Computing power network task unloading method and device

Abstract

The embodiment of the invention discloses a method and a device for processing a computing power network task, which are used for improving the reasonability of unloading the computing power network task so as to improve the utilization rate of computing power resources. The scheme provided by the invention comprises the following steps: acquiring topological information of a computing power network, wherein the topological information comprises respective local computing power resource information of a plurality of computing power equipment nodes in communication connection and respective generated computing power network task information; constructing a computing power model based on the topological information of the computing power network, wherein the computing power model represents computing power resource cost consumed by processing computing power network tasks in the computing power network through a task processing strategy; iteratively optimizing a task processing strategy through a simulated annealing algorithm by taking the computing power resource cost of a minimum computing power model as an optimization target; and processing the computing power network task in the computing power network based on the task processing strategy after iterative optimization.

Description

Technical Field

[0001] The present application relates to the field of computing power networks, and in particular to a method and device for offloading computing power network tasks. Background Art

[0002] The computing network can meet the various task requirements of users in practical applications. In the computing network, it is necessary to reasonably plan the unloading strategy of user tasks and the allocation strategy of resources, so as to connect computing resources with user needs, maximize the value of computing resources, and avoid the waste of computing resources. Improper decisions will not only seriously affect the user experience, but also may lead to increased network traffic, and in severe cases, even bring unpredictable delays and energy consumption to the region.

[0003] How to improve the rationality of computing network task offloading to improve computing resource utilization is the technical problem to be solved by this application. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a method and device for offloading computing power network tasks, so as to improve the rationality of offloading computing power network tasks and improve the utilization rate of computing power resources.

[0005] In a first aspect, a method for processing a computing network task is provided, comprising: Acquire topology information of the computing power network, wherein the topology information includes local computing power resource information of each of the plurality of computing power device nodes in communication connection and computing power network task information generated by each of the plurality of computing power device nodes; A computing power model is constructed based on the topological information of the computing power network, wherein the computing power model represents the computing power resource cost consumed by processing the computing power network tasks in the computing power network through the task processing strategy, wherein the task processing strategy includes using the local computing power resources of the computing power device node to perform processing on at least part of the computing power network tasks generated by the computing power device node itself and / or offloading at least part of the computing power network tasks generated by the computing power device node itself to an adjacent node for processing; Taking minimizing the computing resource cost of the computing model as the optimization goal, the task processing strategy is iteratively optimized by a simulated annealing algorithm; The computing power network tasks in the computing power network are processed based on the iteratively optimized task processing strategy.

[0006] In a second aspect, a device for processing a computing network task is provided, comprising: An acquisition module is used to acquire topological information of a computing power network, wherein the topological information includes local computing power resource information of each of a plurality of computing power device nodes in communication connection and computing power network task information generated by each of them; A construction module, which constructs a computing power model based on the topological information of the computing power network, wherein the computing power model represents the computing power resource cost consumed by processing the computing power network tasks in the computing power network through the task processing strategy, wherein the task processing strategy includes using the local computing power resources of the computing power device node to perform processing on at least part of the computing power network tasks generated by the computing power device node itself and / or offloading at least part of the computing power network tasks generated by the computing power device node itself to an adjacent node for processing; An optimization module, which takes minimizing the computing resource cost of the computing model as an optimization goal and iteratively optimizes the task processing strategy through a simulated annealing algorithm; The processing module processes the computing power network tasks in the computing power network based on the iteratively optimized task processing strategy.

[0007] According to a third aspect, an electronic device is provided. The electronic device comprises a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the steps of the method according to the first aspect are implemented.

[0008] According to a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method according to the first aspect are implemented.

[0009] In a fifth aspect, a computer program product is provided, the computer program product comprising a non-transitory computer-readable storage medium storing a computer program, the computer program being operable to cause a computer to execute part or all of the steps of the method of the first aspect.

[0010] In an embodiment of the present application, first, the topology information of the computing power network is obtained, and the topology information includes the local computing power resource information of each of the multiple computing power device nodes that are communicatively connected and the computing power network task information generated by each of them; then, a computing power model is constructed based on the topology information of the computing power network, and the computing power model represents the computing power resource cost consumed by processing the computing power network tasks in the computing power network through the task processing strategy, wherein the task processing strategy includes using the local computing power resources of the computing power device node to perform processing on at least part of the computing power network tasks generated by itself and / or offloading at least part of the computing power network tasks generated by the computing power device node itself to adjacent nodes for processing; then, with minimizing the computing power resource cost of the computing power model as the optimization goal, the task processing strategy is iteratively optimized through a simulated annealing algorithm; finally, the computing power network tasks in the computing power network are processed based on the iteratively optimized task processing strategy. This solution can build a computing model for the local computing resource information and computing network task information of each computing device node in the computing network. The model can be used to associate local computing resources, computing network tasks, and the topological structure of the computing network, and the computing resource cost consumed by processing computing network tasks can be expressed. Then, based on the computing model, the task processing strategy is iteratively optimized through the simulated annealing algorithm to fully utilize the computing resources in the computing network and reduce the computing resource cost consumed by processing computing network tasks. Finally, the iteratively optimized task processing strategy is used to execute and process computing network tasks, improve the rationality of computing network task unloading, improve computing resource utilization, and achieve cost reduction and efficiency improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 This is one of the flowcharts of a method for processing a computing network task in an embodiment of the present application; Figure 2 It is a schematic diagram of a computing network of a method for processing a computing network task in an embodiment of the present application; Figure 3 This is a second flow chart of a method for processing a computing network task in an embodiment of the present application; Figure 4 This is a flowchart of a method for processing a computing network task according to an embodiment of the present application; Figure 5 This is a fourth flow chart of a method for processing a computing network task according to an embodiment of the present application; Figure 6 This is a flowchart of a method for processing a computing network task according to an embodiment of the present application; Figure 7 It is a structural diagram of a device for processing computing network tasks according to an embodiment of the present application. DETAILED DESCRIPTION

[0012] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application. The numbering of the drawings in this application is only used to distinguish the various steps in the scheme, and is not used to limit the execution order of the various steps. The specific execution order is subject to the description in the specification.

[0013] In a computing network, computing network tasks generated by computing device nodes can be executed locally or offloaded to adjacent nodes in the computing network. In practical applications, for example, they can be offloaded to cloud servers or edge servers for execution. Different execution modes often consume different computing resources and processing time. In practical applications, different computing devices in a computing network have different computing capabilities, and the transmission distances between different computing network devices are also different. The above processing capabilities and transmission distances will limit the actual processing methods of computing network tasks. Unreasonable task processing strategies may result in a high consumption of computing resource costs in the computing network, and the computing network task processing efficiency is low, with the defect of low computing resource utilization.

[0014] In order to solve the problems existing in the prior art, the present application embodiment provides a method for processing computing network tasks, such as Figure 1 As shown, including: S11: Acquire topology information of the computing power network, where the topology information includes local computing power resource information of each of the multiple computing power device nodes that are communicatively connected and computing power network task information generated by each of them.

[0015] In this step, the topology information of the computing power network is obtained. The topology information may include the communication connection relationship between multiple computing power device nodes, and the topology information includes the local computing power resource information and computing power network task information corresponding to the computing power device node. Among them, the local computing power resource information may specifically include the available amount of computing power resources of the local computing power device node, the type of computing power network task that can be processed locally, etc. The local computing power resource information is used to indicate the ability of the corresponding computing power device node to process the computing power network task. The computing power network task information may include a variety of information such as the type of computing power network task, the amount of data, and the processing method. The computing power network task information can be used to indicate the processing conditions required for the computing power network task.

[0016] Optionally, in actual applications, the computing power device nodes included in the computing power network may be of multiple types. For example, the computing power network may include user device nodes and server nodes. Alternatively, it may also include other types of nodes such as relay nodes. Among them, the user device node can generate a computing power network task corresponding to the user's needs, and the computing power network task can be offloaded to the server node in the computing power network for processing. In this step, the local computing power resource information obtained is used to indicate the ability of the corresponding computing power device node to process tasks, and the computing power network task information is used to indicate the processing conditions required for the computing power network task. The topological information of the computing power network can comprehensively characterize the ability of each computing power device node to process tasks and the processing conditions for generating tasks, thereby providing a data basis for the subsequent construction of a computing power model.

[0017] S12: Construct a computing power model based on the topological information of the computing power network, wherein the computing power model represents the computing power resource cost consumed by processing the computing power network tasks in the computing power network through the task processing strategy, wherein the task processing strategy includes using the local computing power resources of the computing power device node to perform processing on at least part of the computing power network tasks generated by itself and / or offloading at least part of the computing power network tasks generated by the computing power device node itself to an adjacent node for processing.

[0018] In this step, a computing power model is constructed based on the topological information obtained in the above steps to represent the computing power resource cost consumed by the computing power network tasks generated by the nodes in the computing power network to execute and process in the computing power network.

[0019] For example, a topology diagram can be first constructed based on the communication connection relationship between multiple computing power device nodes in the computing power network. The nodes in the topology diagram correspond to the multiple computing power device nodes in the computing power network, and the lines between the nodes represent the communication connection relationship between the computing power device nodes. Then, a computing power model is constructed based on the topology diagram, wherein local computing power resource information and computing power network task information are added to the nodes in the topology diagram to characterize the task processing capabilities of the corresponding computing power device nodes and the processing requirements for generating tasks. Information such as the data transmission method and transmission rate are added to the lines between the nodes in the topology diagram to characterize the data transmission cost. Thus, the constructed computing power model can characterize the task processing performance of the computing power device nodes in the computing power network, the processing condition requirements for the computing power device nodes to generate tasks, and the related information transmitted between nodes.

[0020] In the computing power model, the computing power network tasks generated by the computing power device node itself can be processed locally or offloaded to adjacent nodes for processing. Among them, for tasks offloaded to adjacent nodes, the adjacent node can also choose to process locally or offload to adjacent nodes again. For any computing power network task to be processed, it can be split according to demand, and the split parts can be processed on different nodes. The above-mentioned strategy for whether to split and offload the computing power network task is the task processing strategy described in this step. In the computing power network, assuming that the computing power network tasks generated by each computing power device node are fixed, the computing power resource costs consumed by executing task processing using different task processing strategies are often different. Further, in actual applications, the computing power network tasks generated by each computing power device node in the computing power network may be non-fixed, making it difficult to determine the optimal task processing strategy. In this step, by constructing a computing power model, the computing power resource cost consumed by processing the computing power network tasks through the task processing strategy is expressed, providing a data basis for optimizing the processing strategy in subsequent steps.

[0021] S13: Taking minimizing the computing resource cost of the computing model as the optimization goal, iteratively optimize the task processing strategy through a simulated annealing algorithm.

[0022] In this step, based on the above computing power model, a simulated annealing algorithm is used to perform iterative optimization on the task processing strategy to minimize the computing power resource cost of the computing power model.

[0023] Among them, the simulated annealing algorithm is a global search optimization method that is highly efficient in solving large-scale and complex combinatorial problems. In this solution, the simulated annealing algorithm is used to perform solution and optimization iterations on the task processing strategy of the above computing model. This effectively reduces the computing resource cost consumed by processing tasks while ensuring that the task processing strategy can complete the computing network tasks in the computing network.

[0024] S14: Processing the computing power network tasks in the computing power network based on the iteratively optimized task processing strategy.

[0025] In this step, the task processing strategy after the above-mentioned simulated annealing algorithm optimization iteration is applied to the computing power network task, instructing the computing power device nodes in the computing power network to perform task processing and unloading according to the task processing strategy, thereby ensuring that the computing power network tasks are processed while effectively reducing the computing power resource cost consumed by processing tasks, improving the rationality of unloading computing power network tasks, and improving the utilization rate of computing power resources.

[0026] The present invention is further described below with reference to an example.

[0027] The solution provided by the embodiment of the present application can be applied to a variety of computing networks, for example, it can be applied to a computing network including multiple edge nodes and multiple user equipment (UE) nodes. In the computing network of this example, the user equipment UE can use the neighboring UE as an unloading relay node (RN) to perform task unloading, and the neighboring UE can further unload the task to the edge node. Any edge node can be deployed with a mobile edge computing (MEC) server, thereby realizing the unloading processing of computing network tasks. Give full play to the horizontal collaboration and resource sharing capabilities between adjacent nodes and improve the overall resource utilization of the computing network.

[0028] Specifically, different edge nodes can be configured with different MEC servers, which can provide rich computing resources to handle various computing tasks. When performing task offloading, it is possible to determine whether the MEC server has the ability to handle a specific type of computing network task based on the local computing resource information of the MEC server, thereby offloading the computing network task to the MEC server that is capable of handling it.

[0029] In practical applications, an application task can also be divided into several segments, and the capacity of each segment can be different, so that multiple adjacent RNs can be selected for offloading. Through the UE-RN-MEC three-level collaborative partial offloading method, tasks can be offloaded from resource-constrained terminal UEs to other computing entities. Furthermore, the simulated annealing algorithm is used to minimize the computing resource cost, optimize the iterative task processing strategy, and achieve efficient computing network task processing.

[0030] Through this solution, a computing model can be constructed for the local computing resource information and computing network task information of each computing device node in the computing network. The model can be used to associate local computing resources, computing network tasks, and the topological structure of the computing network, and the computing resource cost consumed by processing computing network tasks can be expressed. Then, based on the computing model, the task processing strategy is iteratively optimized through the simulated annealing algorithm to fully utilize the computing resources in the computing network and reduce the computing resource cost consumed by processing computing network tasks. Finally, the iteratively optimized task processing strategy is used to execute and process computing network tasks, and the rationality of computing network task unloading is improved to improve the utilization rate of computing resources and achieve cost reduction and efficiency improvement.

[0031] Based on the solution provided in the above embodiment, optionally, the multiple computing power device nodes include at least one user device, at least one unloading relay node and at least one edge server, wherein any unloading relay node is communicatively connected with at least one user device and at least one edge server; The computing power resource cost includes the computing power resource cost occupied by processing the computing power network task and the time resource cost consumed.

[0032] Figure 2 This is a schematic diagram of a computing network of an embodiment of the present application. In the computing network of this example, U UE user devices and E MEC servers are included, and each computing device node can be randomly distributed in multiple communication ranges. Among them, each task generated by the UE can select one or more RN relay devices and unload through communication methods such as D2D (Device-to-Device) communication links. When tasks are unloaded through multiple RNs, multiple unloaded tasks can be executed in parallel to improve the processing efficiency of the tasks. Among them, on the one hand, the RN can perform computational processing on the task segment unloaded by the UE, and on the other hand, the RN can also choose to further unload the task segment to the MEC server through the cellular link for computational processing.

[0033] Among them, any UE can be expressed as: , Indicates the communication interval where the UE is located. Indicates the computing resources of the UE.

[0034] The tasks generated by UE can be expressed as: , is the input data volume of the task, Indicates the number of CPU instruction cycles required to calculate 1 bit of this type of task. Indicates the maximum delay allowed for the task. Optionally, the above task can be split into multiple tasks for computation offloading.

[0035] Multiple relay nodes RN in the computing power network can be expressed as a set: .

[0036] In any communication range, covering at least one MEC server can be expressed as: , is the communication range of the MEC server, The computing resources of the MEC server.

[0037] Based on the computing power network of the embodiment of the present application, the tasks generated by the user equipment UE can be offloaded through the local processing stage, the RN processing stage, and the MEC processing stage.

[0038] In the local processing phase, the tasks generated by the user equipment UE The proportion of tasks executed locally is expressed as The proportion of unloading to a certain adjacent relay node r is expressed as , then the overall proportion of offloading to each RN is expressed as .

[0039] In the RN processing stage, the proportion of tasks performed by a relay node r for local computation is expressed as , the proportion of relay node r unloading to a certain edge node MEC processor e is expressed as , then the overall proportion of RN offloading to each MEC server is expressed as .

[0040] In the MEC server processing phase, the relay node r offloads the task to a certain edge node MEC processor e This is completed by the MEC server e, where: Indicates whether to choose to offload to MEC server e.

[0041] Based on the above information, the task processing strategy of task u can be expressed as , and satisfy the following formula:

[0042]

[0043]

[0044] In addition, in the computing network, unloading tasks requires the transmission of execution tasks, which will consume a certain amount of time, and the computing device node will also consume a certain amount of time to execute and process the tasks. In the embodiment of the present application, the time consumed by task transmission is called transmission delay, and the time consumed by the computing device node to process the task is called calculation delay. The time resource cost includes the time resource cost corresponding to the above-mentioned transmission delay and calculation delay.

[0045] Among them, the execution delay of the local device task can be expressed as: , .in, It is the local computing resource allocated by the user device to the task.

[0046] The communication delay caused by offloading the task of the user equipment to r can be expressed as: .in, is the transmission rate from task u to r.

[0047] The computation delay on the relay device r can be expressed as: , , .in, It is the local computing power resource allocated to the task by the relay device r.

[0048] The communication delay caused by offloading the task of the user equipment to the edge server e can be expressed as: .in, is the transfer rate of tasks from r to e.

[0049] The time to execute a subtask on edge server e can be expressed as: , .

[0050] In practical applications, after the task is offloaded, the data after execution will be transmitted back to the node that initiated the task. Considering that the data after the offload execution is much smaller than the uploaded data, the delay of data transmission can be ignored. In the embodiment of the present application, all subtasks in the offload scenario can be executed in parallel to improve the execution efficiency. Therefore, the offload delay of the task can be expressed as: , .

[0051] In actual applications, different computing power device nodes have different energy consumption coefficients. When executing tasks with the same amount of data, the energy consumption of edge servers with powerful computing power is much smaller than that of terminal devices. In addition, the transmission of tasks also requires energy consumption. Therefore, the computing power resource cost corresponding to the computing power model can be expressed as:

[0052]

[0053]

[0054]

[0055]

[0056]

[0057] in, , They are respectively the energy consumption of executing tasks in the local UE, the transmission energy consumption of UE offloading to relay r, the energy consumption of local execution of relay r, the transmission energy consumption of r offloading to MEC server e, and the energy consumption of e executing tasks locally; , are energy consumption coefficients; , are the transmit powers of UE and relay r respectively.

[0058] In the example of this application, the computing resource cost of the computing model includes two aspects: time resource cost and computing resource cost. The computing resource costs of these two aspects are used as the optimization target of the simulated annealing algorithm in the subsequent steps. The optimization target of minimizing the computing resource cost can be expressed as: (1) (2) (3) (4) (5) Through the solution provided in the embodiment of the present application, for a computing power network including "user equipment UE-offloading relay node RN-edge server MEC", the computing power resource cost occupied by processing computing power network tasks and the time resource cost consumed are jointly used as the computing power resource cost, and the resources consumed by the task are comprehensively represented from the time dimension and the physical resource quantity dimension, thereby improving the rationality of the optimization target of the simulated annealing algorithm, thereby improving the rationality of the task processing strategy after iterative optimization.

[0059] Based on the solution provided in the above embodiment, optionally, Figure 3 As shown, in the above step S13, the task processing strategy is iteratively optimized by a simulated annealing algorithm with the goal of minimizing the computing resource cost of the computing model, including: S31: Constructing a candidate processing strategy for the computing power network task based on the computing power model, wherein the candidate processing strategy includes offloading at least part of the computing power network task generated by at least one user device to an edge server for execution through an offloading relay node.

[0060] In the embodiment of the present application, solving the task processing strategy of the computing network is an NP-hard problem, and it is difficult to obtain the absolute optimal solution by direct calculation. This solution uses a simulated annealing algorithm to perform iterative optimization to determine the optimal task processing strategy for efficient use of computing network resources.

[0061] In this step, based on the computing power network tasks generated by the computing power device nodes in the above computing power model, a candidate processing strategy that can process the computing power network tasks is constructed. The candidate processing strategy is used as the initial solution for iterative optimization in subsequent steps to perform iterative optimization.

[0062] Among them, in the candidate processing strategies, there is at least one computing power network task generated by a user device, at least part of which is unloaded to the edge server through an unloading relay node for processing. Optionally, at least one computing power network task is split, and one or more task segments after the split are unloaded to an adjacent unloading relay node, and then the relay node further unloads one or more task segments to the edge server for processing. During the processing of a computing power network task, any computing power device node in the path can process at least part of the task through local computing power resources, and can also unload unprocessed tasks to adjacent nodes. Among them, if the edge server at the end of the computing power network does not have an adjacent node that can be unloaded, it is necessary to complete the processing of all received tasks.

[0063] S32: Taking minimizing the computing resource cost of the computing model as the optimization goal, perform iterative optimization on the candidate processing strategies through a simulated annealing algorithm until the simulated annealing algorithm converges.

[0064] In this step, the optimization goal of the simulated annealing algorithm is to minimize the computing resource cost determined in any of the above examples, and iterative optimization is performed on the above initial solution.

[0065] In practical applications, before performing iterative optimization, the initial temperature is set for the simulated annealing algorithm. The candidate processing strategy constructed in the above steps is expressed as the initial solution: , Each component of is the offloading ratio of the task in the local, relay device, and edge server. In addition, the initial cooling rate of the simulated annealing algorithm is set , acceptance rate threshold , the number of iterations for each T value L. Set the objective function value of the simulated annealing algorithm to .in, and is the weight coefficient for controlling delay and energy consumption cost, . is the delay of the entire unloading model, is the energy consumption of the entire unloading model.

[0066] Based on the above parameters, the objective function value is used to reduce the time resource cost and computing power resource cost as the optimization goal, and the above candidate processing strategies are iteratively optimized through the simulated annealing algorithm until the simulated annealing algorithm meets the preset convergence conditions.

[0067] S33: Determine the task processing strategy after iterative optimization based on the converged candidate processing strategy.

[0068] In this step, the candidate processing strategy after the simulated annealing algorithm converges can be determined as the task processing strategy execution output after iterative optimization. The task processing strategy after iterative optimization can process the computing power network tasks in the computing power network, and at the same time can make full use of the physical resources and time resources in the computing power network, improve the rationality of computing power network task unloading, and improve the utilization rate of computing power resources.

[0069] Based on the solution provided in the above embodiment, optionally, Figure 4 As shown, in the above step S32, the optimization goal is to minimize the computing resource cost of the computing model, and the candidate processing strategies are iteratively optimized by the simulated annealing algorithm until the simulated annealing algorithm converges, including: S41: Constructing a neighborhood processing strategy based on the candidate processing strategy.

[0070] In this step, the initial solution corresponding to the above candidate processing strategy is , corresponding to the calculation Neighborhood of . Optionally, for the initial solution Apply disturbances, such as Add a uniform random perturbation or a Gaussian perturbation to at least one component of The neighborhood solution of is used as the neighborhood processing strategy.

[0071] S42: Determine whether to accept the neighborhood processing strategy based on the acceptance condition of the simulated annealing algorithm.

[0072] In this step, the newly constructed neighborhood processing strategy can be judged according to the constraint conditions and the acceptance conditions. The constraint conditions can specifically include the constraint conditions of formula (2) to formula (5) in the above embodiment. The acceptance conditions can specifically include If the new solution satisfies the above acceptance conditions and constraints, the new solution is used to iterate the current solution, otherwise the new solution is used according to the probability Determine whether to use the new solution to iterate the current solution, where is the cost function, and T is the current temperature.

[0073] S43: performing iterative optimization on the candidate processing strategies based on the determination result of the neighborhood processing strategies until the simulated annealing algorithm converges.

[0074] Among them, after the initial solution generates the neighborhood solution for the first time, the current solution is the initial solution, and the initial solution is iterated based on the judgment result of the neighborhood solution. The solution after each iteration optimization is used as the current solution to perform the next iteration. After the simulated annealing algorithm converges, the current solution after the latest iteration is output.

[0075] In the simulated annealing algorithm, if the preset number of iterations L for each T value is reached, the T value is cooled to the next temperature according to the cooling rate, and then the cooling is performed again after the temperature is iterated L times again until the temperature is cooled to the preset minimum temperature, and then the simulated annealing algorithm is determined to have converged.

[0076] Through the solution provided in the embodiment of the present application, the candidate processing strategies are optimized and iterated by the simulated annealing algorithm. During the optimization and iteration process, whether to accept the neighborhood processing strategy is determined according to the acceptance conditions, so that the candidate processing strategy can be iterated towards the optimal state as a whole. And there is a certain probability of iterating the neighborhood solution with poor quality to avoid the local optimum, which is conducive to making the solution output after convergence globally optimal.

[0077] Based on the solution provided in the above embodiment, optionally, Figure 5 As shown, after the above step S42, that is, after determining whether to accept the neighborhood processing strategy through the acceptance condition of the simulated annealing algorithm, it also includes: S51: Determine the neighborhood acceptance rate of the simulated annealing algorithm, wherein the neighborhood acceptance rate is the ratio of the number of neighborhood processing strategies that have been determined to be accepted to the number of neighborhood processing strategies that have been constructed.

[0078] In the present application example, after the judgment of the simulated annealing algorithm is executed, the neighborhood acceptance rate is accumulated according to the judgment result. The calculation method of the acceptance rate can be expressed as: .

[0079] S52: Adjusting the cooling rate of the simulated annealing algorithm based on the neighborhood acceptance rate.

[0080] The simulated annealing algorithm usually adopts a fixed cooling rate, such as linear cooling (the temperature is reduced by a fixed value each time iteratively) or exponential cooling (the temperature is multiplied by a constant less than 1 each time iteratively). The cooling speed of these methods may be too fast or too slow, and cannot adapt well to the complexity of the problem. In the example of the present application, in order to make the algorithm better converge to the global optimal solution, the cooling rate of the simulated annealing algorithm is adjusted based on the above-mentioned neighborhood acceptance rate, so that the cooling rate changes with the actual acceptance rate, and then the temperature cooling process of the simulated annealing algorithm changes with the effect of iterative optimization, thereby accelerating the convergence of the simulated annealing algorithm while ensuring the effect of iterative optimization, and improving the overall efficiency of iterative optimization.

[0081] Based on the solution provided in the above embodiment, optionally, Figure 6 As shown, in the above step S52, adjusting the cooling rate of the simulated annealing algorithm based on the neighborhood acceptance rate includes: S61: if the neighborhood acceptance rate is greater than a first preset acceptance rate, reducing the value of the cooling rate; S62: If the neighborhood acceptance rate is less than a second preset acceptance rate, increasing the value of the cooling rate, wherein the second preset acceptance rate is less than or equal to the first preset acceptance rate.

[0082] In the solution provided in the embodiment of the present application, the cooling rate is adjusted based on the acceptance rate. Specifically, assuming that the first preset acceptance rate is equal to the second preset acceptance rate, it is expressed as , then if > , then the algorithm is considered to search well, and the (For example ) to speed up cooling. < , then the algorithm needs more exploration, so increase (For example ) to slow down cooling.

[0083] Thus, in the cooling step, the cooling rate is adjusted Update Temperature: .

[0084] Optionally, the first preset acceptance rate may be greater than the second preset acceptance rate. In this case, if the neighborhood acceptance rate is greater than or equal to the second preset acceptance rate and less than or equal to the first preset acceptance rate, the current cooling rate is kept unchanged, and the cooling coefficient is adjusted only when the acceptance rate is less than the second preset acceptance rate or greater than the first preset acceptance rate.

[0085] The solution provided in the embodiment of the present application can be used to use neighboring UE as an offloading relay node (RN) between the local UE and the edge node, and to offload tasks from resource-constrained terminals to other computing entities through the UE-RN-MEC three-level collaborative partial offloading method, and jointly optimize the two offloading targets of task offloading delay and energy consumption. Combined with the characteristics of the model, a task offloading model that minimizes total delay and total energy consumption under delay constraints is proposed, and an adaptive simulated annealing algorithm based on acceptance rate is proposed to solve the model, and the resource allocation scheme for collaborative offloading, the selection of RN and MEC, and the task segmentation ratio are determined.

[0086] This solution applies a multi-objective collaborative task partial offloading and resource allocation model of multiple edge nodes and multiple UEs. It considers the scenario where a group of UEs and a group of edge nodes are adjacent, and uses the neighboring UEs as offloading relay nodes (RN) between the local UE and the edge nodes, making full use of the computing resources of the surrounding terminals to reduce latency. The task is offloaded from resource-constrained terminals to other computing entities through the UE-RN-MEC three-level collaborative partial offloading method. For divisible tasks, combined with the characteristics of the model, a task offloading model that minimizes total latency and total energy consumption under latency constraints is proposed.

[0087] This scheme aims at the multi-objective optimization problem. By introducing the weight coefficients of control delay and energy consumption cost, it transforms it into a single-objective optimization, and proposes an adaptive simulated annealing algorithm based on acceptance rate to solve the model, and determines the resource allocation scheme of collaborative offloading, the selection of RN and MEC, and the task segmentation ratio.

[0088] In the process of iterative optimization, this scheme optimizes the cooling process and introduces the concept of acceptance rate. If the value is higher, the algorithm is considered to be searching better, and the cooling rate value is reduced to speed up the cooling. If it is lower, it means that the algorithm needs more exploration and iterative optimization, so increase the cooling rate value to slow down the cooling.

[0089] In order to solve the problems existing in the related art, the embodiment of the present application also provides a device 70 for processing computing network tasks, such as Figure 7 As shown, including: An acquisition module 71 is configured to acquire topology information of a computing power network, wherein the topology information includes local computing power resource information of each of a plurality of computing power device nodes in communication connection and computing power network task information generated by each of the nodes; A construction module 72 is configured to construct a computing power model based on the topological information of the computing power network, wherein the computing power model represents the computing power resource cost consumed by processing the computing power network tasks in the computing power network through the task processing strategy, wherein the task processing strategy includes using the local computing power resources of the computing power device node to perform processing on at least part of the computing power network tasks generated by the computing power device node itself and / or offloading at least part of the computing power network tasks generated by the computing power device node itself to an adjacent node for processing; An optimization module 73, which takes minimizing the computing resource cost of the computing model as an optimization goal and iteratively optimizes the task processing strategy through a simulated annealing algorithm; The processing module 74 processes the computing power network tasks in the computing power network based on the iteratively optimized task processing strategy.

[0090] Through the device provided in the embodiment of the present application, a computing model can be constructed for the local computing resource information and computing network task information of each computing device node in the computing network, and the local computing resources, computing network tasks and the topological structure of the computing network can be associated through the model to express the computing resource cost consumed by processing the computing network tasks. Then, based on the computing model, the task processing strategy is iteratively optimized through the simulated annealing algorithm to fully utilize the computing resources in the computing network and reduce the computing resource cost consumed by processing the computing network tasks. Finally, the task processing strategy after iterative optimization is used to execute the computing network tasks, and the rationality of unloading the computing network tasks is improved to improve the utilization rate of computing resources and achieve cost reduction and efficiency improvement.

[0091] Among them, the above modules in the device provided by the embodiment of the present application can also implement the method steps provided by the above method embodiment. Alternatively, the device provided by the embodiment of the present application can also include other modules in addition to the above modules to implement the method steps provided by the above method embodiment. And the device provided by the embodiment of the present application can achieve the technical effects that can be achieved by the above method embodiment.

[0092] Preferably, an embodiment of the present application also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the various processes of the above-mentioned method embodiment for processing computing power network tasks are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0093] The embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, each process of the above-mentioned method for processing computing network tasks is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0094] An embodiment of the present application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program. The computer program can be operated to enable a computer to execute part or all of the steps of the above-mentioned method embodiment for processing computing power network tasks, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0095] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0096] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0097] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0098] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0099] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0100] Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0101] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0102] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0103] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0104] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for processing computing network tasks, characterized in that: include: Acquire topology information of the computing power network, wherein the topology information includes local computing power resource information of each of the plurality of computing power device nodes in communication connection and computing power network task information generated by each of the plurality of computing power device nodes; A computing power model is constructed based on the topological information of the computing power network, wherein the computing power model represents the computing power resource cost consumed by processing the computing power network tasks in the computing power network through the task processing strategy, wherein the task processing strategy includes using the local computing power resources of the computing power device node to perform processing on at least part of the computing power network tasks generated by the computing power device node itself and / or offloading at least part of the computing power network tasks generated by the computing power device node itself to an adjacent node for processing; Taking minimizing the computing resource cost of the computing model as the optimization goal, the task processing strategy is iteratively optimized by a simulated annealing algorithm; The computing power network tasks in the computing power network are processed based on the iteratively optimized task processing strategy.

2. The method according to claim 1, characterized in that The multiple computing power device nodes include at least one user device, at least one unloading relay node and at least one edge server, wherein any unloading relay node is communicatively connected with at least one user device and at least one edge server; The computing power resource cost includes the computing power resource cost occupied by processing the computing power network task and the time resource cost consumed.

3. The method according to claim 2, characterized in that Taking minimizing the computing resource cost of the computing model as the optimization goal, the task processing strategy is iteratively optimized through a simulated annealing algorithm, including: Constructing a candidate processing strategy for computing network tasks based on the computing model, wherein the candidate processing strategy includes offloading at least part of the computing network tasks generated by at least one user device to an edge server for execution through an offloading relay node; Taking minimizing the computing resource cost of the computing model as an optimization goal, performing iterative optimization on the candidate processing strategies through a simulated annealing algorithm until the simulated annealing algorithm converges; The task processing strategy after iterative optimization is determined based on the converged candidate processing strategy.

4. The method according to claim 3, characterized in that Taking minimizing the computing resource cost of the computing model as the optimization goal, performing iterative optimization on the candidate processing strategies through a simulated annealing algorithm until the simulated annealing algorithm converges, including: constructing a neighborhood processing strategy based on the candidate processing strategies; Determining whether to accept the neighborhood processing strategy through the acceptance condition of the simulated annealing algorithm; The candidate processing strategies are iteratively optimized based on the determination result of the neighborhood processing strategies until the simulated annealing algorithm converges.

5. The method according to claim 4, characterized in that After determining whether to accept the neighborhood processing strategy through the acceptance condition of the simulated annealing algorithm, the method further includes: Determining a neighborhood acceptance rate of the simulated annealing algorithm, wherein the neighborhood acceptance rate is a ratio of the number of neighborhood processing strategies that have been determined to be accepted to the number of neighborhood processing strategies that have been constructed; A cooling rate of the simulated annealing algorithm is adjusted based on the neighborhood acceptance rate.

6. The method according to claim 5, characterized in that Adjusting a cooling rate of the simulated annealing algorithm based on the neighborhood acceptance rate includes: If the neighborhood acceptance rate is greater than a first preset acceptance rate, reducing the value of the cooling rate; If the neighborhood acceptance rate is less than a second preset acceptance rate, the value of the cooling rate is increased, wherein the second preset acceptance rate is less than or equal to the first preset acceptance rate.

7. A device for processing computing network tasks, characterized in that: include: An acquisition module is used to acquire topological information of a computing power network, wherein the topological information includes local computing power resource information of each of a plurality of computing power device nodes in communication connection and computing power network task information generated by each of them; A construction module, which constructs a computing power model based on the topological information of the computing power network, wherein the computing power model represents the computing power resource cost consumed by processing the computing power network tasks in the computing power network through the task processing strategy, wherein the task processing strategy includes using the local computing power resources of the computing power device node to perform processing on at least part of the computing power network tasks generated by the computing power device node itself and / or offloading at least part of the computing power network tasks generated by the computing power device node itself to an adjacent node for processing; An optimization module, which takes minimizing the computing resource cost of the computing model as an optimization goal and iteratively optimizes the task processing strategy through a simulated annealing algorithm; The processing module processes the computing power network tasks in the computing power network based on the iteratively optimized task processing strategy.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the steps of the method according to any one of claims 1 to 6 when executed by the processor.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product comprises a non-transitory computer-readable storage medium storing a computer program, wherein the computer program is operable to cause a computer to execute part or all of the steps of the method according to any one of claims 1 to 6.

Images