Calculation task allocation method and device, electronic equipment and storage medium
By obtaining the energy consumption and delay parameters of the computing task, and classifying and allocating edge servers to network nodes, solving the problem of low offloading efficiency of computing task, realizing the reduction of mobile device burden and improving system performance.
Patent Information
- Application Number
- CN202410918048.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-09
- Publication Date
- 2024-11-12
AI Technical Summary
The prior art lacks a solution that can efficiently and reasonably offload computing tasks to edge servers, resulting in the failure to effectively reduce the computing burden of mobile devices and limited system performance improvement.
By obtaining the task type, processing energy consumption, transmission delay and transmission energy consumption during the calculation task transmission process, determining the offload computing power energy consumption, and classifying network nodes based on these parameters to form a computing grid, thereby efficiently allocating edge servers .
It realizes rapid and accurate calculation task unloading, reduces the computing burden of mobile devices, improves overall system performance, and ensures efficient utilization of edge server resources.
Smart Images

Figure CN118939411A_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the field of Internet technology, and in particular, relates to a method, device, electronic device and storage medium for allocating computing tasks. Background Art
[0002] With the development of information technology, mobile edge computing has been widely used. Mobile edge computing is a way to move computing and data storage to the edge of the network. In this computing model, task offloading management has become a core issue.
[0003] Task offloading refers to transferring computing tasks from mobile devices to more powerful computing resources for processing. This technology can effectively reduce the computing burden of mobile devices and improve the performance of the overall system.
[0004] Currently, there is a lack of a solution that can efficiently and reasonably offload computing tasks to edge servers. Summary of the invention
[0005] The embodiments of the present application provide a method, apparatus, device and storage medium for allocating computing tasks, which can solve the problem of the current lack of a solution that can efficiently and reasonably offload computing tasks to edge servers.
[0006] In a first aspect, an embodiment of the present application provides a method for allocating computing tasks, the method comprising:
[0007] During the transmission of computing tasks, the task type, task processing energy consumption, task transmission delay and task transmission energy consumption of each computing task are obtained; the computing task is initiated by a network node and assigned to an edge server for processing;
[0008] Determine the offloading computing power energy consumption according to the task processing energy consumption, the task transmission delay and the task transmission energy consumption, wherein the offloading computing power energy consumption is the total energy consumption from sending the computing task to the edge server to returning the computing task to the network node;
[0009] Classifying the plurality of network nodes according to the task type, the offload computing power energy consumption and the task transmission delay to obtain a plurality of computing grids, each of the computing grids including at least one network node;
[0010] Based on the computing grid, edge servers are allocated to the network nodes.
[0011] In a second aspect, an embodiment of the present application provides a computing task allocation device, the computing task allocation device comprising:
[0012] An acquisition module is used to acquire the task type, task processing energy consumption, task transmission delay and task transmission energy consumption of each computing task during the transmission process of the computing task; the computing task is initiated by the network node and assigned to the edge server for processing;
[0013] A determination module, configured to determine the offload computing power energy consumption according to the task processing energy consumption, the task transmission delay and the task transmission energy consumption, wherein the offload computing power energy consumption is the total energy consumption from the computing task being sent to the edge server to the computing task being returned to the network node;
[0014] A classification module, configured to classify the plurality of network nodes according to the task type, the offload computing power energy consumption and the task transmission delay, to obtain a plurality of computing grids, each of which includes at least one network node;
[0015] The allocation module is used to allocate edge servers to the network nodes based on the computing grid.
[0016] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, it implements the method in the first aspect or any possible implementation of the first aspect.
[0017] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having computer program instructions stored thereon. When the computer program instructions are executed by a processor, the method in the first aspect or any possible implementation of the first aspect is implemented.
[0018] In an embodiment of the present application, during the transmission process of the computing task, the task type, task processing energy consumption, task transmission delay and task transmission energy consumption of each computing task are obtained; the computing task is initiated by a network node and assigned to an edge server for processing; based on the task processing energy consumption, task transmission delay and task transmission energy consumption, the unloaded computing power energy consumption from the computing task being sent to the edge server to the computing task being returned to the network node can be quickly and accurately determined, and based on the task type, unloaded computing power energy consumption and task transmission delay, multiple network nodes are classified to obtain multiple computing grids, each computing grid including at least one network node, here, network nodes with high similarity can be divided into the same computing grid, so as to facilitate the unified allocation of edge servers to network nodes in the computing grid, and finally, based on the computing grid, edge servers can be efficiently and accurately allocated to network nodes. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] In order to more clearly illustrate the technical solution of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0020] Figure 1 This is a schematic diagram of edge computing provided by an embodiment of the present application;
[0021] Figure 2 is a flow chart of a method for allocating computing tasks provided in an embodiment of the present application;
[0022] Figure 3 It is a structural diagram of a computing task allocation device provided in an embodiment of the present application;
[0023] Figure 4 It is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0024] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only configured to explain the present application and are not configured to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is only to provide a better understanding of the present application by illustrating the examples of the present application.
[0025] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article 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, article or device. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.
[0026] The following is a brief introduction to the technical terms involved in this application.
[0027] Edge computing refers to the use of an open platform that integrates network, computing, storage, and application core capabilities to provide the nearest end service at the side close to the object or data source. Its application is initiated at the edge side, resulting in faster network service response, meeting the industry's basic needs in real-time business, application intelligence, security and privacy protection, etc.
[0028] Edge servers are a distributed computing architecture in which data processing is performed at the edge of the network, close to the data source. Designed to improve response speed and reduce bandwidth usage, they are usually deployed where users access the network, such as at the edge of a carrier's network.
[0029] The computing task allocation method provided in the embodiment of the present application can be applied to at least the following application scenarios, which are described below.
[0030] From the edge computing standard documents released by the European Telecommunications Standards Institute (ETSI), the Edge Computing Consortium (ECC) and other organizations, it can be seen that the core idea of edge computing is to move the corresponding functions of the computing platform from the network core side (cloud) to the network access side to provide users with close-range services, so as to reduce the latency of end users and enhance the user experience.
[0031] Mobile edge computing is a new computing model that moves computing and data storage to the edge of the network, that is, devices or terminals, to improve response speed and reduce network bandwidth requirements. In this computing model, task offloading and cache management become core issues.
[0032] This definition provides the basic framework of edge computing, which is to deploy nodes with certain computing capabilities at the edge of the network to provide services to users nearby. Users complete various computing tasks by offloading computing tasks to edge servers. Currently, there is a lack of a reasonable plan for offloading edge computing users to improve user service experience.
[0033] Task offloading refers to transferring computing tasks from mobile devices to more powerful computing resources for processing. This technology can effectively reduce the computing burden of mobile devices and improve the performance of the overall system.
[0034] like Figure 1 As shown, a typical edge computing task offloading scenario is shown.
[0035] To offload computationally intensive and delay-sensitive tasks to edge servers for processing, edge servers are usually deployed near access base stations of mobile cellular networks to facilitate data communication.
[0036] Figure 2 It is a flowchart of a method for allocating computing tasks provided in an embodiment of the present application.
[0037] like Figure 2 As shown, the method for allocating computing tasks may include steps 210 to 240, and the method is applied to a computing task allocation device, as shown below:
[0038] Step 210, during the transmission process of the computing task, obtaining the task type, task processing energy consumption, task transmission delay and task transmission energy consumption of each computing task; the computing task is initiated by the network node and assigned to the edge server for processing;
[0039] Step 220, determining the offloading computing power energy consumption according to the task processing energy consumption, the task transmission delay and the task transmission energy consumption, wherein the offloading computing power energy consumption is the total energy consumption from sending the computing task to the edge server to returning the computing task to the network node;
[0040] Step 230, classifying the plurality of network nodes according to the task type, the offload computing power energy consumption and the task transmission delay to obtain a plurality of computing grids, each of which includes at least one network node;
[0041] Step 240: Allocate edge servers to the network nodes based on the computing grid.
[0042] In an embodiment of the present application, during the transmission process of the computing task, the task type, task processing energy consumption, task transmission delay and task transmission energy consumption of each computing task are obtained; the computing task is initiated by a network node and assigned to an edge server for processing; based on the task processing energy consumption, task transmission delay and task transmission energy consumption, the unloaded computing power energy consumption from the computing task being sent to the edge server to the computing task being returned to the network node can be quickly and accurately determined, and based on the task type, unloaded computing power energy consumption and task transmission delay, multiple network nodes are classified to obtain multiple computing grids, each computing grid including at least one network node, here, network nodes with high similarity can be divided into the same computing grid, so as to facilitate the unified allocation of edge servers to network nodes in the computing grid, and finally, based on the computing grid, edge servers can be efficiently and accurately allocated to network nodes.
[0043] The contents of step 210 to step 240 are described below respectively:
[0044] This involves step 210 .
[0045] During the transmission of computing tasks, the task type, task processing energy consumption, task transmission delay and task transmission energy consumption of each computing task are obtained; the computing task is initiated by a network node and assigned to an edge server for processing;
[0046] Task offloading is generally divided into three main steps.
[0047] During the task offloading and task uploading stages, users use reliable wireless transmission technology to offload tasks to edge servers through mobile access points. At this time, if the corresponding services are deployed in the server, only the input of the task needs to be uploaded, such as pictures collected by the terminal device in AR applications. If the server only provides a computing environment but does not deploy the corresponding services, then the edge server needs to request or upload the code of the corresponding service to the cloud for deployment on the server.
[0048] Among them, task types include: single task, multi-task and multi-weight task.
[0049] Single-Task Computation: In single-task computing, the model is trained and optimized only for a specific computing task. For example, one model is trained only to perform addition operations, and another model is trained only to perform multiplication operations. This method is relatively simple, but the model's generalization ability may be weak and it can only handle a single type of computing task.
[0050] Multi-Task Computation: In multi-task computing, the model is trained and optimized for multiple related computing tasks at the same time. For example, a model can perform multiple basic operations such as addition, subtraction, multiplication, and division at the same time. This method can take advantage of the correlation between different computing tasks and improve the generalization ability and computing efficiency of the model. However, it requires a reasonable design of the task relationship and loss function weight.
[0051] Multi-Weighted Task Computation is similar to multi-task computing, but each computing task has a different loss function weight. The weight can be adjusted based on factors such as the complexity and importance of the task. For example, for addition and multiplication operations, a higher weight can be given to multiplication operations because multiplication operations are usually more complex.
[0052] The three types of computing tasks, single-task, multi-task and multi-weight task, reflect the different requirements for model training and optimization in computing applications.
[0053] The task transmission process includes:
[0054] Receiving computing tasks sent by each of the network nodes;
[0055] The computing task is distributed to the edge server, which is a server deployed at the edge of the network for processing the computing task.
[0056] When the task is completed on the edge server, the final task offloading computing power consumption is recorded, and the task result is returned to the end user through the user access node.
[0057] Task offloading is generally divided into three main steps.
[0058] Task unloading-task uploading steps:
[0059] In this stage, users use reliable wireless transmission technology to offload tasks to edge servers through mobile access points. If the corresponding services are deployed on the server, you only need to upload the input of the task, such as pictures collected by the terminal device in AR applications. If the server only provides a computing environment but does not deploy the corresponding services, the edge server needs to request or upload the code of the corresponding service to the cloud for deployment on the server.
[0060] For system scenarios, we need to select appropriate communication technologies based on the characteristics of user services. For example, when the number of users is much greater than the number of access nodes, orthogonal frequency division multiplexing (OFM) is used. For another example, when low-power IoT devices are far away from the base station, linear frequency modulation technologies such as LoRa can be used to increase the transmission distance, and the spreading factor can be adjusted according to the distance.
[0061] With the introduction of the concept of ultra-dense networks (UN), there may be multiple optional access nodes near the user. At this time, it is necessary to coordinate the upload strategies of different users (node selection, user communication timing arrangement) to reduce the task upload delay.
[0062] When a task arrives at the server, the server allocates reasonable resources and processing timing according to the computational complexity of the task and the required user service requirements. In an edge network composed of multiple servers, the collaborative processing of tasks between multiple servers also needs to be considered.
[0063] Specific process of result return: The amount of data returned by a task is often much smaller than the amount of data uploaded by the task. For example, in a video surveillance scenario, the camera uploads the collected images to the edge network for analysis and processing, and the result returned is often just an instruction on whether an alarm is needed. In a mobile scenario, the edge server accessed by the user and the server that receives the returned results are often not the same node, and the results need to be routed to the corresponding return node based on the user's movement trajectory.
[0064] This involves step 120 .
[0065] Determine the offloading computing power energy consumption according to the task processing energy consumption, the task transmission delay and the task transmission energy consumption, wherein the offloading computing power energy consumption is the total energy consumption from sending the computing task to the edge server to returning the computing task to the network node;
[0066] Task processing energy consumption can determine the energy consumption required to execute tasks on the local device. This includes the power consumption of hardware resources such as CPU and GPU.
[0067] Task transmission delay analysis can estimate the time delay required to transmit tasks from local devices to the cloud, which is affected by factors such as network bandwidth and distance.
[0068] Task transmission energy consumption analysis can be performed by calculating the energy consumption required to transmit tasks from local devices to the cloud. This includes the power consumption of wireless communication modules.
[0069] The energy consumption of offloaded computing power is calculated by weighing the task processing energy consumption, task transmission delay and task transmission energy consumption to obtain the overall energy consumption of offloaded computing power.
[0070] For example: offloading computing power energy consumption = task transmission energy consumption + cloud task execution energy consumption - local task execution energy consumption;
[0071] Task transmission energy consumption = task transmission delay × task transmission energy consumption rate;
[0072] Cloud task execution energy consumption = cloud task execution delay × cloud task execution power;
[0073] Local task execution energy consumption = local task execution delay × local task execution power;
[0074] By calculating various parameters, the overall energy consumption of offloaded computing power is finally obtained, which helps to make the best computing power offloading decision between local devices and the cloud.
[0075] This involves step 130 .
[0076] Classifying the plurality of network nodes according to the task type, the offload computing power energy consumption and the task transmission delay to obtain a plurality of computing grids, each of the computing grids including at least one network node;
[0077] By calculating the computing power consumption and latency of the task from the user service delay to the terminal device energy consumption to the edge server and the computing power consumption of the task, we can obtain computing power consumption and path transmission quality that is close to the actual situation.
[0078] All network nodes are organized into a matrix and a matrix decomposition model is used to merge nodes with high similarity into the same grid for management. The same edge server is matched to the grid to maximize the computing and resource utilization efficiency of edge task assistance.
[0079] In a possible embodiment, step 130 includes:
[0080] According to the task type, the offload computing power energy consumption and the task transmission delay, all network nodes are formed into a network node matrix;
[0081] The network node matrix is decomposed by a matrix decomposition model to obtain multiple computing grids.
[0082] First, a formal method is used to describe the relationship between network nodes, offload task types, and edge servers. The relationship between them can be defined as:
[0083]
[0084] Assume TA={TA1,TA2,...TA n} is the offloading task type (single task, multiple tasks, business weight category task). The big data business includes seven basic logical relationships: sequential relationship (SegR), conditional relationship (CndR), and relationship (AndR), or relationship (OrR), concurrent relationship (ConcR), synchronous relationship (SynR) and cyclic relationship (CycR).
[0085] Since relationships are communicable, various new relationships can be formed between various types of business big data through relationship reasoning, which can be used as a method of distributing data and merging them.
[0086] Then, the grid is divided into three types of edge-assisted task types: single task, multi-task, and multi-weight task according to the computing task type. Finally, the network nodes classified by matrix decomposition are aggregated into the same grid and assigned a separate edge server.
[0087] In the step of building a network node matrix, all available network nodes (including local devices and cloud resources) can be formed into a matrix. The rows of the matrix represent different network nodes, and the columns of the matrix represent attributes such as task type, offload computing power consumption, and task transmission delay.
[0088] In the step of performing matrix decomposition on the network node matrix, a suitable matrix decomposition model may be selected, such as non-negative matrix decomposition (NMF), singular value decomposition (SVD), etc.;
[0089] Through matrix decomposition, the original network node matrix is decomposed into two or more sub-matrices. The sub-matrices represent different computational grids, each with different characteristics.
[0090] In the steps involving the construction of the computing grid, the network nodes can be divided into different computing grids according to the sub-matrices obtained by decomposition. The nodes in each computing grid have similar task types, offload computing power consumption, and task transmission delays, so that fine-grained management and optimization of network resources can be achieved.
[0091] Therefore, according to the task type, the offloading computing power consumption and the task transmission delay, all network nodes are formed into a network node matrix, and the network node matrix is decomposed through a matrix decomposition model to obtain multiple computing grids. The network nodes can be effectively grouped according to the task characteristics, and personalized resource scheduling strategies can be adopted for different computing grids to maximize the utilization efficiency of network resources and improve computing performance.
[0092] In a possible embodiment, step 130 includes:
[0093] According to the offload computing power energy consumption, the task transmission delay and the task type, all computing tasks sent by each of the network nodes to the corresponding edge servers are sorted to obtain a task sequence;
[0094] According to the task sequence, the plurality of network nodes are classified to obtain a plurality of task computing grids.
[0095] The task sequence includes: a task sequence based on offloading computing power energy consumption and a task sequence based on the transmission delay of the task under different task types.
[0096] Statistics are collected for all computing tasks sent by the current network node to the edge server, and sorted by energy consumption, latency, and computing task type. Statistics are collected for which type of tasks are transmitted fastest and with the least consumption from each network node to the edge node.
[0097] The steps involving task sorting can sort all computing tasks sent to the edge server for each network node according to the offload computing power consumption, task transmission delay and task type, and obtain three different task sequences:
[0098] Task sequence sorted by offload computing power and energy consumption;
[0099] Task sequence sorted by task transmission delay;
[0100] A sequence of tasks sorted by task type.
[0101] The steps involving network node classification are to classify network nodes according to the above three task sequences, and divide network nodes with similar offload computing power consumption, task transmission delay and task type into the same task computing grid, so as to ensure that nodes in the same grid have similar task characteristics.
[0102] The steps involved in building a computing grid are as follows: According to the above classification results, multiple task computing grids are built. Nodes in each computing grid can share resources to achieve more efficient task scheduling and load balancing. Resource allocation between different computing grids can be dynamically adjusted according to demand.
[0103] Therefore, network nodes can be effectively classified and organized according to task characteristics, and personalized resource management strategies can be adopted for different computing grids to improve task scheduling efficiency and reduce energy consumption and latency.
[0104] The step of classifying the plurality of network nodes according to the task sequence to obtain a plurality of task computing grids mentioned above includes:
[0105] Determine a target computing task according to each of the task sequences, wherein the target computing task is a computing task with the fastest transmission speed or the lowest energy consumption initiated by each of the network nodes to the edge server;
[0106] The plurality of network nodes are classified according to the target computing tasks to obtain a plurality of task computing grids.
[0107] The task sequence based on the offloading computing power energy consumption is a task sequence from small to large offloading computing power energy consumption;
[0108] The task sequence based on the task transmission delay is a task sequence with task transmission delay from short to long.
[0109] According to each of the task sequences, it is possible to determine the target computing task with the fastest transmission speed or the lowest energy consumption initiated by each of the network nodes to the edge server.
[0110] For each network node, the three task sequences obtained previously are sorted by offloading computing power consumption, task transmission delay, and task type. The first task in each sequence is selected as the target computing task of the node, which ensures that the task with the fastest transmission speed or the lowest energy consumption is selected.
[0111] The steps involved in classifying network nodes are to classify network nodes according to their target computing tasks, and to divide network nodes with the same target computing tasks into the same task computing grid, so as to ensure that nodes in the same grid have similar task characteristics.
[0112] The steps involved in building a computing grid are to build multiple task computing grids according to the above classification results. The nodes in each computing grid can share resources to achieve more efficient task scheduling and load balancing. Resource allocation between different computing grids can be dynamically adjusted according to demand.
[0113] Therefore, network nodes can be effectively classified and organized according to task transmission speed and energy consumption characteristics, and personalized resource management strategies can be adopted for different computing grids to improve task scheduling efficiency and reduce energy consumption and latency.
[0114] In a possible embodiment, a service weight value of each computing task is obtained;
[0115] Allocating transmission resources and edge servers to each of the computing tasks according to the service weight value;
[0116] The transmission resource is used to transmit the computing task to the edge server.
[0117] First, users complete various computing tasks by offloading computing tasks to edge servers. Priority is marked according to business weights during task upload, so that when edge computing resources reach bottlenecks, resources are scheduled in order through different business weights assigned by priority marking, so as to achieve reasonable resource scheduling when resources are bottlenecked.
[0118] This involves step 140 .
[0119] Based on the computing grid, edge servers are allocated to the network nodes.
[0120] From the perspective of user service latency, service deployment mainly affects user service latency from the following aspects: different service deployment locations and allocated bandwidth resources. The computing resources allocated to the service will affect the time required for service calculation; the access nodes allocated to the service will affect the user's wireless transmission speed.
[0121] The service deployment optimization problem can be considered as an optimization problem with user service delay as the goal. The user group in the network is represented by U, the edge server group is represented by H, the service required by the user is represented by N, and the transmission delay in the edge network is represented by the matrix l. It is necessary to deploy the required service for each user to minimize the end-to-end delay of the user. The calculation formula is as follows:
[0122]
[0123] Where P k represents a specific path in the edge network, n represents the specific service required by the user, and h j represents a specific edge server, X ijk Is a 0 / 1 bit flag that indicates whether a particular service n has been deployed on server h j and the user accesses it through path p. ijkIt represents the delay caused by the user accessing service n deployed on server h through path P. From the above formula, we can see that the only variable is X, which indicates which server the service computing resources required by the user are deployed on, and gives the path selection corresponding to the user accessing the service in the edge network.
[0124] To solve the above problem, we need to traverse all possible service deployments and all possible paths for users to access the service in order to minimize user latency. Solving this problem does not simply mean that deploying the service as close to the user as possible is the optimal solution to this problem, because resource constraints also need to be considered during the service deployment stage. Specifically:
[0125] The limited server resources, including CPU computing resources, I / O transmission resources, and memory storage resources. The transmission bandwidth resources within the edge network. The end-to-end delay of the service must be less than the maximum delay that the service can tolerate.
[0126] Energy consumption of terminal equipment. Another indicator for evaluating the quality of service deployment from the user's perspective is the energy consumption of terminal equipment. The energy consumption of terminal equipment mainly comes from the following two situations:
[0127] When tasks are executed locally, the energy overhead mainly comes from the computational energy consumption caused by the user processing the task.
[0128] When the service is executed at the edge, the energy consumption mainly comes from the transmission energy consumption caused by the user uploading the corresponding service to the edge network (although the device needs to monitor the channel all the time when the user transmits the result, this overhead is negligible compared with the overhead of uploading the task). From the energy consumption comparison of the above two, it can be seen that from the user's perspective, when the task is offloaded to the edge network for execution, the energy consumption can be effectively reduced. Therefore, from the perspective of the overall energy consumption of the edge user group, this problem can be summarized as follows:
[0129]
[0130] Specifically, x represents the deployment strategy of the service, y represents the uninstallation strategy of the user, and x ns =1 means that service s is deployed in server n. y represents the user's uninstallation policy. us =1 indicates that service s offloads the task to the edge network through base station u. When all base stations u in the edge network satisfy y = 0, it means that the task is executed locally.
[0131] The optimization goal of formula 2 is to maximize the number of users served by the edge network, which means that the edge network can serve edge users as much as possible to reduce user energy consumption. This problem cannot be solved by uploading all services to reduce the energy consumption of the edge system, because the following points need to be considered:
[0132] User service delay. An important indicator of edge computing is the end-to-end delay of user services. Therefore, when more and more users upload tasks, the channel competition between users will become fierce, resulting in an increase in the transmission delay of the wireless channel, which cannot meet the user's requirements for delay. At this time, users will not choose to offload tasks to the edge network. us = 0. Indicators that exceed the user service delay threshold are not included.
[0133] Transmission energy consumption of terminal devices. When users choose to perform computing on the edge network, it will cause congestion of the wireless link and the transmission power of the device will increase significantly.
[0134] Computing power quantization model:
[0135] Among them, C br is the total computing power requirement; f(x) is the mapping function; a, B and y are the mapping ratio coefficients; q is the redundant computing power. Taking parallel computing power as an example, assuming there are three different types of parallel computing chip resources: b1, b2, and b3, then f(b j ) represents the mapping function of the parallel computing capability that the j-th parallel computing chip b can provide, and q2 represents the redundant computing power of parallel computing.
[0136] In an embodiment of the present application, during the transmission process of the computing task, the task type, task processing energy consumption, task transmission delay and task transmission energy consumption of each computing task are obtained; the computing task is initiated by a network node and assigned to an edge server for processing; based on the task processing energy consumption, task transmission delay and task transmission energy consumption, the unloaded computing power energy consumption from the computing task being sent to the edge server to the computing task being returned to the network node can be quickly and accurately determined, and based on the task type, unloaded computing power energy consumption and task transmission delay, multiple network nodes are classified to obtain multiple computing grids, each computing grid including at least one network node, here, network nodes with high similarity can be divided into the same computing grid, so as to facilitate the unified allocation of edge servers to network nodes in the computing grid, and finally, based on the computing grid, edge servers can be efficiently and accurately allocated to network nodes.
[0137] Based on the above Figure 1 The method for allocating computing tasks shown in the figure, the embodiment of the present application also provides a computing task allocation device, such as Figure 3 As shown, the computing task allocation device 300 may include:
[0138] The acquisition module 310 is used to acquire the task type, task processing energy consumption, task transmission delay and task transmission energy consumption of each computing task during the transmission process of the computing task; the computing task is initiated by the network node and assigned to the edge server for processing;
[0139] A determination module 320 is used to determine the offload computing power energy consumption according to the task processing energy consumption, the task transmission delay and the task transmission energy consumption, wherein the offload computing power energy consumption is the total energy consumption from sending the computing task to the edge server to returning the computing task to the network node;
[0140] A classification module 330, configured to classify the plurality of network nodes according to the task type, the offload computing power energy consumption and the task transmission delay, to obtain a plurality of computing grids, each of which includes at least one network node;
[0141] The allocation module 340 is configured to allocate an edge server to the network node based on the computing grid.
[0142] In a possible embodiment, the classification module 330 is specifically configured to:
[0143] According to the task type, the offload computing power energy consumption and the task transmission delay, all network nodes are formed into a network node matrix;
[0144] The network node matrix is decomposed by a matrix decomposition model to obtain multiple computing grids.
[0145] In a possible embodiment, the classification module 330 is specifically configured to:
[0146] According to the offload computing power energy consumption, the task transmission delay and the task type, all computing tasks sent by each of the network nodes to the corresponding edge servers are sorted to obtain a task sequence;
[0147] According to the task sequence, the plurality of network nodes are classified to obtain a plurality of task computing grids.
[0148] In a possible embodiment, the classification module 330 is specifically configured to:
[0149] Determine a target computing task according to each of the task sequences, wherein the target computing task is a computing task with the fastest transmission speed or the lowest energy consumption initiated by each of the network nodes to the edge server;
[0150] The plurality of network nodes are classified according to the target computing tasks to obtain a plurality of task computing grids.
[0151] In a possible embodiment, the task types include: single task, multi-task and multi-weight task.
[0152] In a possible embodiment, the task transmission process includes:
[0153] Receiving computing tasks sent by each of the network nodes;
[0154] The computing task is distributed to the edge server, which is a server deployed at the edge of the network for processing the computing task.
[0155] In a possible embodiment, the acquisition module 310 is further used to obtain the service weight value of each computing task;
[0156] The allocation module 340 is further configured to allocate transmission resources and edge servers to each of the computing tasks according to the service weight value;
[0157] The transmission resource is used to transmit the computing task to the edge server.
[0158] In an embodiment of the present application, during the transmission process of the computing task, the task type, task processing energy consumption, task transmission delay and task transmission energy consumption of each computing task are obtained; the computing task is initiated by a network node and assigned to an edge server for processing; based on the task processing energy consumption, task transmission delay and task transmission energy consumption, the unloaded computing power energy consumption from the computing task being sent to the edge server to the computing task being returned to the network node can be quickly and accurately determined, and based on the task type, unloaded computing power energy consumption and task transmission delay, multiple network nodes are classified to obtain multiple computing grids, each computing grid including at least one network node, here, network nodes with high similarity can be divided into the same computing grid, so as to facilitate the unified allocation of edge servers to network nodes in the computing grid, and finally, based on the computing grid, edge servers can be efficiently and accurately allocated to network nodes.
[0159] Figure 4 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application is shown.
[0160] The electronic device may include a processor 401 and a memory 402 storing computer program instructions.
[0161] Specifically, the processor 401 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.
[0162] The memory 402 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 402 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 402 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 402 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 402 is a non-volatile solid-state memory. In a specific embodiment, the memory 402 includes a read-only memory (ROM). In appropriate cases, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM) or a flash memory or a combination of two or more of these.
[0163] The processor 401 reads and executes computer program instructions stored in the memory 402 to implement any one of the computing task allocation methods in the embodiments shown in the figure.
[0164] In one example, the electronic device may further include a communication interface 404 and a bus 410. Figure 4 As shown, the processor 401, the memory 402, and the communication interface 404 are connected via a bus 410 and communicate with each other.
[0165] The communication interface 404 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.
[0166] Bus 410 includes hardware, software or both, and the parts of electronic equipment are coupled to each other. For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front side bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 410 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the application considers any suitable bus or interconnection.
[0167] The electronic device can execute the method for allocating computing tasks in the embodiment of the present application, thereby realizing the combination of Figure 2 Describes the method for allocating computing tasks.
[0168] In addition, in combination with the method for allocating computing tasks in the above embodiment, the present application embodiment can provide a computer-readable storage medium for implementation. The computer-readable storage medium stores computer program instructions; when the computer program instructions are executed by the processor, the Figure 1 The method of allocating computing tasks.
[0169] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.
[0170] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present application are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.
[0171] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.
[0172] The above is only a specific implementation of the present application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present application is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in this application, and these modifications or replacements should be included in the protection scope of this application.
Claims
1. A method for allocating computing tasks, It is characterized in that The method comprises: During the transmission of computing tasks, the task type, task processing energy consumption, task transmission delay and task transmission energy consumption of each computing task are obtained; the computing task is initiated by a network node and assigned to an edge server for processing; Determine the offloading computing power energy consumption according to the task processing energy consumption, the task transmission delay and the task transmission energy consumption, wherein the offloading computing power energy consumption is the total energy consumption from sending the computing task to the edge server to returning the computing task to the network node; Classifying the plurality of network nodes according to the task type, the offload computing power energy consumption and the task transmission delay to obtain a plurality of computing grids, each of the computing grids including at least one network node; Based on the computing grid, edge servers are allocated to the network nodes.
2. The method according to claim 1, It is characterized in that The multiple network nodes are classified according to the task type, the offload computing power energy consumption and the task transmission delay to obtain multiple computing grids, including: According to the task type, the offload computing power energy consumption and the task transmission delay, all network nodes are formed into a network node matrix; The network node matrix is decomposed by a matrix decomposition model to obtain multiple computing grids.
3. The method according to claim 1, It is characterized in that The multiple network nodes are classified according to the task type, the offload computing power energy consumption and the task transmission delay to obtain multiple computing grids, including: According to the offload computing power energy consumption, the task transmission delay and the task type, all computing tasks sent by each of the network nodes to the corresponding edge servers are sorted to obtain a task sequence; According to the task sequence, the plurality of network nodes are classified to obtain a plurality of task computing grids.
4. The method according to claim 3, It is characterized in that The method of classifying the plurality of network nodes according to the task sequence to obtain a plurality of task computing grids includes: Determine a target computing task according to each of the task sequences, wherein the target computing task is a computing task with the fastest transmission speed or the lowest energy consumption initiated by each of the network nodes to the edge server; The plurality of network nodes are classified according to the target computing tasks to obtain a plurality of task computing grids.
5. The method according to claim 1, It is characterized in that The task types include: single task, multi-task and multi-weight task.
6. The method according to claim 1, It is characterized in that The task transmission process includes: Receiving computing tasks sent by each of the network nodes; The computing task is distributed to the edge server, which is a server deployed at the edge of the network for processing the computing task.
7. The method according to claim 1, It is characterized in that The method further comprises: Get the business weight value of each computing task; Allocating transmission resources and edge servers to each of the computing tasks according to the service weight value; The transmission resource is used to transmit the computing task to the edge server.
8. A computing task distribution device, It is characterized in that The computing task allocation device comprises: An acquisition module is used to acquire the task type, task processing energy consumption, task transmission delay and task transmission energy consumption of each computing task during the transmission process of the computing task; the computing task is initiated by the network node and assigned to the edge server for processing; A determination module, configured to determine the offload computing power energy consumption according to the task processing energy consumption, the task transmission delay and the task transmission energy consumption, wherein the offload computing power energy consumption is the total energy consumption from the computing task being sent to the edge server to the computing task being returned to the network node; A classification module, configured to classify the plurality of network nodes according to the task type, the offload computing power energy consumption and the task transmission delay, to obtain a plurality of computing grids, each of which includes at least one network node; The allocation module is used to allocate edge servers to the network nodes based on the computing grid.
9. An electronic device, It is characterized in that The electronic device comprises: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the method for allocating computing tasks as claimed in any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed by a processor, the method for allocating computing tasks as described in any one of claims 1 to 7 is implemented.
