技术领域Technical Field
本发明涉及任务调度领域，尤其涉及一种基于分布式云平台的任务调度方法、装置、计算机设备及存储介质。The present invention relates to the field of task scheduling, and in particular to a task scheduling method, device, computer equipment and storage medium based on a distributed cloud platform.
背景技术Background technique
在网络数据爆炸式增长的时代，云平台因其高性能计算能力越来越受到人们的欢迎。目前使用的分布式云平台是一种将云计算服务分布式下沉到边缘节点的新兴云平台架构，该分布式云平台由虚拟机或容器驱动的方式使得计算资源可以更好地隔离与管理。但由于接入云平台的终端设备呈现爆炸式增长，云平台的调度方式为用户将业务发送至调度节点，调度节点生成调度策略之后将业务数据中转至目标计算节点，导致调度阶段的网络负载重。In an era of explosive growth in network data, cloud platforms are becoming increasingly popular due to their high-performance computing capabilities. The currently used distributed cloud platform is an emerging cloud platform architecture that distributes cloud computing services to edge nodes. The distributed cloud platform is driven by virtual machines or containers, which allows better isolation and management of computing resources. However, due to the explosive growth of terminal devices connected to the cloud platform, the scheduling method of the cloud platform is that users send services to the scheduling node, and the scheduling node generates a scheduling strategy and then transfers the service data to the target computing node, resulting in heavy network load in the scheduling stage.
因此，现有的云平台的调度节点存在网络负载重的问题。Therefore, the scheduling nodes of existing cloud platforms have the problem of heavy network load.
发明内容Summary of the invention
本发明实施例提供一种基于分布式云平台的任务调度方法、装置、计算机设备和存储介质，以提高基于分布式云平台的任务调度的智能化，降低调度节点的网络负载。The embodiments of the present invention provide a distributed cloud platform-based task scheduling method, apparatus, computer equipment and storage medium to improve the intelligence of distributed cloud platform-based task scheduling and reduce the network load of scheduling nodes.
为了解决上述技术问题，本申请实施例提供一种基于分布式云平台的任务调度方法，包括：In order to solve the above technical problems, the embodiment of the present application provides a task scheduling method based on a distributed cloud platform, including:
获取目标从节点发送的任务请求，所述任务请求包括基于所述目标从节点上各任务确定的图结构特征 ；Obtaining a task request sent by a target slave node, wherein the task request includes a graph structure feature determined based on each task on the target slave node;
基于所述目标从节点所在区域的主节点上的特征提取模块 ，针对所述图结构特征上的每个任务节点，对所述任务节点以及所述任务节点的领域信息进行特征聚合处理，得到聚合图结构特征；Based on the feature extraction module on the master node of the area where the target slave node is located, for each task node on the graph structure feature, feature aggregation processing is performed on the task node and the domain information of the task node to obtain an aggregated graph structure feature;
将所述聚合图结构特征输入到深度强化学习模块中，得到所述目标从节点的任务调度策略；Inputting the aggregate graph structure features into a deep reinforcement learning module to obtain a task scheduling strategy for the target slave node;
根据所述任务调度策略，对所述目标从节点上各任务进行调度分配。According to the task scheduling strategy, each task on the target slave node is scheduled and allocated.
为了解决上述技术问题，本申请实施例还提供一种基于分布式云平台的任务调度装置，包括：In order to solve the above technical problems, the embodiment of the present application further provides a task scheduling device based on a distributed cloud platform, including:
任务请求获取模块，用于获取目标从节点发送的任务请求，所述任务请求包括基于所述目标从节点上各任务确定对应的图结构特征 ；A task request acquisition module, used to acquire a task request sent by a target slave node, wherein the task request includes graph structure features corresponding to each task on the target slave node;
聚合模块，用于基于所述目标从节点所在区域的主节点上的特征提取模块 ，针对所述图结构特征上的每个任务节点，对所述任务节点以及所述任务节点的领域信息进行特征聚合处理，得到聚合图结构特征；An aggregation module is used to perform feature aggregation processing on each task node and the domain information of the task node based on the feature extraction module on the master node of the area where the target slave node is located, so as to obtain an aggregated graph structure feature;
任务调度策略确定模块，用于将所述聚合图结构特征输入到深度强化学习模块中，得到所述目标从节点的任务调度策略；A task scheduling strategy determination module, used to input the aggregate graph structure features into a deep reinforcement learning module to obtain the task scheduling strategy of the target slave node;
调度模块，用于根据所述任务调度策略，对所述目标从节点上各任务进行调度分配。The scheduling module is used to schedule and allocate each task on the target slave node according to the task scheduling strategy.
为了解决上述技术问题，本申请实施例还提供一种计算机设备，包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现上述基于分布式云平台的任务调度方法的步骤。In order to solve the above technical problems, an embodiment of the present application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned task scheduling method based on a distributed cloud platform when executing the computer program.
为了解决上述技术问题，本申请实施例还提供一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行时实现上述基于分布式云平台的任务调度方法的步骤。In order to solve the above technical problems, an embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned task scheduling method based on a distributed cloud platform are implemented.
本发明实施例提供的基于分布式云平台的任务调度方法、装置、计算机设备及存储介质，通过获取目标从节点发送的任务请求，所述任务请求包括基于所述目标从节点上各任务确定的图结构特征；基于所述目标从节点所在区域的主节点上的特征提取模块，针对所述图结构特征上的每个任务节点，对所述任务节点以及所述任务节点的领域信息进行特征聚合处理，得到聚合图结构特征；将所述聚合图结构特征输入到深度强化学习模块中，得到所述目标从节点的任务调度策略；根据所述任务调度策略，对所述目标从节点上各任务进行调度分配。采用本发明提高基于分布式云平台的任务调度的智能化，降低调度节点的网络负载。The task scheduling method, device, computer equipment and storage medium based on the distributed cloud platform provided by the embodiment of the present invention obtain the task request sent by the target slave node, the task request includes the graph structure features determined based on each task on the target slave node; based on the feature extraction module on the master node of the area where the target slave node is located, for each task node on the graph structure feature, the task node and the domain information of the task node are subjected to feature aggregation processing to obtain the aggregated graph structure feature; the aggregated graph structure feature is input into the deep reinforcement learning module to obtain the task scheduling strategy of the target slave node; according to the task scheduling strategy, each task on the target slave node is scheduled and allocated. The present invention is adopted to improve the intelligence of task scheduling based on the distributed cloud platform and reduce the network load of the scheduling node.
附图说明BRIEF DESCRIPTION OF THE DRAWINGS
为了更清楚地说明本发明实施例的技术方案，下面将对本发明实施例的描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.
图1是本申请可以应用于其中的示例性系统架构图；FIG1 is a diagram of an exemplary system architecture in which the present application may be applied;
图2是本申请的基于分布式云平台的任务调度方法的一个实施例的流程图；FIG2 is a flow chart of an embodiment of a task scheduling method based on a distributed cloud platform of the present application;
图3是本申请的基于分布式云平台的任务调度方法的一具体实施例的调度应用框架示意图；FIG3 is a schematic diagram of a scheduling application framework of a specific embodiment of a task scheduling method based on a distributed cloud platform of the present application;
图4是本申请的基于分布式云平台的任务调度方法的又一具体实施例的调度应用框架示意图；FIG4 is a schematic diagram of a scheduling application framework of another specific embodiment of the task scheduling method based on a distributed cloud platform of the present application;
图5是本申请的基于分布式云平台的任务调度方法的另一具体实施例的调度应用框架示意图；FIG5 is a schematic diagram of a scheduling application framework of another specific embodiment of the task scheduling method based on a distributed cloud platform of the present application;
图6是根据本申请的基于分布式云平台的任务调度装置的一个实施例的结构示意图；FIG6 is a schematic diagram of the structure of an embodiment of a task scheduling device based on a distributed cloud platform according to the present application;
图7是根据本申请的计算机设备的一个实施例的结构示意图。FIG. 7 is a schematic structural diagram of a computer device according to an embodiment of the present application.
具体实施方式Detailed ways
除非另有定义，本文所使用的所有的技术和科学术语与属于本申请的技术领域的技术人员通常理解的含义相同；本文中在申请的说明书中所使用的术语只是为了描述具体的实施例的目的，不是旨在于限制本申请；本申请的说明书和权利要求书及上述附图说明中的术语“包括”和“具有”以及它们的任何变形，意图在于覆盖不排他的包含。本申请的说明书和权利要求书或上述附图中的术语“第一”、“第二”等是用于区别不同对象，而不是用于描述特定顺序。Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.
在本文中提及“实施例”意味着，结合实施例描述的特定特征、结构或特性可以包含在本申请的至少一个实施例中。在说明书中的各个位置出现该短语并不一定均是指相同的实施例，也不是与其它实施例互斥的独立的或备选的实施例。本领域技术人员显式地和隐式地理解的是，本文所描述的实施例可以与其它实施例相结合。Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
请参阅图1，如图1所示，分布式云平台100可以包括多个区域，每个区域可以包括从节点101、102、103，网络104和主节点105。网络104用以在从节点101、102、103和主节点105之间提供通信链路的介质。网络104可以包括各种连接类型，例如有线、无线通信链路或者光纤电缆等等。Please refer to FIG. 1 , as shown in FIG. 1 , a distributed cloud platform 100 may include multiple regions, each of which may include slave nodes 101, 102, 103, a network 104, and a master node 105. The network 104 is used to provide a medium for a communication link between the slave nodes 101, 102, 103 and the master node 105. The network 104 may include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.
用户可以使用从节点101、102、103通过网络104与主节点105交互，以接收或发送消息等。Users can use slave nodes 101, 102, 103 to interact with master node 105 through network 104 to receive or send messages, etc.
从节点101、102、103可以是具有显示屏并且支持网页浏览的各种电子设备，包括但不限于智能手机、平板电脑、电子书阅读器、MP3播放器( Moving Picture Eperts GroupAudio Layer III，动态影像专家压缩标准音频层面3 )、MP4(Moving Picture EpertsGroup Audio Layer IV，动态影像专家压缩标准音频层面4 )播放器、膝上型便携计算机和台式计算机等等。Slave nodes 101, 102, 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Eperts Group Audio Layer III), MP4 (Moving Picture Eperts Group Audio Layer IV) players, laptop computers, desktop computers, etc.
主节点105可以是提供各种服务的主节点，例如对从节点101、102、103上显示的页面提供支持的后台服务器。The master node 105 may be a master node that provides various services, such as a background server that provides support for pages displayed on the slave nodes 101 , 102 , and 103 .
需要说明的是，由于分布式云平台在地理位置上广泛分布，提供给客户端延迟更低、容错性和可用性更好的服务。在本申请实施例中，将分布式云平台划分为不同的服务区域，每个区域内包括多个分布式云节点，在一个区域内，分布式云节点可以包括一个主节点和多个从节点，主节点负责管理和协调同一区域内的所有从节点。同一区域内的从节点的负载情况周期性报告给该区域的主节点。It should be noted that, since the distributed cloud platform is widely distributed geographically, it provides clients with services with lower latency, better fault tolerance and availability. In an embodiment of the present application, the distributed cloud platform is divided into different service areas, each area includes multiple distributed cloud nodes, and within a region, the distributed cloud nodes may include a master node and multiple slave nodes, and the master node is responsible for managing and coordinating all slave nodes in the same region. The load conditions of the slave nodes in the same region are periodically reported to the master node of the region.
示例性的，以一个基站的信号覆盖范围为一个服务区域。主节点可以直接选择基站里面或者附近的云节点。随着边缘计算和物联网技术的发展，一些基站开始集成一定程度的计算能力。一个服务区域下有很多从节点，比如路边监控等等都可以算作一个从节点，可以提供服务。For example, the signal coverage of a base station is a service area. The master node can directly select a cloud node in or near the base station. With the development of edge computing and IoT technology, some base stations have begun to integrate a certain degree of computing power. There are many slave nodes under a service area. For example, roadside monitoring can be counted as a slave node and can provide services.
需要说明的是，本申请实施例所提供的基于分布式云平台的任务调度方法由目标区域的主节点执行，相应地，基于分布式云平台的任务调度装置设置于目标区域的主节点中。It should be noted that the task scheduling method based on the distributed cloud platform provided in the embodiment of the present application is executed by the master node of the target area, and accordingly, the task scheduling device based on the distributed cloud platform is set in the master node of the target area.
应该理解，图1中的从节点、网络和主节点的数目仅仅是示意性的。根据实现需要，可以具有任意数目的从节点、网络和主节点，本申请实施例中的从节点101、102、103具体可以对应的是实际生产中的应用系统。It should be understood that the number of slave nodes, networks and master nodes in Figure 1 is only schematic. According to the implementation requirements, there can be any number of slave nodes, networks and master nodes. The slave nodes 101, 102, 103 in the embodiment of the present application can specifically correspond to the application system in actual production.
请参阅图2，图2示出本发明实施例提供的一种基于分布式云平台的任务调度方法，以该方法应用在图1中的目标区域的主节点为例进行说明，分布式云平台中目标区域的主节点，所述分布式云平台包括多个区域，每个所述区域包括主节点以及多个从节点，详述如下：Please refer to FIG. 2 , which shows a task scheduling method based on a distributed cloud platform provided by an embodiment of the present invention. The method is applied to the master node of the target area in FIG. 1 as an example for explanation. The master node of the target area in the distributed cloud platform includes multiple areas, each of which includes a master node and multiple slave nodes, as described in detail as follows:
S201、获取目标从节点发送的任务请求，任务请求包括基于目标从节点上各任务确定的图结构特征。S201. Obtain a task request sent by a target slave node, where the task request includes graph structure features determined based on each task on the target slave node.
其中，目标从节点是指在目标区域中发起任务请求的从节点。任务请求是指对目标从节点上的任务进行任务调度的请求。The target slave node refers to the slave node that initiates the task request in the target area. The task request refers to a request for scheduling the task on the target slave node.
在本申请实施例中，其具体是，基于目标从节点所在区域的主节点，获取目标从节点发送的任务请求。In the embodiment of the present application, specifically, based on the master node of the area where the target slave node is located, the task request sent by the target slave node is obtained.
在本申请实施例中，图结构特征是指根据目标从节点上各任务之间的优先级约束以及依赖关系构建得到的图结构特征。在图结构特征中，包括顶点以及连接两个顶点之间的边。其中，顶点即为任务节点，用于表示目标从节点上的任务，顶点的数量与目标从节点上的任务数量相等。边表示任务之间的继承顺序。In the embodiment of the present application, the graph structure feature refers to the graph structure feature constructed based on the priority constraints and dependencies between the tasks on the target slave node. The graph structure feature includes vertices and edges connecting two vertices. Among them, the vertex is the task node, which is used to represent the task on the target slave node, and the number of vertices is equal to the number of tasks on the target slave node. The edge represents the inheritance order between tasks.
在本申请实施例中，图结构特征可以是有向无环图。图结构特征也可以是仅包括任务顶点以及任务顶点之间的连接关系的结构图。图结构特征的具体内容可根据实际情况进行调整，本申请实施例不做限制。In the embodiment of the present application, the graph structure feature may be a directed acyclic graph. The graph structure feature may also be a structure graph that only includes task vertices and the connection relationship between task vertices. The specific content of the graph structure feature may be adjusted according to actual conditions, and the embodiment of the present application does not limit it.
示例性的，图结构特征为有向无环图，工作流通常由顶点和边组成的有向无环图进行表示。有向无环图是一个元组G=<T,E>，其中，T是工作流任务对应的顶点集，N表示任务总数，/>是反映任务间数据依赖关系的有向边集。 例如，边/>意味着/>和/>之间存在优先约束，即/>是/>的直接前驱（父），/>是/>的直接后继（子）。 每个边/>都有一个权重来表示从/>到/>传输的数据的大小。 一个任务可以有一个或多个父任务或子任务，在执行完所有父任务并且接收到任务所需的所有输入数据之前，该任务不能执行。For example, the graph structure feature is a directed acyclic graph, and the workflow is usually represented by a directed acyclic graph consisting of vertices and edges. A directed acyclic graph is a tuple G=<T,E>, where , T is the vertex set corresponding to the workflow task, N represents the total number of tasks,/> is a set of directed edges that reflect the data dependencies between tasks. For example, the edge/> Means/> and/> There is a precedence constraint between them, namely/> Yes/> The immediate predecessor (parent), /> Yes/> The direct successor (child) of . Each edge/> There is a weight to indicate the To/> The size of the data transferred. A task can have one or more parent tasks or child tasks, and a task cannot be executed until all parent tasks have been executed and all the input data required by the task has been received.
通过有向无环图，让目标从节点所在区域的主节点（调度节点）细粒度地理解目标从节点上业务（各任务组成）的特征。同时，由于发送图结构特征比发送各任务的全部数据的数量少，提高了任务请求的发送速率。Through the directed acyclic graph, the master node (scheduling node) in the area where the target slave node is located can understand the characteristics of the business (composition of each task) on the target slave node in a fine-grained manner. At the same time, since the number of graph structure features sent is less than the number of all data sent for each task, the sending rate of task requests is increased.
S202、基于目标从节点所在区域的主节点上的特征提取模块 ，针对图结构特征上的每个任务节点，对任务节点以及任务节点的领域信息进行特征聚合处理，得到聚合图结构特征。S202. Based on the feature extraction module on the master node in the area where the target slave node is located, for each task node on the graph structure feature, feature aggregation processing is performed on the task node and the domain information of the task node to obtain an aggregated graph structure feature.
其中，特征提取模块包括用于提取图结构特征的全局特征的模块。特征提取模块的实现方式为图神经网络，图神经网络包括但不限于递归神经网络、循环神经网络、马尔可夫网络，其具体选择可根据图结构特征的调整而进行调整，本申请实施例不做限制。Among them, the feature extraction module includes a module for extracting global features of graph structural features. The feature extraction module is implemented as a graph neural network, which includes but is not limited to a recursive neural network, a recurrent neural network, and a Markov network. The specific selection can be adjusted according to the adjustment of the graph structural features, and the embodiments of this application are not limited.
在本申请实施例中，领域信息是指与任务节点具有连接关系的节点对应的信息。该具有连接关系的节点包括邻居节点以及多跳节点。多跳节点是指与任务节点具有连接关系的非邻居节点的其他节点。In the embodiment of the present application, domain information refers to information corresponding to a node having a connection relationship with a task node. The node having a connection relationship includes a neighbor node and a multi-hop node. A multi-hop node refers to other nodes that are not neighbor nodes and have a connection relationship with the task node.
在本申请实施例中，其具体是，基于目标从节点所在区域的主节点上的特征提取模块 ，针对图结构特征上的每个任务节点，在每个任务节点上聚合领域信息，学习任务之间的依赖关系（边的关系信息）和特征表示，从而形成图结构特征对应的全局特征，即聚合图结构特征。In an embodiment of the present application, specifically, based on a feature extraction module on a master node in the region where a target slave node is located, for each task node on a graph structure feature, domain information is aggregated on each task node, and dependencies between tasks (edge relationship information) and feature representations are learned, thereby forming global features corresponding to the graph structure features, i.e., aggregated graph structure features.
S203、将聚合图结构特征输入到深度强化学习模块中，得到目标从节点的任务调度策略。S203: Input the aggregate graph structure features into the deep reinforcement learning module to obtain the task scheduling strategy of the target slave node.
其中，深度强化学习模块包括智能体，用于接收目标从节点的相关信息，并根据目标从节点的相关信息，智能化决策对目标从节点上各任务的任务调度策略。该任务调度策略包括计算任务调度策略以及通信任务调度策略。The deep reinforcement learning module includes an agent, which is used to receive relevant information of the target slave node and make intelligent decisions on the task scheduling strategy for each task on the target slave node based on the relevant information of the target slave node. The task scheduling strategy includes a computing task scheduling strategy and a communication task scheduling strategy.
在本申请实施例中，任务请求包括与目标从节点相连的至少一个其他从节点的信道状态信息，其中，目标从节点可以通过探测数据包发送与接收，在发送任务请求前评估目标从节点与各个区域从节点之间的信道状态信息，并将该信道状态信息作为任务请求中的一个数据发送给目标从节点所在区域的主节点。In an embodiment of the present application, the task request includes channel state information of at least one other slave node connected to the target slave node, wherein the target slave node can send and receive detection data packets, evaluate the channel state information between the target slave node and the slave nodes in each area before sending the task request, and send the channel state information as a data in the task request to the master node in the area where the target slave node is located.
S204、根据任务调度策略，对目标从节点上各任务进行调度分配。S204: According to the task scheduling strategy, schedule and allocate each task on the target slave node.
其中，其具体过程为主节点向目标从节点回复任务调度策略，目标从节点根据任务调度策略对目标从节点上各任务进行调度分配。The specific process is that the master node replies the task scheduling strategy to the target slave node, and the target slave node schedules and allocates each task on the target slave node according to the task scheduling strategy.
在本实施例中，通过获取目标从节点发送的任务请求，任务请求包括基于目标从节点上各任务确定的图结构特征；基于目标从节点所在区域的主节点上的特征提取模块，针对图结构特征上的每个任务节点，对任务节点以及任务节点的领域信息进行特征聚合处理，得到聚合图结构特征；将聚合图结构特征输入到深度强化学习模块中，得到目标从节点的任务调度策略；根据任务调度策略，对目标从节点上各任务进行调度分配。采用本发明提高基于分布式云平台的任务调度的智能化，降低调度节点的网络负载。In this embodiment, by obtaining the task request sent by the target slave node, the task request includes the graph structure features determined based on each task on the target slave node; based on the feature extraction module on the master node in the area where the target slave node is located, for each task node on the graph structure feature, the task node and the domain information of the task node are subjected to feature aggregation processing to obtain the aggregated graph structure feature; the aggregated graph structure feature is input into the deep reinforcement learning module to obtain the task scheduling strategy of the target slave node; according to the task scheduling strategy, each task on the target slave node is scheduled and allocated. The present invention is adopted to improve the intelligence of task scheduling based on the distributed cloud platform and reduce the network load of the scheduling node.
请参阅图3，图3是本申请的基于分布式云平台的任务调度方法的一具体实施例的调度应用框架示意图，如图3所示，客户端（目标从节点）为发送业务请求（任务请求）做准备，客户端向客户端所在区域的主节点发送业务请求，主节点根据获取的信息确定资源调度策略，并向目标从节点回复资源调度策略（任务调度策略），客户端根据任务调度策略发送业务数据（各任务相关数据）。Please refer to Figure 3, which is a schematic diagram of the scheduling application framework of a specific embodiment of the task scheduling method based on a distributed cloud platform of the present application. As shown in Figure 3, the client (target slave node) prepares to send a business request (task request), and the client sends a business request to the master node in the area where the client is located. The master node determines the resource scheduling strategy based on the information obtained, and replies to the target slave node with the resource scheduling strategy (task scheduling strategy). The client sends business data (data related to each task) according to the task scheduling strategy.
在该实施例中，将聚合图结构特征输入到深度强化学习模块中，得到目标从节点的任务调度策略之前，还包括：In this embodiment, before the aggregate graph structure features are input into the deep reinforcement learning module to obtain the task scheduling strategy of the target slave node, the following is also included:
基于目标从节点所在区域的主节点上的感知机模块，对信道状态信息进行特征提取，得到信道状态特征。Based on the perception machine module on the master node in the area where the target slave node is located, the channel state information is feature extracted to obtain the channel state feature.
其中，感知机模块可以是单层感知机模块，也可以是多层感知机模块，其具体可根据实际情况进行调整，本申请实施例不做限制。Among them, the perception machine module can be a single-layer perception machine module or a multi-layer perception machine module, which can be adjusted according to actual conditions and is not limited in the embodiments of the present application.
在该实施例中，使用目标从节点所在区域的主节点上的感知机模块处理信道状态信息，目标从节点到各个从节点的信道状态信息表示为，其中/>表示客户端到第/>个从节点的信道状态信息，J是服务区域中从节点的数量。将目标从节点到各个从节点的信道状态信息作为输入，经过多层感知机网络，提取其特征表示，得到信道状态特征。In this embodiment, the channel state information is processed by the sensor module on the master node in the area where the target slave node is located. The channel state information from the target slave node to each slave node is expressed as , where/> Indicates that the client goes to /> The channel state information of the slave nodes is obtained by taking the channel state information from the target slave node to each slave node as input, and extracting its feature representation through the multi-layer perceptron network to obtain the channel state feature.
在本申请实施例中，在获取目标从节点发送的任务请求之前，还包括：In the embodiment of the present application, before obtaining the task request sent by the target slave node, the method further includes:
获取目标从节点所在区域的各从节点在每一时间步上的负载特征。Obtain the load characteristics of each slave node in the area where the target slave node is located at each time step.
当满足时序特征提取条件时，基于目标从节点所在区域的主节点，对负载特征进行时序特征提取，确定目标从节点所在区域的各从节点的时序特征。When the timing feature extraction conditions are met, the timing feature extraction is performed on the load feature based on the master node in the area where the target slave node is located, and the timing features of each slave node in the area where the target slave node is located are determined.
在该实施例中，可以理解的是，主节点负责管理和协调同一区域内所有从节点，从节点的负载情况周期性或者实时报告给其区域所在的主节点。In this embodiment, it can be understood that the master node is responsible for managing and coordinating all slave nodes in the same area, and the load conditions of the slave nodes are reported to the master node in its area periodically or in real time.
在该实施例中，目标从节点所在区域的主节点周期性或者实时获取目标从节点所在区域的各从节点在每一时间步上的负载特征。In this embodiment, the master node in the area where the target slave node is located obtains the load characteristics of each slave node in the area where the target slave node is located at each time step periodically or in real time.
可以理解的是，每个从节点中包含多个容器，在从节点内部，一个容器会计算一个或几个任务，表示第/>个从节点上第/>个容器的整体负载情况，其中/>表示CPU利用率、/>表示内存使用率、/>，/>表示磁盘I/O速率。用矩阵M表示每个时间步中区域分布式云的负载情况。服务区域中有N个从节点，每个从节点有K个容器。则可以按照如下公式（1）表示矩阵M。It is understandable that each slave node contains multiple containers. Inside the slave node, a container will calculate one or several tasks. Indicates the first/> The first slave node/> The overall load of the container, where /> Indicates CPU utilization,/> Indicates memory usage, /> ,/> Represents the disk I/O rate. The matrix M represents the load of the regional distributed cloud at each time step. There are N slave nodes in the service area, and each slave node has K containers. The matrix M can be expressed as follows: (1).
（1） (1)
在该实施例中，获取目标从节点所在区域的各从节点在每一时间步上的负载特征的方式包括但不限于循环神经网络、递归神经网络。其具体选择可根据实际情况进行调整，本申请实施例不做限制。In this embodiment, the method of obtaining the load characteristics of each slave node in the area where the target slave node is located at each time step includes but is not limited to a recurrent neural network and a recursive neural network. The specific selection can be adjusted according to the actual situation, and the embodiment of the present application does not limit it.
示例性的，循环神经网络按照从节点顺序对T个时间步的区域分布式云负载情况提取时序特征（对矩阵M提取时序特征），其中，每个时间步的输入是/>矩阵中的一行。T为时间步的总数，所有时间步的输入按时间顺序组成序列，其中N是每个中的节点数。循环神经网络按照时间步遍历输入序列，对每个/>矩阵中的每一行数据进行处理，形成每个节点的时序特征，并对得到的时序特征进行拼接，得到目标从节点所在区域的各从节点的时序特征。For example, the recurrent neural network calculates the regional distributed cloud load of T time steps in the order of nodes. Extract time series features (extract time series features from matrix M), where the input for each time step is/> A row in the matrix. T is the total number of time steps, and the inputs of all time steps form a sequence in chronological order. , where N is each The number of nodes in the recurrent neural network. The recurrent neural network traverses the input sequence in time steps, and for each /> Each row of data in the matrix is processed to form the timing characteristics of each node, and the obtained timing characteristics are spliced to obtain the timing characteristics of each slave node in the area where the target slave node is located.
在本申请实施例中，时序特征提取条件包括获取目标从节点发送的任务请求，将聚合图结构特征输入到深度强化学习模块中，得到目标从节点的任务调度策略，包括：将信道状态特征、聚合图结构特征以及时序特征输入到深度强化学习模块中 ，得到目标从节点的任务调度策略。In an embodiment of the present application, the timing feature extraction conditions include obtaining a task request sent by a target slave node, inputting the aggregate graph structure features into a deep reinforcement learning module, and obtaining a task scheduling strategy for the target slave node, including: inputting channel state features, aggregate graph structure features, and timing features into a deep reinforcement learning module to obtain a task scheduling strategy for the target slave node.
在本申请实施例中，任务调度策略包括计算任务调度策略，深度强化学习模块包括估计网络子模块和调度子模块，估计网络子模块的输出为调度子模块的输入，估计网络子模块包括状态价值函数估计网络和优势函数估计网络，状态价值函数估计网络和优势函数估计网络共享输入深度强化学习模块的特征。In an embodiment of the present application, the task scheduling strategy includes a computing task scheduling strategy, the deep reinforcement learning module includes an estimation network sub-module and a scheduling sub-module, the output of the estimation network sub-module is the input of the scheduling sub-module, the estimation network sub-module includes a state-value function estimation network and an advantage function estimation network, and the state-value function estimation network and the advantage function estimation network share the characteristics of the input deep reinforcement learning module.
得到目标从节点的任务调度策略，包括：Get the task scheduling strategy of the target slave node, including:
获取状态价值函数估计网络的第一输出值和优势函数估计网络的第二输出值。Obtain a first output value of the state-value function estimation network and a second output value of the advantage function estimation network.
将第一输出值和第二输出值作为调度子模块的输入，调度子模块还包括状态价值网络。The first output value and the second output value are used as inputs of a scheduling submodule, and the scheduling submodule also includes a state value network.
基于状态价值网络中设置的预定状态价值条件，对第一输出值和第二输出值进行计算资源分析，并根据得到的计算资源结果信息，确定对目标从节点的计算任务调度策略，计算资源结果信息至少包括计算资源结果信息对应的从节点。Based on the predetermined state value conditions set in the state value network, computing resource analysis is performed on the first output value and the second output value, and based on the obtained computing resource result information, a computing task scheduling strategy for the target slave node is determined, and the computing resource result information at least includes the slave node corresponding to the computing resource result information.
其中，调度子模块为状态价值网络，可以理解的是，状态价值函数估计网络和优势函数估计网络以及状态价值网络均有对应的函数实现，其具体可根据实际情况调整，本申请实施例不做限制。Among them, the scheduling submodule is a state value network. It can be understood that the state value function estimation network, the advantage function estimation network and the state value network all have corresponding function implementations, which can be adjusted according to actual conditions and are not limited in the embodiments of this application.
其中，将信道状态特征、聚合图结构特征以及时序特征作为状态输入，输入深度强化学习模块的智能体中，以使深度强化学习模块确定目标从节点的计算任务调度策略。Among them, the channel state characteristics, aggregate graph structure characteristics and timing characteristics are used as state inputs and input into the intelligent agent of the deep reinforcement learning module, so that the deep reinforcement learning module determines the computing task scheduling strategy of the target slave node.
其中，深度强化学习模块包括状态空间，该状态空间包括分布式云平台的计算资源状态（各从节点的状态）、信道状态信息以及目标从节点的各任务。Among them, the deep reinforcement learning module includes a state space, which includes the computing resource state of the distributed cloud platform (the state of each slave node), channel state information, and each task of the target slave node.
其中，信道状态特征、聚合图结构特征以及时序特征被估计网络子模块中的估计网络共享，估计网络子模块包括状态价值函数估计网络和优势函数估计网络，两个估计网络的输出值组成调度子模块的输入。调度子模块为状态价值函数Q，Q以最大化奖励值为目标，总共输出δ个值，δ由动作空间决定，动作空间是深度强化学习模块的策略空间，动作空间具有多个维度，其维度包括从节点序号、四个计算资源的规格大小、参考码本。动作的所有维度的取值不得超出其对应资源约束范围，δ是动作中不同维度取值的总数量。取出使得输出Q值序列最大的动作作为针对当前状态的资源调度策略，即根据得到的计算资源结果信息，确定对目标从节点的计算任务调度策略，计算资源结果信息至少包括计算资源结果信息对应的从节点。Among them, the channel state features, aggregate graph structure features and timing features are shared by the estimation network in the estimation network submodule. The estimation network submodule includes a state value function estimation network and an advantage function estimation network. The output values of the two estimation networks constitute the input of the scheduling submodule. The scheduling submodule is the state value function Q. Q aims to maximize the reward value and outputs a total of δ values. δ is determined by the action space. The action space is the strategy space of the deep reinforcement learning module. The action space has multiple dimensions, including the slave node number, the size of the four computing resources, and the reference codebook. The values of all dimensions of the action must not exceed the corresponding resource constraint range. δ is the total number of values of different dimensions in the action. Take out the action that maximizes the output Q value sequence as the resource scheduling strategy for the current state, that is, according to the obtained computing resource result information, determine the computing task scheduling strategy for the target slave node. The computing resource result information at least includes the slave node corresponding to the computing resource result information.
其中，在确定计算任务调度策略的过程中，还需要进行目标函数的优化过程。Among them, in the process of determining the computing task scheduling strategy, it is also necessary to optimize the objective function.
示例性的，以各任务执行时间和成本作为双重优化目标，各任务的执行时间T分为传输时间和计算时间/>。任务执行的成本主要是从节点上的计算能耗P。For example, the execution time and cost of each task are used as dual optimization objectives. The execution time T of each task is divided into transmission time and calculation time/> The cost of task execution is mainly derived from the computational energy consumption P on the node.
则可以按照如下公式（2）来进行目标函数的优化：Then the objective function can be optimized according to the following formula (2):
（2） (2)
其中，是过去所有任务的平均执行时间，/>是平均计算能耗。/>是权衡因子用来调整深度强化学习模块调度的重点是放在时间上还是能耗上。/>用于监测执行失败的情况。如果任务执行失败或资源分配不合理，导致 />、/>或P趋于无穷大，则会引入一个负值来惩罚这种情况。in, is the average execution time of all tasks in the past, /> is the average computing energy consumption. /> It is a trade-off factor used to adjust the scheduling of deep reinforcement learning modules to focus on time or energy consumption. /> Used to monitor execution failures. If a task fails or resource allocation is unreasonable, it will result in/> 、/> Or if P tends to infinity, a negative value is introduced to penalize this situation.
深度强化学习模块根据状态为当前任务做出计算方面的决策，即选择一个最合适的从节点和计算资源。值得注意的是，计算资源的选择是为当前业务的计算限定其最大使用的CPU利用率、内存使用率/>、磁盘I/O速率/>，/>。为减小动作空间的维度。The deep reinforcement learning module makes computing decisions for the current task based on the state, that is, selects the most suitable slave node and computing resources. It is worth noting that the choice of computing resources is to limit the maximum CPU utilization for the current business calculation. , memory usage/> , disk I/O rate/> ,/> . To reduce the dimension of the action space.
具体的，请参阅图4，图4为本申请的基于分布式云平台的任务调度方法的又一具体实施例的调度应用框架示意图，如图4所示，深度强化学习模块的智能体接收来自分布式云平台的多个输入（信道状态特征、聚合图结构特征以及时序特征），根据智能体上的智能体决策网络以及奖励函数，进行从节点选择，计算资源规格的确定以及参考码本的选择，其中，参考码本的选择可根据定制化码本池进行构造，从而确定对目标从节点的计算任务调度策略（从节点选择，计算资源规格的确定等）以及通信任务调度策略（参考码本的选择）。Specifically, please refer to Figure 4, which is a schematic diagram of a scheduling application framework of another specific embodiment of the task scheduling method based on a distributed cloud platform of the present application. As shown in Figure 4, the intelligent agent of the deep reinforcement learning module receives multiple inputs (channel state characteristics, aggregate graph structure characteristics, and timing characteristics) from the distributed cloud platform, and selects slave nodes, determines computing resource specifications, and selects reference codebooks based on the intelligent agent decision network and reward function on the intelligent agent. The selection of the reference codebook can be constructed based on a customized codebook pool to determine the computing task scheduling strategy (slave node selection, determination of computing resource specifications, etc.) and the communication task scheduling strategy (reference codebook selection) for the target slave node.
在本申请实施例中，任务调度策略包括通信任务调度策略，得到目标从节点的任务调度策略，包括：In the embodiment of the present application, the task scheduling strategy includes a communication task scheduling strategy, and obtaining the task scheduling strategy of the target slave node includes:
将聚合图结构特征与目标从节点上各任务的历史记录进行聚类处理，确定与聚合图结构特征对应的第一参考码本，其中，目标从节点上各任务的历史记录包括历史聚合图结构特征以及与历史聚合图结构特征对应的第二参考码本。The aggregate graph structure feature and the historical records of each task on the target slave node are clustered to determine a first reference codebook corresponding to the aggregate graph structure feature, wherein the historical records of each task on the target slave node include the historical aggregate graph structure feature and a second reference codebook corresponding to the historical aggregate graph structure feature.
基于预定码本池，将第一参考码本与预定码本池中的码本进行相似度计算，并根据计算得到的结果确定对目标从节点的通信任务调度策略。Based on the predetermined codebook pool, similarity calculation is performed between the first reference codebook and the codebooks in the predetermined codebook pool, and a communication task scheduling strategy for the target slave node is determined according to the calculation result.
在本申请实施例中，第一参考码本为目标从节点向各从节点发送任务时的参考码本。In the embodiment of the present application, the first reference codebook is a reference codebook used when the target slave node sends a task to each slave node.
可以理解的是，在传统的通信资源调度场景下，一次性传输大容量数据（例如，大文件或大规模数据集）会给无线通信网络带来巨大的负载和加剧资源竞争。通信方面的决策是模糊的，可以通过参考码本来确定目标从节点向各从节点发送任务时的参考码本，参考码本的规格与数据传输时间息息相关。It is understandable that in traditional communication resource scheduling scenarios, transmitting large amounts of data (e.g., large files or large data sets) at one time will bring huge loads to the wireless communication network and intensify resource competition. Communication decisions are ambiguous, and the reference codebook can be used to determine the reference codebook when the target slave node sends tasks to each slave node. The specifications of the reference codebook are closely related to the data transmission time.
在本申请实施例中，参考码本包括但不限于IrSCMA码本（Irregular Sparse CodeMultiple Access，非正规稀疏码分多址）、LDS_CDMA码本（low-density spreading CDMA，低密度扩频CDMA）、LDS-OFDM码本（low-density spreading OFDM，低密度扩频OFDM）、MUSA码本（Multi User Shared Access，多用户共享接入）。则预定码本池为参考码本对应的码本池。例如，当参考码本为IrSCMA码本时，预定码本池为IrSCMA码本对应的码本池。参考码本以及预定码本池的具体内容可根据实际情况进行调整，本申请实施例不做限制。In the embodiment of the present application, the reference codebook includes but is not limited to the IrSCMA codebook (Irregular Sparse Code Multiple Access), the LDS_CDMA codebook (low-density spreading CDMA), the LDS-OFDM codebook (low-density spreading OFDM), and the MUSA codebook (Multi User Shared Access). The predetermined codebook pool is the codebook pool corresponding to the reference codebook. For example, when the reference codebook is the IrSCMA codebook, the predetermined codebook pool is the codebook pool corresponding to the IrSCMA codebook. The specific contents of the reference codebook and the predetermined codebook pool can be adjusted according to actual conditions, and the embodiment of the present application does not limit it.
示例性的，为解决上述问题，深度强化学习模块给予的IrSCMA码本 仅作为该任务的参考码本，从同属于一类的历史任务对应的码本中提取该时刻未使用的码本以及总IrSCMA码本池（预定码本池）中构造针对当前任务的定制化码本池，支持目标从节点上各任务的细粒化数据传输，从而提高通信效率降低网络负载。也就是说，码本与通信有关，若动作空间太大，通信直接按照参考码本去调度。Exemplarily, to solve the above problem, the IrSCMA codebook given by the deep reinforcement learning module is only used as a reference codebook for the task. The unused codebook at this moment is extracted from the codebook corresponding to the historical tasks of the same category, and a customized codebook pool for the current task is constructed from the total IrSCMA codebook pool (predetermined codebook pool), supporting fine-grained data transmission of each task on the target slave node, thereby improving communication efficiency and reducing network load. In other words, the codebook is related to communication. If the action space is too large, the communication is directly scheduled according to the reference codebook.
在本申请实施例中，其具体过程为；将聚合图结构特征或者图结构特征与历史记录（[聚合图结构特征，第二参考码本]）进行聚类，从中获得相似任务的参考码本（第一参考码本）。然后从第一参考码本出发，通过与IrSCMA码本池中其他码本的结构相似度计算，结构相似度的计算是码本的行数、列数、非零行的位置、数量和其中元素取值。In the embodiment of the present application, the specific process is: clustering the aggregated graph structure features or graph structure features with the historical records ([aggregated graph structure features, second reference codebook]), and obtaining a reference codebook (first reference codebook) for similar tasks. Then, starting from the first reference codebook, the structural similarity with other codebooks in the IrSCMA codebook pool is calculated, and the calculation of the structural similarity is the number of rows, columns, positions and number of non-zero rows of the codebook and the values of the elements therein.
示例性的，参考码本为A，IrSCMA中其他码本表示为。NZ(A)表示A中的非零行的索引集合，NZ(B) 表示B中的非零行的索引集合。其中，，其中，/>表示码本的行数、表示码本的列数、/>表示码本的非零行数。/> ，分别是矩阵A和B的非零行元素的标准差。按照如下公式（3）计算相似度：For example, the reference codebook is A, and other codebooks in IrSCMA are represented as NZ(A) represents the index set of non-zero rows in A, and NZ(B) represents the index set of non-zero rows in B. , where /> Indicates the number of lines in the codebook, Indicates the number of columns in the codebook, /> Indicates the number of non-zero rows in the codebook. /> , are the standard deviations of the non-zero row elements of matrices A and B, respectively. The similarity is calculated according to the following formula (3):
（3） (3)
通过上述方式，筛选出与参考码本相似度在范围的码本，码本相似度过小会增减传输过程中数据间的干扰，过大又会使任务的传输达不到细粒度，最终共同构成针对当前任务的定制化IrSCMA码本池。码本数量不小于目标从节点内部任务的数量。Through the above method, the codebooks with similarity in the range If the codebook similarity is too small, it will increase or decrease the interference between data during transmission. If it is too large, the transmission of the task will not reach fine granularity. Finally, they together constitute a customized IrSCMA codebook pool for the current task. The number of codebooks is not less than the number of tasks inside the target slave node.
需要理解的是，在通常情况下，继承关系涉及到某些任务必须在其他任务之前完成，这对于任务计算顺序的影响是显著的。然而，对于数据传输而言，任务之间的依赖关系更多地关注数据的可用性，而非任务的顺序性。因此，任务之间的继承关系对于数据传输的顺序并没有直接影响，数据的可用性可能会根据任务执行的实际情况和需求进行调整和优化。此外，数据是并行传输的意味着可以同时处理多个数据传输任务，而不必等待之前的传输完成。It is important to understand that, in general, inheritance relationships involve certain tasks that must be completed before other tasks, which has a significant impact on the order in which tasks are calculated. However, for data transfer, the dependencies between tasks focus more on the availability of data rather than the order of tasks. Therefore, the inheritance relationship between tasks has no direct impact on the order of data transfer, and the availability of data may be adjusted and optimized based on the actual situation and needs of task execution. In addition, the fact that data is transferred in parallel means that multiple data transfer tasks can be processed simultaneously without having to wait for the previous transfer to complete.
请参阅图5，图5为本申请的基于分布式云平台的任务调度方法中在IrSCMA中通过johnson规则的调度应用框架示意图。Please refer to FIG. 5 , which is a schematic diagram of a scheduling application framework using the Johnson rule in IrSCMA in the distributed cloud platform-based task scheduling method of the present application.
如图5所示，根据图结构特征中任务之间的继承关系将任务划分为数据传输的不同阶段，具体而言，将所有的入口任务视为第一个阶段，入口任务的直接后继任务视为第二个阶段，以此类推，直到没有直接后继任务的出口任务被视为最后一个阶段。每个任务都要被先编码后调制，每个并行的IrSCMA发送设备可以同时执行数据块的编码和传输操作，但一个任务的两个操作均由同一个IrSCMA发送设备执行。以最小化makespan为目标，利用johnson规则优化两个工序不同任务执行的重叠时间。在每一个阶段中，根据Johnson规则重新调整任务数据包的顺序，使得最后一个完成第二个工序传输的任务的结束时间（makespan）最小化。对于每个任务首先确定编码和传输两个操作的所需时间。将每个阶段中所有任务按照第一个工序所需的时间排序。对于第一阶段的排好序的任务，再按照第二个工序传输时间最短的顺序重新调整每个阶段中任务的顺序，按照调整后的顺序传输任务，以最小化最后一个完成第二个工序传输的任务结束时间。目标从节点上各任务（业务）的数据被目标从节点（客户端）直接传输至决策指定的从节点上。As shown in Figure 5, tasks are divided into different stages of data transmission according to the inheritance relationship between tasks in the graph structure feature. Specifically, all entry tasks are regarded as the first stage, the direct successor tasks of the entry tasks are regarded as the second stage, and so on, until the exit tasks without direct successor tasks are regarded as the last stage. Each task must be encoded and modulated first. Each parallel IrSCMA sending device can perform the encoding and transmission operations of the data block at the same time, but the two operations of a task are performed by the same IrSCMA sending device. With the goal of minimizing makespan, the Johnson rule is used to optimize the overlapping time of the execution of different tasks in two processes. In each stage, the order of task data packets is readjusted according to the Johnson rule so that the end time (makespan) of the last task to complete the transmission of the second process is minimized. For each task, the time required for the two operations of encoding and transmission is first determined. All tasks in each stage are sorted according to the time required for the first process. For the sorted tasks in the first stage, the order of tasks in each stage is readjusted according to the order with the shortest transmission time of the second process, and the tasks are transmitted in the adjusted order to minimize the end time of the last task to complete the transmission of the second process. The data of each task (business) on the target slave node is directly transmitted by the target slave node (client) to the slave node specified by the decision.
其中，Johnson规则是一种现有的流水线调度方法。其具体过程包括：Among them, Johnson rule is an existing pipeline scheduling method. Its specific process includes:
a、列出业务的工序矩阵，每个阶段的任务的工序矩阵视为一个单独的子工序矩阵；a. List the process matrix of the business, and the process matrix of the task in each stage is regarded as a separate sub-process matrix;
b、在每个子工序矩阵中选出加工时间最短的工序。如果该工序属于第1工序，则将该工序所属工件排在前面。反之，最小工序是第2工序，则将该工序所属的工件排在最后面；b. Select the process with the shortest processing time in each sub-process matrix. If the process belongs to process 1, the workpiece to which the process belongs is placed in the front. On the contrary, if the shortest process is process 2, the workpiece to which the process belongs is placed in the back;
c、将已排序的任务从工序矩阵中消去；c. Eliminate the sorted tasks from the process matrix;
d、继续按步骤a、b、c进行排序，若所有任务都已排定投产顺序，排序即告结束。d. Continue to sort according to steps a, b, and c. If all tasks have been scheduled for production, the sorting is complete.
在本申请实施例中，在IrSCMA中通过johnson规则来减少数据传输时间。In an embodiment of the present application, the Johnson rule is used in IrSCMA to reduce data transmission time.
在本申请实施例中，实现调度策略下发至用户、业务数据直接发送至从节点的同时，结合IrSCMA通信技术，有效缓解调度节点的网络拥塞情况。进一步的，基于深度强化学习的混合资源分配方法，利用图卷积神经网络提取业务的有向无环图特征，通过将深度强化学习的注意力转移到计算资源的分配，同时采用模糊化的方式处理通信资源分配，加速决策过程。最后，使用聚类算法从历史调度记录和IrSCMA码本池构建针对当前业务的定制化码本池，将任务之间的继承关系转化为数据传输的不同阶段，利用Johnson规则优化整个传输过程的时间开销，以提高传输效率和整体性能。In an embodiment of the present application, the scheduling strategy is sent to the user and the service data is sent directly to the slave node, and the IrSCMA communication technology is combined to effectively alleviate the network congestion of the scheduling node. Furthermore, a hybrid resource allocation method based on deep reinforcement learning uses a graph convolutional neural network to extract the directed acyclic graph features of the service, and by shifting the attention of deep reinforcement learning to the allocation of computing resources, a fuzzy method is used to process the allocation of communication resources to accelerate the decision-making process. Finally, a clustering algorithm is used to build a customized codebook pool for the current business from historical scheduling records and the IrSCMA codebook pool, and the inheritance relationship between tasks is converted into different stages of data transmission. The Johnson rule is used to optimize the time overhead of the entire transmission process to improve transmission efficiency and overall performance.
应理解，上述实施例中各步骤的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本发明实施例的实施过程构成任何限定。It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.
图6示出与上述实施例基于分布式云平台的任务调度方法一一对应的基于分布式云平台的任务调度装置的原理框图。如图6所示，该基于分布式云平台的任务调度装置包括任务请求获取模块301、聚合模块302、任务调度策略确定模块303和调度模块304。各功能模块详细说明如下：FIG6 shows a principle block diagram of a task scheduling device based on a distributed cloud platform corresponding to the task scheduling method based on a distributed cloud platform in the above embodiment. As shown in FIG6 , the task scheduling device based on a distributed cloud platform includes a task request acquisition module 301, an aggregation module 302, a task scheduling strategy determination module 303 and a scheduling module 304. Each functional module is described in detail as follows:
任务请求获取模块301，用于获取目标从节点发送的任务请求，其中，任务请求包括基于目标从节点上各任务确定对应的图结构特征 。The task request acquisition module 301 is used to acquire the task request sent by the target slave node, wherein the task request includes the graph structure features corresponding to each task on the target slave node.
聚合模块302，用于基于目标从节点所在区域的主节点上的特征提取模块 ，针对图结构特征上的每个任务节点，对任务节点以及任务节点的领域信息进行特征聚合处理，得到聚合图结构特征。Aggregation module 302 is used to perform feature aggregation processing on task nodes and domain information of task nodes for each task node on the graph structure feature based on the feature extraction module on the master node of the area where the target slave node is located, so as to obtain aggregated graph structure features.
任务调度策略确定模块303，用于将聚合图结构特征输入到深度强化学习模块中，得到目标从节点的任务调度策略。The task scheduling strategy determination module 303 is used to input the aggregate graph structure features into the deep reinforcement learning module to obtain the task scheduling strategy of the target slave node.
调度模块304，用于根据任务调度策略，对目标从节点上各任务进行调度分配。The scheduling module 304 is used to schedule and allocate each task on the target slave node according to the task scheduling strategy.
可选的，在本申请实施例中的装置上，任务请求包括与目标从节点相连的至少一个其他从节点的信道状态信息。Optionally, on the device in the embodiment of the present application, the task request includes channel state information of at least one other slave node connected to the target slave node.
在任务调度策略确定模块303之前，任务调度策略确定模块之前，还包括：Before the task scheduling strategy determination module 303, before the task scheduling strategy determination module, it also includes:
信道状态特征确定模块，用于基于目标从节点所在区域的主节点上的感知机模块，对信道状态信息进行特征提取，得到信道状态特征。The channel state feature determination module is used to extract features of the channel state information based on the perception machine module on the master node in the area where the target slave node is located to obtain the channel state features.
可选的，在本申请实施例中的装置上，在任务调度策略确定模块303之前，在获取目标从节点发送的任务请求之前，还包括：Optionally, on the device in the embodiment of the present application, before the task scheduling strategy determination module 303, before obtaining the task request sent by the target slave node, it also includes:
负载特征获取模块，用于获取目标从节点所在区域的各从节点在每一时间步上的负载特征。The load characteristic acquisition module is used to obtain the load characteristics of each slave node in the area where the target slave node is located at each time step.
时序特征获取模块，用于当满足时序特征提取条件时，基于目标从节点所在区域的主节点，对负载特征进行时序特征提取，确定目标从节点所在区域的各从节点的时序特征。The timing feature acquisition module is used to extract the timing features of the load features based on the master node in the area where the target slave node is located when the timing feature extraction conditions are met, and determine the timing features of each slave node in the area where the target slave node is located.
可选的，在本申请实施例中的装置上，时序特征提取条件包括获取目标从节点发送的任务请求，任务调度策略确定模块303中，将聚合图结构特征输入到深度强化学习模块中，得到目标从节点的任务调度策略，包括：Optionally, in the device in the embodiment of the present application, the time series feature extraction condition includes obtaining a task request sent by the target slave node, and in the task scheduling strategy determination module 303, the aggregate graph structure feature is input into the deep reinforcement learning module to obtain the task scheduling strategy of the target slave node, including:
任务调度策略确定单元，用于将信道状态特征、聚合图结构特征以及时序特征输入到深度强化学习模块中 ，得到目标从节点的任务调度策略。The task scheduling strategy determination unit is used to input the channel state features, the aggregate graph structure features and the timing features into the deep reinforcement learning module to obtain the task scheduling strategy of the target slave node.
可选的，在本申请实施例中的装置上，任务调度策略包括计算任务调度策略，深度强化学习模块包括估计网络子模块和调度子模块，估计网络子模块的输出为调度子模块的输入，估计网络子模块包括状态价值函数估计网络和优势函数估计网络，状态价值函数估计网络和优势函数估计网络共享输入深度强化学习模块的特征。Optionally, on the device in the embodiment of the present application, the task scheduling strategy includes a calculation task scheduling strategy, the deep reinforcement learning module includes an estimation network sub-module and a scheduling sub-module, the output of the estimation network sub-module is the input of the scheduling sub-module, the estimation network sub-module includes a state-value function estimation network and an advantage function estimation network, and the state-value function estimation network and the advantage function estimation network share the characteristics of the input deep reinforcement learning module.
任务调度策略确定模块303中，得到目标从节点的任务调度策略，包括：In the task scheduling strategy determination module 303, the task scheduling strategy of the target slave node is obtained, including:
输出值确定单元，用于获取状态价值函数估计网络的第一输出值和优势函数估计网络的第二输出值。The output value determination unit is used to obtain the first output value of the state value function estimation network and the second output value of the advantage function estimation network.
输入单元，用于将第一输出值和第二输出值作为调度子模块的输入，调度子模块还包括状态价值网络。The input unit is used to use the first output value and the second output value as inputs of the scheduling submodule, and the scheduling submodule also includes a state value network.
计算任务调度策略确定单元，用于基于状态价值网络中设置的预定状态价值条件，对第一输出值和第二输出值进行计算资源分析，并根据得到的计算资源结果信息，确定对目标从节点的计算任务调度策略，计算资源结果信息至少包括计算资源结果信息对应的从节点。A computing task scheduling strategy determination unit is used to perform computing resource analysis on the first output value and the second output value based on a predetermined state value condition set in the state value network, and determine the computing task scheduling strategy for the target slave node based on the obtained computing resource result information, and the computing resource result information at least includes the slave node corresponding to the computing resource result information.
可选的，在本申请实施例中的装置上，任务调度策略包括通信任务调度策略，任务调度策略确定模块303中，得到目标从节点的任务调度策略，包括：Optionally, in the device in the embodiment of the present application, the task scheduling strategy includes a communication task scheduling strategy, and in the task scheduling strategy determination module 303, the task scheduling strategy of the target slave node is obtained, including:
第二参考码本确定单元，用于将聚合图结构特征与目标从节点上各任务的历史记录进行聚类处理，确定与聚合图结构特征对应的第一参考码本，其中，目标从节点上各任务的历史记录包括历史聚合图结构特征以及与历史聚合图结构特征对应的第二参考码本。The second reference codebook determination unit is used to cluster the aggregate graph structure feature and the historical records of each task on the target slave node to determine a first reference codebook corresponding to the aggregate graph structure feature, wherein the historical records of each task on the target slave node include the historical aggregate graph structure feature and the second reference codebook corresponding to the historical aggregate graph structure feature.
通信任务调度策略确定单元，用于基于预定码本池，将第一参考码本与预定码本池中的码本进行相似度计算，并根据计算得到的结果确定对目标从节点的通信任务调度策略。The communication task scheduling strategy determining unit is used to calculate the similarity between the first reference codebook and the codebooks in the predetermined codebook pool based on the predetermined codebook pool, and determine the communication task scheduling strategy for the target slave node according to the calculated result.
关于基于分布式云平台的任务调度装置的具体限定可以参见上文中对于基于分布式云平台的任务调度方法的限定，在此不再赘述。上述基于分布式云平台的任务调度装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中，也可以以软件形式存储于计算机设备中的存储器中，以便于处理器调用执行以上各个模块对应的操作。For the specific limitations of the task scheduling device based on the distributed cloud platform, please refer to the limitations of the task scheduling method based on the distributed cloud platform above, which will not be repeated here. Each module in the above-mentioned task scheduling device based on the distributed cloud platform can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.
为解决上述技术问题，本申请实施例还提供计算机设备。具体请参阅图7，图7为本实施例计算机设备基本结构框图。In order to solve the above technical problems, the embodiment of the present application also provides a computer device. Please refer to Figure 7 for details, which is a basic structural block diagram of the computer device of this embodiment.
所述计算机设备4包括通过系统总线相互通信连接存储器41、处理器42、网络接口43。需要指出的是，图中仅示出了具有组件连接存储器41、处理器42、网络接口43的计算机设备4，但是应理解的是，并不要求实施所有示出的组件，可以替代的实施更多或者更少的组件。其中，本技术领域技术人员可以理解，这里的计算机设备是一种能够按照事先设定或存储的指令，自动进行数值计算和/或信息处理的设备，其硬件包括但不限于微处理器、专用集成电路(Application Specific Integrated Circuit，ASIC)、可编程门阵列(Field－Programmable Gate Array，FPGA)、数字处理器 (Digital Signal Processor，DSP)、嵌入式设备等。The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components connected to the memory 41, the processor 42, and the network interface 43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and/or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.
所述计算机设备可以是桌上型计算机、笔记本、掌上电脑及云端主节点等计算设备。所述计算机设备可以与用户通过键盘、鼠标、遥控器、触摸板或声控设备等方式进行人机交互。The computer device may be a desktop computer, a notebook, a PDA, a cloud master node, etc. The computer device may interact with the user through a keyboard, a mouse, a remote control, a touch pad, or a voice control device.
所述存储器41至少包括一种类型的可读存储介质，所述可读存储介质包括闪存、硬盘、多媒体卡、卡型存储器（例如，SD或D界面显示存储器等）、随机访问存储器（RAM）、静态随机访问存储器（SRAM）、只读存储器（ROM）、电可擦除可编程只读存储器（EEPROM）、可编程只读存储器（PROM）、磁性存储器、磁盘、光盘等。在一些实施例中，所述存储器41可以是所述计算机设备4的内部存储单元，例如该计算机设备4的硬盘或内存。在另一些实施例中，所述存储器41也可以是所述计算机设备4的外部存储设备，例如该计算机设备4上配备的插接式硬盘，智能存储卡（Smart Media Card, SMC），安全数字（Secure Digital, SD）卡，闪存卡（Flash Card）等。当然，所述存储器41还可以既包括所述计算机设备4的内部存储单元也包括其外部存储设备。本实施例中，所述存储器41通常用于存储安装于所述计算机设备4的操作系统和各类应用软件，例如电子文件的控制的程序代码等。此外，所述存储器41还可以用于暂时地存储已经输出或者将要输出的各类数据。The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or D interface display memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as the program code for controlling electronic files, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.
所述处理器42在一些实施例中可以是中央处理器（Central Processing Unit，CPU）、控制器、微控制器、微处理器、或其他数据处理芯片。该处理器42通常用于控制所述计算机设备4的总体操作。本实施例中，所述处理器42用于运行所述存储器41中存储的程序代码或者处理数据，例如运行电子文件的控制的程序代码。The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run the program code stored in the memory 41 or process data, such as running the program code for controlling the electronic file.
所述网络接口43可包括无线网络接口或有线网络接口，该网络接口43通常用于在所述计算机设备4与其他电子设备之间建立通信连接。The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.
本申请还提供了另一种实施方式，即提供一种计算机可读存储介质，所述计算机可读存储介质存储有界面显示程序，所述界面显示程序可被至少一个处理器执行，以使所述至少一个处理器执行如上述的基于分布式云平台的任务调度方法的步骤。The present application also provides another implementation, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores an interface display program, and the interface display program can be executed by at least one processor so that the at least one processor performs the steps of the task scheduling method based on the distributed cloud platform as described above.
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质（如ROM/RAM、磁碟、光盘）中，包括若干指令用以使得一台从节点（可以是手机，计算机，主节点，空调器，或者网络设备等）执行本申请各个实施例所述的方法。Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, a disk, or an optical disk), and includes a number of instructions for enabling a slave node (which can be a mobile phone, a computer, a master node, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.
显然，以上所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例，附图中给出了本申请的较佳实施例，但并不限制本申请的专利范围。本申请可以以许多不同的形式来实现，相反地，提供这些实施例的目的是使对本申请的公开内容的理解更加透彻全面。尽管参照前述实施例对本申请进行了详细的说明，对于本领域的技术人员来而言，其依然可以对前述各具体实施方式所记载的技术方案进行修改，或者对其中部分技术特征进行等效替换。凡是利用本申请说明书及附图内容所做的等效结构，直接或间接运用在其他相关的技术领域，均同理在本申请专利保护范围之内。Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to replace some of the technical features therein with equivalents. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is also within the scope of patent protection of this application.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410168488.2ACN117707797B (en) | 2024-02-06 | 2024-02-06 | Task scheduling method and device based on distributed cloud platform and related equipment |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410168488.2ACN117707797B (en) | 2024-02-06 | 2024-02-06 | Task scheduling method and device based on distributed cloud platform and related equipment |
| Publication Number | Publication Date |
|---|---|
| CN117707797A CN117707797A (en) | 2024-03-15 |
| CN117707797Btrue CN117707797B (en) | 2024-05-03 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410168488.2AActiveCN117707797B (en) | 2024-02-06 | 2024-02-06 | Task scheduling method and device based on distributed cloud platform and related equipment |
| Country | Link |
|---|---|
| CN (1) | CN117707797B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119071292B (en)* | 2024-08-26 | 2025-09-26 | 浙江大学 | A distributed single-node scheduling method based on multi-layer perceptron decision making |
| CN119276863B (en)* | 2024-09-27 | 2025-09-26 | 蔚来汽车科技(安徽)有限公司 | A distributed framework implementation method, electronic device, and storage medium |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105005570A (en)* | 2014-04-23 | 2015-10-28 | 国家电网公司 | Method and apparatus for mining massive intelligent power consumption data based on cloud computing |
| CN112148451A (en)* | 2020-09-27 | 2020-12-29 | 南京大学 | Low-delay collaborative self-adaptive CNN inference system and method |
| CN112486641A (en)* | 2020-11-18 | 2021-03-12 | 鹏城实验室 | Task scheduling method based on graph neural network |
| CN112817728A (en)* | 2021-02-20 | 2021-05-18 | 咪咕音乐有限公司 | Task scheduling method, network device and storage medium |
| CN114860398A (en)* | 2022-04-21 | 2022-08-05 | 郑州大学 | A task scheduling method, device and device for an intelligent cloud platform |
| CN115309521A (en)* | 2022-07-25 | 2022-11-08 | 哈尔滨工业大学（深圳） | Deep reinforcement learning task scheduling method and device for maritime unmanned equipment |
| CN115480882A (en)* | 2021-05-31 | 2022-12-16 | 中移雄安信息通信科技有限公司 | Distributed edge cloud resource scheduling method and system |
| CN115794341A (en)* | 2022-11-16 | 2023-03-14 | 中国平安财产保险股份有限公司 | Task scheduling method, device, equipment and storage medium based on artificial intelligence |
| CN116302467A (en)* | 2022-12-09 | 2023-06-23 | 中国联合网络通信集团有限公司 | Task allocation method, device and storage medium |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8059733B2 (en)* | 2006-12-20 | 2011-11-15 | Nec Laboratories America, Inc. | Multi-user downlink linear MIMO precoding systems |
| CN110580196B (en)* | 2019-09-12 | 2021-04-06 | 北京邮电大学 | Multi-task reinforcement learning method for realizing parallel task scheduling |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105005570A (en)* | 2014-04-23 | 2015-10-28 | 国家电网公司 | Method and apparatus for mining massive intelligent power consumption data based on cloud computing |
| CN112148451A (en)* | 2020-09-27 | 2020-12-29 | 南京大学 | Low-delay collaborative self-adaptive CNN inference system and method |
| CN112486641A (en)* | 2020-11-18 | 2021-03-12 | 鹏城实验室 | Task scheduling method based on graph neural network |
| CN112817728A (en)* | 2021-02-20 | 2021-05-18 | 咪咕音乐有限公司 | Task scheduling method, network device and storage medium |
| CN115480882A (en)* | 2021-05-31 | 2022-12-16 | 中移雄安信息通信科技有限公司 | Distributed edge cloud resource scheduling method and system |
| CN114860398A (en)* | 2022-04-21 | 2022-08-05 | 郑州大学 | A task scheduling method, device and device for an intelligent cloud platform |
| CN115309521A (en)* | 2022-07-25 | 2022-11-08 | 哈尔滨工业大学（深圳） | Deep reinforcement learning task scheduling method and device for maritime unmanned equipment |
| CN115794341A (en)* | 2022-11-16 | 2023-03-14 | 中国平安财产保险股份有限公司 | Task scheduling method, device, equipment and storage medium based on artificial intelligence |
| CN116302467A (en)* | 2022-12-09 | 2023-06-23 | 中国联合网络通信集团有限公司 | Task allocation method, device and storage medium |
| Title |
|---|
| 云系统中面向海量多媒体数据的动态任务调度算法;朱映映 等;小型微型计算机系统;20130415(第04期);全文* |
| Publication number | Publication date |
|---|---|
| CN117707797A (en) | 2024-03-15 |
| Publication | Publication Date | Title |
|---|---|---|
| Chen et al. | Latency minimization for mobile edge computing networks | |
| CN117707797B (en) | Task scheduling method and device based on distributed cloud platform and related equipment | |
| CN112162865A (en) | Server scheduling method and device and server | |
| CN104484233B (en) | A kind of resource allocation methods | |
| CN111511028B (en) | Multi-user resource allocation method, device, system and storage medium | |
| CN114980216B (en) | Dependency task unloading system and method based on mobile edge calculation | |
| CN108304256A (en) | A low-overhead task scheduling method and device in edge computing | |
| CN114675845B (en) | Information age optimization method, device, computer equipment and storage medium | |
| CN113778675A (en) | Calculation task distribution system and method based on block chain network | |
| CN114860449A (en) | Data processing method, device, equipment and storage medium | |
| CN120123099B (en) | Resource scheduling method, system, equipment and medium in edge cloud computing platform | |
| CN110489224A (en) | A kind of method and apparatus of task schedule | |
| CN110912967A (en) | Service node scheduling method, device, equipment and storage medium | |
| Li et al. | A cost‐efficient and QoS‐aware adaptive placement of applications in fog computing | |
| CN114866612B (en) | Electric power micro-service unloading method and device | |
| Guo et al. | PARA: Performability‐aware resource allocation on the edges for cloud‐native services | |
| CN113891466A (en) | Online scheduling system and method for UDL task in edge wireless network | |
| CN118502898A (en) | Task processing method, system, device and storage medium | |
| CN118349333A (en) | Task scheduling method, system, device and readable storage medium | |
| CN115361285B (en) | Method, device, equipment and medium for realizing off-line service mixed deployment | |
| CN117651044A (en) | Edge computing task scheduling method and device | |
| CN117950833A (en) | Task scheduling method, device, computer equipment and storage medium | |
| CN117806951A (en) | Intelligent scheduling method and related equipment applied to distributed operation of test cases | |
| CN117873640A (en) | Virtual machine scheduling method, device, equipment and storage medium based on NUMA architecture | |
| CN116974747A (en) | Resource allocation method, device, equipment, medium and program product |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |