A cloud network edge network task offloading and resource allocation method

Abstract

The application discloses a cloud network edge network task unloading and resource allocation method, and belongs to the technical field of calculation, estimation or counting. The method adopts a solution based on the maximum minimum criterion, proposes a strategy for solving the problem that there is a large difference between the communication calculation capabilities and the load conditions of different nodes, and realizes efficient utilization and sharing of resources. The method constructs a cloud network edge network model, including a communication model and a calculation model, describes the user unloading decision and the task processing mode, defines a user energy efficiency function and constraint conditions to jointly unload the decision and allocate resources. The generalized fraction theory and the slack variable are introduced to convert the optimization problem into an equivalent convex optimization problem, and then an iterative algorithm is proposed to obtain the optimal solution of the problem. The simulation results show that the method has good performance and effectively guarantees the fairness of the resources obtained by the user.

Description

Technical Field

[0001] This invention relates to communication technology, specifically disclosing a method for task offloading and resource allocation in cloud-network-edge-device networks, belonging to the technical field of calculation, estimation, or counting. Background Technology

[0002] With the rapid development of mobile communications, various new applications such as virtual / augmented reality, facial recognition, and autonomous driving are experiencing explosive growth, placing stringent demands on ultra-low latency and high reliability. While these new applications bring immense convenience to social life and production, the amount of data and energy consumed for computation and transmission is increasing at an astronomical rate, placing a heavy processing burden on devices. Cloud-network-edge-device converged networks can achieve integrated applications of communication computing and resource collaboration, effectively alleviating the heavy burden on devices for communication computing and transmission. Therefore, research on cloud-network-edge-device converged networks is of great significance.

[0003] In cloud-network-edge-device networks, computation and communication interact, making resource optimization a highly complex and difficult problem to solve. Furthermore, the network contains diverse edge nodes, with significant differences in communication and computing capabilities and load conditions among different nodes. Therefore, leveraging node collaboration in converged cloud-network-edge-device networks to improve network efficiency and performance, and to achieve efficient resource utilization and sharing, is a highly worthy research topic. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of the aforementioned background technology by providing a method for task offloading and resource allocation in cloud-network-edge-device networks. This method employs a solution based on the maximum-minimum criterion to construct a cloud-network-edge-device network model and proposes a strategy to address the significant differences in communication computing capabilities and load conditions between different nodes, thereby achieving the invention's objective of optimal task offloading and resource allocation.

[0005] To achieve the above-mentioned objectives, the present invention employs the following technical solution:

[0006] Consider M edge servers, I user terminals, and one remote cloud server. Each user terminal has a computing task to process. User terminals can choose to process the task locally or upload it to the MEC server. When a user terminal sends a task to the MEC server, the MEC server can process the task itself, forward it to other MEC servers with more computing resources, or further offload the task to the cloud server for processing. To address the issue of unfair user energy efficiency, with the goal of minimizing user energy efficiency, joint offloading decisions and resource allocation ensure fairness in resource acquisition for users.

[0007] The cloud-network edge network task offloading and resource allocation method of the present invention includes the following three steps:

[0008] Step 1: Construct a cloud-network-edge-device network collaboration model. Users can choose to process their computing tasks locally, via the MEC server, or on the cloud server. When user terminal i sends a task to MEC server m, the channel gain is expressed as... Where g i，m d is the channel power gain coefficient when user terminal i sends a task to MEC server m. i，m Let g be the distance between user terminal i and MEC server m, and α be the path loss factor. Assume the user's movement speed is very slow during unloading. i，m This can be considered a constant; the uplink transmission rate between end user terminal i and MEC server m is:

[0009]

[0010] Where B is the available spectrum bandwidth, σ 2 It is noise power, p i It is the uplink transmit power of user terminal i.

[0011] User terminal i local computing power f i L The power consumed during local calculation is represented by P. i L This means that, during local processing, the computation latency can be expressed as...

[0012]

[0013] Energy consumption can be expressed as

[0014]

[0015] When user terminal i sends a task to MEC server m for processing, the computing power of MEC server m is represented by F. m This indicates that the computing resources allocated to user terminal i are f. i，m This means that when user terminal i sends a task, the transmission delay can be expressed as...

[0016]

[0017] During transmission, the transmission energy consumed by user terminal i can be expressed as:

[0018]

[0019] At this point, the computational latency of MEC server m for task processing can be expressed as:

[0020]

[0021] When the task of user terminal i is offloaded to the cloud server for processing, the cloud computing power is used by f. i c This means that the latency of user terminal i's task being processed on the cloud server can be expressed as:

[0022]

[0023] After the task is completed, the size of the returned data is much smaller than the size of the data before processing, so the transmission delay of the task result return is negligible.

[0024] Step 2: Define the user energy efficiency function as follows

[0025]

[0026] The time constraints for unloading and computation processes are as follows: Base station resource allocation constraints are This indicates that the base station's computing resources cannot exceed its maximum computing capacity, and the offloading decision constraint is... x i y i，m , z i ∈{0,1} indicates that the computation task can only be processed by selecting one node.

[0027] Step 3: By introducing generalized fraction theory and slack variables, the optimization problem is transformed into an equivalent convex optimization problem. Then, an iterative algorithm is proposed to find the optimal solution. The proposed algorithm and a non-cooperative scheme are selected for simulation experiments for comparison. The simulation results show that as the number of iterations increases, the proposed method can converge to a stable value, has better performance, and can effectively ensure user fairness.

[0028] The present invention adopts the above technical solution and has the following beneficial effects: The present invention comprehensively considers the offloading decision and resource allocation under the cloud-network-edge-device network, takes the maximum and minimum user energy efficiency as the objective function, and takes the task processing latency, base station computing capacity and user offloading mode as constraints. By utilizing the cooperation between nodes, the network efficiency and performance are improved, and the high-efficiency utilization and sharing of resources are achieved. Attached Figure Description

[0029] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0030] Figure 1 This is a cloud-network-edge-device network model;

[0031] Figure 2 This is a schematic diagram of the iterative results of the proposed method;

[0032] Figure 3A schematic diagram illustrating how the proposed method ensures energy efficiency fairness for users. Detailed Implementation

[0033] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0034] Figure 1 The cloud-network-edge-device network model consists of M edge servers, I user terminals, and one remote cloud server. The set of edge servers can be represented as M = {1, 2, ..., m, ..., M}, and the set of user terminals can be represented as I = {1, 2, ..., i, ..., I}.

[0035] User terminal i has a computational task that needs to be processed, using Q. i =(L i K i The computational task is described, L i K represents the data size of the task. i This indicates the number of CPU cycles required to complete the terminal task. User terminal i can choose to process the task locally or offload it to an MEC server or a cloud server for processing. When the user terminal sends a task to an MEC server, the MEC server can process the task itself, forward the task to other MEC servers with more computing resources, or further offload the task to a cloud server for processing.

[0036] Define a binary variable x i y i，m , z i x ∈{0,1} i =1 indicates that the task of user terminal i is processed locally, x i =0 indicates that the task of user terminal i is not processed locally; y i，m =1 indicates that the task of user terminal i is processed on MEC server m, y i，m =0 indicates that the task of user terminal i is not processed on MEC server m; z i =1 indicates that the task of user terminal i is processed on the cloud server, z i =0 indicates that the task of user terminal i is not processed on the cloud server.

[0037] To ensure fairness in user energy efficiency, a task unloading optimization problem was constructed based on the maximum and minimum user energy efficiency. The optimization problem is specifically described as follows:

[0038]

[0039]

[0040]

[0041]

[0042] x i y i，m , z i ∈{0,1} (13)

[0043] Formula (9) represents the objective function of the problem, constraint (10) represents the user terminal's requirement for the delay of the task offloading process, constraint (11) represents that the base station's computing resources cannot exceed the maximum computing capacity, and constraints (12) and (13) represent that the computing task can only be processed by one node. Let variable Q represent the value of this optimization problem, i.e., the maximum and minimum energy efficiency. Through generalized fraction theory and the introduction of the relaxation variable θ, the binary variable is relaxed to x. i y i，m , z i ∈[0,1], the optimization problem is transformed into

[0044]

[0045]

[0046]

[0047]

[0048] x i y i，m , z i ∈[0,1] (18)

[0049]

[0050] The problem is solved by using an iterative algorithm. First, the initial value Q = 0 is set and substituted into the optimization problem (14) to obtain the optimal solution. Then, the value of Q is updated according to the optimal solution. The convergence is judged according to the updated Q value. If convergence is achieved, the iteration is stopped; otherwise, the iteration continues.

[0051] Figure 2 The proposed algorithm and a non-cooperative approach were compared through simulation experiments. The horizontal axis represents the number of iterations, and the vertical axis represents energy efficiency. The simulation results show that as the number of iterations increases, the proposed method converges to a stable value and exhibits better performance than the non-cooperative approach.

[0052] Figure 3This comparison focuses on the scenarios considering fairness and those not considering fairness. The difference lies in the objective function: without fairness, the objective function is to maximize total energy efficiency, i.e., maximizing the ratio of total user rate to total user energy consumption. The figure shows that, compared to the scenario without fairness, the energy efficiency gap between the best and worst users is relatively smaller when fairness is considered. Therefore, the proposed method better ensures fairness in resource acquisition among users.

[0053] The above specific embodiments further illustrate the inventive purpose, technical solution and beneficial effects of the present invention. It should be understood that the above specific embodiments are only illustrative examples and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions or alterations made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for task offloading and resource allocation in a cloud-network-edge-device network, characterized in that, Construct a cloud-network-edge-device network collaboration model, comprising a communication model and a computing model, describing the user offloading decision and task processing methods; construct the cloud-network-edge-device network model, where user terminal i has a computing task to process, represented by Q. i =(L i K i The computational task is described, L i K represents the data size of the task. i This represents the number of CPU cycles required to complete the terminal task; user terminal i can choose to process the task locally or offload it to an MEC server or a cloud server for processing; when the user terminal sends a task to an MEC server, the MEC server can process the task itself, forward the task to other MEC servers with richer computing resources, or further offload the task to a cloud server for processing; define a binary variable x. i y i,m ,z i x ∈{0,1} i =1 indicates that the task of user terminal i is processed locally, x i =0 indicates that the task of user terminal i is not processed locally; y i，m =1 indicates that the task of user terminal i is processed on MEC server m, y i，m =0 indicates that the task of user terminal i is not processed on MEC server m; z i =1 indicates that the task of user terminal i is processed on the cloud server, z i =0 indicates that the task of user terminal i is not processed on the cloud server; The uplink transmission rate between end user terminal i and MEC server m is: Where B is the available spectrum bandwidth, σ 2 It is noise power, p i It is the uplink transmit power of user terminal i. It is the channel gain, where g i，m d is the channel power gain coefficient when user terminal i sends a task to MEC server m. i，m α is the distance between user terminal i and MEC server m, and α is the path loss factor. User terminal i local computing power f i L This indicates that the power consumed during local calculations is represented by P. i L This means that, during local processing, the computation latency can be expressed as... Energy consumption can be expressed as When user terminal i sends a task to MEC server m for processing, the computing power of MEC server m is represented by F. m This indicates that the computing resources allocated to user terminal i are f. i，m This means that when user terminal i sends a task, the transmission delay can be expressed as... During transmission, the transmission energy consumed by user terminal i can be expressed as: At this point, the computational latency of MEC server m for task processing can be expressed as: When the task of user terminal i is offloaded to the cloud server for processing, the cloud computing power is used by f. i c This means that the latency of user terminal i's task being processed on the cloud server can be expressed as: After the task is completed, the size of the returned data is much smaller than the size of the data before processing, so the transmission delay of the task result return is negligible. Define the user energy efficiency function as follows The time constraints for unloading and computation processes are as follows: Base station resource allocation constraints are Unloading decision constraints are And satisfy x i y i,m ,z i ∈{0,1} indicates that the computing task can only be processed by one node; based on the user energy efficiency function and constraints, the offloading decision and resource allocation are jointly performed.

2. The method for cloud-network edge-end network task offloading and resource allocation according to claim 1, characterized in that, In step 3, the optimization problem is transformed into an equivalent convex optimization problem by introducing generalized fractional theory and slack variables. Let variable Q represent the value of the optimization problem, i.e., the maximum and minimum energy efficiency. Through generalized fractional theory and the introduction of slack variables θ, the bivariate variable is relaxed to x. i y i,m ,z i The optimization problem is transformed into {0, 1}. x i ,y i，m ,z i ∈{0,1} Then, an iterative algorithm is proposed to find the optimal solution to the problem. First, the initial value Q = 0 is set and substituted into the optimization problem to obtain the optimal solution. Then, the value of Q is updated according to the optimal solution. The convergence is judged according to the updated Q value. If convergence is achieved, the iteration stops; otherwise, the iteration continues. Finally, the algorithm is verified by simulation.

Images