algorithm can more effectively adapt to market changes and participant behavior, improving the accuracy and efficiency of resource pricing.

[0044] This invention achieves collaborative resource allocation between cloud service providers and edge service providers through a resource allocation algorithm based on a two-layer Stackelberg Game. This collaborative allocation can better meet the needs of end users and improve service quality and satisfaction.

[0045] This invention optimizes end-user utility by improving resource allocation methods and pricing strategies. Simultaneously, it optimizes data processing latency and computing performance in edge computing scenarios, improving the real-time performance and efficiency of data processing, thereby enhancing end-user satisfaction and service quality. Attached Figure Description

[0046] Figure 1 This is a structural diagram of the cloud network converged service computing resource allocation system model according to an embodiment of the present invention;

[0047] Figure 2 This is a resource allocation structure diagram based on a two-layer Stackelberg Game in an embodiment of the present invention;

[0048] Figure 3 This is a schematic diagram of the resource pricing iteration process of the edge service provider in an embodiment of the present invention;

[0049] Figure 4 This is a schematic diagram of the resource pricing iteration process of a cloud service provider according to an embodiment of the present invention;

[0050] Figure 5 This is a schematic diagram of the end-user uninstallation strategy (user 1) according to an embodiment of the present invention;

[0051] Figure 6 This is a schematic diagram of a resource request from an edge service provider according to an embodiment of the present invention;

[0052] Figure 7 This is a schematic diagram illustrating the average utility of end users under different numbers of end users in an embodiment of the present invention;

[0053] Figure 8 This is a schematic diagram illustrating the average end-user utility under different numbers of end-user service providers in embodiments of the present invention;

[0054] Figure 9 This is a schematic diagram illustrating the utility of the edge service provider under different numbers of terminal users in an embodiment of the present invention;

[0055] Figure 10 This is a schematic diagram illustrating the utility of edge service providers under different numbers of edge service providers in an embodiment of the present invention;

[0056] Figure 11 This is a schematic diagram illustrating the utility of cloud service providers under different numbers of edge service providers in embodiments of the present invention;

[0057] Figure 12 This is a schematic diagram illustrating cloud service provider resource pricing under different numbers of edge service providers in an embodiment of the present invention. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and examples.

[0059] This invention provides a method for allocating computing resources for cloud network converged services based on a two-layer Stackelberg game. Figure 1 This is the model of the entire system.

[0060] (1) Hierarchical resource allocation framework

[0061] This framework primarily comprises three layers: Cloud Service Providers (CSPs), Edge Service Providers (ESPs), and end users. Within this framework, the optimal resource requests for each edge server are first determined by estimating the data processing needs of end users. Then, the CSP maximizes its utility function by determining appropriate pricing for computing resources, thereby achieving optimal resource allocation.

[0062] First, the data processing needs of end users are estimated. By considering factors such as the distance from the terminal device to the edge server, the size of the task data, and the computational requirements of the task, the range of data processing needs for each end user is determined. Based on the estimation of the end user's data processing needs, the optimal resource request for each edge server is determined. This process involves the edge server formulating a resource request strategy based on the end user's data processing needs to meet user demands and maximize its own utility.

[0063] Determining appropriate computing resource pricing maximizes the utility function of cloud service providers. Cloud service providers formulate suitable resource pricing strategies based on resource requests from edge servers and market supply and demand, achieving optimal resource allocation. This hierarchical resource allocation framework enables collaborative work among cloud service providers, edge service providers, and end users, effectively solving resource allocation and pricing issues.

[0064] (2) Two-layer Stackelberg Game model

[0065] In this model, participants include CSPs, ESPs, and end users, who maximize their respective interests through mutual game theory and decision-making. Backward induction is employed to solve the game theory model, optimizing end-user data processing needs and resource pricing strategies to achieve coordinated resource allocation and profit maximization.

[0066] The two-layer Stackelberg Game model proposed in this invention involves the interaction between cloud service providers (CSPs), edge service providers (ESPs), and terminal devices. Specifically, the model consists of the following two subgames:

[0067] ① End-Edge Layer Subgame:

[0068] In this subgame, the terminal device acts as a follower, and the edge service provider acts as a leader. The terminal device decides the quantity of computing resources to purchase by considering the edge service provider's pricing strategy. The edge service provider formulates a pricing strategy for computing resources to maximize its profits. Its pricing strategy takes into account the pricing strategies of other edge service providers and the resource needs of the terminal device to achieve profit maximization.

[0069] ② Edge-Cloud Layer Subgame:

[0070] In this subgame, the edge service provider acts as a follower, and the cloud service provider acts as a leader. The edge service provider decides to purchase corresponding computing resources based on the cloud service provider's data processing resource pricing and allocates appropriate computing resources to tasks. The cloud service provider formulates a suitable computing resource pricing strategy to maximize its utility function. The pricing strategy considers the edge service provider's resource needs and other market factors to maximize profits.

[0071] Through the interaction of two subgames, the coordination of resource allocation and pricing strategies among the participants in the entire system is achieved, maximizing the overall utility.

[0072] (3) Dynamic resource pricing strategy

[0073] This strategy dynamically adjusts resource pricing using a gradient descent algorithm. First, resource pricing is initialized and transmitted to all ESPs (Entity Purchase Strategies). Then, each ESP calculates its resource purchasing strategy based on the received pricing information and sends its resource request to the CSP (Consumer Service Strategist). Finally, the CSP dynamically adjusts resource pricing based on the received resource requests using the gradient descent algorithm until convergence is achieved.

[0074] The dynamic resource pricing strategy is a key part of this invention. It uses the gradient descent method to enable cloud service providers (CSPs) to dynamically price computing resources.

[0075] (4) Resource allocation algorithm based on two-layer Stackelberg Game

[0076] This algorithm achieves coordinated resource allocation between the CSP and ESP through a two-layer Stackelberg Game model. First, Algorithm 1 calculates the task resource requests of the terminal device and the resource allocation strategy of the edge server. Then, Algorithm 2 calculates the resource requests of the ESP and the resource pricing strategy of the CSP. Finally, through this two-layer game model, equilibrium resource allocation and optimal pricing are achieved.

[0077] The resource allocation algorithm based on the two-layer Stackelberg Game is one of the key algorithms in this invention, used to solve the resource allocation problem among cloud service providers (CSP), edge service providers (ESP), and end users.

[0078] Edge service providers update their resource pricing strategies based on the best resource requests they receive. Cloud service providers dynamically adjust resource pricing based on all resource requests received from edge service providers.

[0079] For each edge service provider and terminal device, resources are allocated based on the optimal resource allocation and pricing obtained from Algorithm 1 and Algorithm 2.

[0080] Edge service providers and cloud service providers provide services to end users and generate revenue based on optimal resource allocation and pricing.

[0081] In a game theory model between cloud service providers (CSPs) and edge service providers (ESPs), each player transmits its information to a central controller to ensure that other players can access each other's strategies. Backward induction is used to solve for the optimal strategies of the players in the game. First, the optimal demand for data processing resources is obtained by differentiation. Then, the optimal pricing scheme is determined based on the optimal resource demand.

[0082] Specifically, first, the number of iterations is set to k=0, and then the CSP sets the initial price p within the pricing range. c Then p c The CSP sends the resource purchase requests to all ESPs. Next, each ESP calculates its initial resource purchasing strategy. Once the CSP receives resource purchase requests from all ESPs, it updates its pricing strategy. The iterative update expression for the CSP's resource pricing strategy is as follows:

[0083]

[0084] Where θ is the update step size; then k is set to 1, and the above process is repeated until the required accuracy is achieved, at which point the algorithm stops iterating. Approximate optimal solutions for CSP and ESP can be found, denoted as follows: and Therefore, the optimal utility of CSP and ESP can be obtained.

[0085] This resource allocation algorithm, based on a two-layer Stackelberg Game, enables competition and cooperation among end users, edge service providers, and cloud service providers, maximizing the utility of each party and effectively solving the optimization problem of resource allocation.

[0086] (5) Theoretical Analysis

[0087] The computational complexity of the equilibrium summation algorithm for a two-layer Stackelberg game is analyzed, and related theorems are given.

[0088] (6) Simulation experiments and performance analysis

[0089] Simulation experiments verified the effectiveness and superiority of the proposed method. Different schemes were compared, including a full-edge processing scheme and a full-cloud processing scheme. Performance analysis demonstrated the performance of different schemes in terms of end-user utility, ESP utility, and CSP utility, and the computational complexity of the algorithm was also analyzed.

[0090] Through simulation experiments and performance analysis, the performance of the proposed algorithm can be comprehensively evaluated, and its effectiveness and feasibility in practical applications can be verified.

[0091] like Figure 3 As shown, the resource pricing for each edge service provider eventually converges to a fixed price, verifying the convergence of Algorithm 1. Figure 4 This shows how cloud service provider resource pricing changes with the number of iterations. It can be seen that as the number of iterations increases, the cloud service provider's resource pricing gradually stabilizes, and after approximately 25 iterations, it tends to stabilize and approach its optimum.

[0092] Figure 5 The presentation shows the trend of end-user 1's task offloading strategy as the number of iterations increases. Observing user 1's task offloading strategy, it can be found that the order of task offloading from largest to smallest is ESP1, ESP4, ESP2, and ESP3. This is because the end-user's task offloading strategy takes into account the service pricing of ESPs in order to formulate an appropriate offloading strategy. Figure 6 This illustrates the trend of resource requests for each ESP during the iteration process. It can be seen that as the number of iterations increases, the resource requests for each ESP tend to stabilize.

[0093] Figure 7 The average end-user utility for each scheme under different numbers of end-users is explained, where the price of ESP is set to 3 in the FOO scheme. Considering that the FC and FE schemes have no impact on end-user utility, i.e., the end-user utility of the FC and FE schemes is the same as that of the scheme proposed in this invention, therefore... Figure 7 and Figure 8The average end-user utility of the FC and FE schemes was not shown.

[0094] Figure 8 The average end-user utility for each scheme under different numbers of ESPs is shown. It can be observed that the average end-user utility increases with the number of ESPs. This is because end-users gain computational utility by offloading tasks to ESPs. End-users can obtain positive utility (i.e., a utility value greater than 0) by choosing an appropriate offloading strategy based on the service pricing of the ESPs. Therefore, the end-user utility increases with the number of ESPs. Compared to other schemes, the average end-user utility of the method proposed in this invention is higher than the PP scheme but lower than the FOO scheme. This is because the final service pricing of the scheme proposed in this invention is higher than the FOO scheme but lower than the PP scheme.

[0095] Figure 11 The utility of CSP under different pricing strategies is shown. It can be observed that the utility of CSP gradually increases with the number of ESPs. This is because... Figure 12 As shown, the resource pricing trend of CSP varies with different numbers of ESPs. It can be observed that as the number of iterations increases, the resource pricing strategy of CSP tends to stabilize and eventually converges.

[0096] The methods described above according to the invention can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium (such as a CD-ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code originally stored on a remote recording medium or a non-transitory machine-readable medium and subsequently stored on a local recording medium, downloaded via a network. Thus, the methods described herein can be stored as software processing on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It is understood that the computer, processor, microprocessor controller, or programmable hardware includes storage components (e.g., RAM, ROM, flash memory, etc.) capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the cloud-network converged service computing resource allocation method described herein. Furthermore, when a general-purpose computer accesses the code used to implement the processing shown herein, the execution of the code transforms the general-purpose computer into a dedicated computer for performing the processing shown herein.

[0097] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the implementation methods of the present invention, and should be understood that the scope of protection of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of the present invention.

Claims

1. A method for allocating computing resources for cloud-network converged services, characterized in that, include: A hierarchical resource allocation framework, a two-layer Stackelberg Game model, and a resource allocation algorithm based on the two-layer Stackelberg Game model; The hierarchical resource allocation framework includes collaboration at three levels: cloud service provider (CSP), edge service provider (ESP), and end user. In the hierarchical resource allocation framework, the data processing needs of end users are first estimated to determine the optimal resource request for each edge server. Factors such as the distance from the terminal device to the edge server, the size of the task data, and the task computing requirements are considered to determine the range of data processing needs for each end user. Then, based on the estimation of the data processing needs of end users, the edge server formulates a resource request strategy to meet user needs and maximize its own utility. CSPs need to determine appropriate pricing for computing resources to maximize their utility function; they need to consider resource requests from edge servers and market supply and demand to formulate appropriate resource pricing strategies to achieve optimal resource allocation. The two-layer Stackelberg Game model aims to maximize the interests of CSP, ESP, and end users through mutual game and decision-making. It uses reverse induction to solve the game model, determine the optimal resource allocation scheme, and achieve coordinated resource allocation and profit maximization by optimizing the data processing needs of end users and resource pricing strategies. The resource allocation algorithm includes: using a gradient descent algorithm to determine the optimal pricing strategy for resources; and using an iterative update method to optimize the resource allocation strategy.

2. The cloud-network converged service computing resource allocation method according to claim 1, characterized in that: The two-layer Stackelberg Game model is divided into two subgames, specifically: Edge-end subgame: In this subgame, the terminal device acts as a follower and the ESP acts as the leader; the terminal device decides the quantity of computing resources to purchase by considering the ESP's pricing of computing resources; the ESP formulates a pricing strategy for computing resources to maximize its profits. Its pricing strategy takes into account the pricing strategies of other ESPs and the resource requirements of terminal devices in order to maximize profits; Edge-Cloud Subgame: In this subgame, ESP acts as a follower and CSP acts as a leader; ESP decides to purchase corresponding computing resources based on CSP's data processing resource pricing and allocates appropriate computing resources to tasks; CSP formulates appropriate computing resource pricing strategies to maximize its utility function; the pricing strategy takes into account ESP's resource needs and other market factors to maximize profits.

3. The cloud-network converged service computing resource allocation method according to claim 1, characterized in that: The resource allocation algorithm includes: Algorithm 1: Terminal device task resource request and edge server resource allocation strategy: 1) Initialize the iteration count t to 0; 2) Initialization precision; 3) Begin the iterative process until the predetermined accuracy is reached; 4) In each iteration, each edge node observes the pricing of other edge nodes; 5) Then, each edge node updates its pricing strategy using an update strategy based on the pricing behavior of other nodes it observes; 6) Increase the iteration number t by 1; 7) Continue iterating until the predetermined accuracy is reached; Algorithm 2: Adjustment of CSP's resource pricing strategy: 1) Initialization: Set the iteration count k = 0, and initialize the CSP pricing p. c ; 2) Conveying pricing information: CSP will deliver pricing p c Send to all ESPs; 3) Calculate resource purchasing strategy: Each ESP calculates its initial resource purchasing strategy; 4) Update pricing strategy: Once the CSP receives all resource purchase requests from the ESPs, it will update its pricing strategy; the update formula is: Where θ is the update step size, The CSP utility function represents the price p. c The partial derivatives; 5) Iterative update: Set k to 1 and repeat the above process until the accuracy requirement is met, at which point the algorithm stops iterating; find the approximate optimal solution for CSP and ESP, and obtain the optimal utility.

Images