Dispersed resource sensing fusion method and device, medium, equipment and product

Abstract

The invention discloses a distributed resource sensing fusion method and device, a medium, equipment and a product in the technical field of distributed computing, and aims to solve the problems of limited sensing range, relatively weak sensing capability and relatively poor fusion in the prior art. The method comprises the following steps: carrying out resource modeling according to computing resources, storage resources and communication resources of network heterogeneous computing equipment to obtain a resource model; according to the basic attributes of the network separable tasks, the dependency relationship between the tasks and the execution constraint of the tasks, performing task modeling by using a directed acyclic graph to obtain a task model; dynamically sensing and adjusting the use state of the resources in the resource model by adopting a bottom-up clustering fusion method; and according to task requirements in the task model and the use states of the resources, dynamically sensing in a task execution process by adopting a local multi-hop fusion method, and fusing the sensed resources. According to the method, the problems of information fragmentation and redundancy in scattered resources are solved, and the task processing efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of distributed computing technology, and in particular to a distributed resource perception fusion method, device, medium, equipment and product. Background Art

[0002] Distributed computing is a computing paradigm that relies on modern information infrastructure and equipment. It can distribute task requirements to spatially distributed heterogeneous computing device node tasks in a self-organizing dynamic networking manner based on real-time network characteristics, and use the available computing resources on the communication network path to jointly optimize computing costs and communication overheads to achieve more efficient, faster, and more accurate task processing. From the perspective of technical origin, the emergence of distributed computing stems from the needs of rescue and disaster relief tasks. When the disaster site is far away from the city command center, there will be limited communication and computing resources, rapid changes in the disaster area, and unstable network connections.

[0003] In a distributed computing environment, a single heterogeneous computing device has a limited perception range, weak perception capabilities, and different models of heterogeneous devices. Different heterogeneous devices need to be integrated into the same architecture to cooperate. It is necessary to propose a distributed environment heterogeneous computing device perception framework so that heterogeneous devices with different perception algorithms can cooperate for perception. Without sacrificing the perception capability of a single device, the perception range and perception accuracy of all intelligent agents can be logically improved.

[0004] The purpose of decentralized resource perception is to collect and integrate computing resource information from different heterogeneous data sources, and provide key computing device information collection and maintenance functions for decentralized resource scheduling and task load balancing allocation models. Through perception, a single heterogeneous computing device can understand the status, performance, load, etc. of each computing device, so as to reasonably allocate and schedule tasks according to demand. At the same time, it is necessary to establish a unified modeling method to integrate the perceived resource information into a unified framework to facilitate subsequent analysis and processing. Summary of the invention

[0005] The purpose of the present invention is to overcome the deficiencies in the prior art and to provide a distributed resource perception fusion method, device, medium, equipment and product that can solve the problems of information fragmentation and redundancy in distributed resources and improve the processing efficiency of tasks.

[0006] In order to solve the above technical problems, the present invention is implemented by adopting the following technical solutions:

[0007] The present invention provides a decentralized resource perception fusion method, comprising:

[0008] Perform resource modeling based on computing resources, storage resources and communication resources of network heterogeneous computing devices to obtain a resource model;

[0009] According to the basic properties of network separable tasks, the dependencies between tasks and the execution constraints of tasks, the task modeling is carried out using directed acyclic graph to obtain the task model;

[0010] A bottom-up clustering fusion method is used to dynamically sense and adjust the usage status of computing resources, storage resources and communication resources in the resource model, so as to realize resource-driven decentralized resource perception fusion;

[0011] According to the task requirements in the task model and the usage status of computing resources, storage resources and communication resources in the resource model, a local multi-hop fusion method is used to dynamically perceive resources during task execution, and the perceived resources are fused to achieve task-driven decentralized resource perception fusion.

[0012] Optionally, the resource modeling includes:

[0013] The computing resources, storage resources, and communication resources of each heterogeneous computing device are represented in the form of tuple feature vectors to obtain the tuple definition of the resource model. , a single heterogeneous computing device The resource representation is ;

[0014] in, represents a vertex set, where the vertices represent heterogeneous computing devices, represents an edge set, where an edge represents the connection between two heterogeneous computing devices. represents the set of computing speeds of heterogeneous computing devices processing tasks, Represents the maximum capacity set that heterogeneous computing devices can store. represents the set of communication speeds of heterogeneous computing devices, Indicates Heterogeneous computing devices, Indicates The computing speed of tasks processed by heterogeneous computing devices, Indicates The maximum capacity that a heterogeneous computing device can store, Indicates The communication speed of heterogeneous computing devices.

[0015] Optionally, the task modeling includes:

[0016] (a1) Based on the basic properties of network separable tasks, the following definitions are made:

[0017]

[0018] Among them, each task is divided into a finite number of node tasks. Indicates the program The size of the data stored and processed by the task is a collection of Indicates the program The set of task workloads, Indicates the program The maximum processing time set that the tasks can accept, Indicates the program The number of node tasks, Indicates the program No. The amount of data stored and processed by each node task, Indicates the program No. The task workload of each node task, Indicates the program No. The maximum processing time that a node task can accept;

[0019] (b1) Define the amount of data to be transferred , represents the data information carried by the communication between adjacent node tasks with sequential dependencies;

[0020] (c1) Use directed acyclic graph DAG to describe the task set. Each task includes several node tasks, and the following expression is obtained;

[0021]

[0022] in, represents the task graph, Representing a task graph The set of finite node tasks in , Representing a task graph The set of directed edges in Indicates the program No. Node tasks, Indicates the program From the first Node tasks to There are directed edges between node tasks, and node tasks For node tasks Parent task, node task The execution priority is greater than the node task ;

[0023] Combining the definitions and descriptions in (a1) to (c1), we get the tuple definition of the task model. .

[0024] Optionally, the dynamically sensing and adjusting the usage status of resources in the resource model by using a bottom-up clustering fusion method includes:

[0025] The heterogeneous computing devices are associated by periodically broadcasting heartbeat packets, wherein the heartbeat packets contain resource representations of the heterogeneous computing devices;

[0026] According to the sensing range of each heterogeneous computing device, each heterogeneous computing device is divided into several clusters; wherein the sensing ranges of any two heterogeneous computing devices in a cluster overlap and can sense each other;

[0027] The computing resources, storage resources and communication resources of each heterogeneous computing device in each cluster are comprehensively scored according to different weights, and the heterogeneous computing device with the highest comprehensive score is used as the cluster head device of the cluster;

[0028] Generate or update the resource adjacency table of the heterogeneous computing devices of all clusters until all heterogeneous computing devices are sensed to form a global virtual resource pool; wherein the resource adjacency table is a one-dimensional array, and the elements in the group are the numbers of other heterogeneous computing devices accessible to the current heterogeneous computing device, representing the set of other heterogeneous computing devices accessible to the current heterogeneous computing device, and the resource adjacency table of the cluster head device stores the resource representation of all heterogeneous computing devices in the cluster.

[0029] Optionally, the comprehensive score is calculated using the following formula:

[0030]

[0031] in, Indicates The comprehensive score of heterogeneous computing devices, Indicates the weight of computing resources. Indicates the weight of storage resources. Indicates the weight of communication resources.

[0032] Optionally, a local multi-hop fusion method is used to dynamically sense resources during task execution and to fuse the sensed resources, including:

[0033] According to the task requirements in the task model, it is determined whether the computing resources, storage resources, and communication resources of the heterogeneous computing device currently receiving the task meet the task requirements. If they meet, the task is processed directly; otherwise, recursive fusion within the one-hop perception range is triggered:

[0034] According to the resource adjacency table of the heterogeneous computing device currently receiving the task, obtain the first-level perception device of the heterogeneous computing device currently receiving the task; perform first-level fusion of the computing resources, storage resources and communication resources of the heterogeneous computing device currently receiving the task and the first-level perception device of the heterogeneous computing device currently receiving the task to obtain the first-level fused resources; determine whether the first-level fused resources meet the task requirements, and directly process them if they meet; otherwise, trigger recursive fusion within the two-hop perception range:

[0035] According to the resource adjacency list of other heterogeneous computing devices accessible to the heterogeneous computing device currently receiving the task, obtain the secondary perception device of the heterogeneous computing device currently receiving the task; perform secondary fusion of the computing resources, storage resources and communication resources of the secondary perception device of the heterogeneous computing device currently receiving the task with the primary fusion resources to obtain the secondary fusion resources; determine whether the secondary fusion resources meet the task requirements, and directly process them if they do; otherwise, assign the current task to any heterogeneous computing device in the resource model except the heterogeneous computing device currently receiving the task and its primary perception device and secondary perception device;

[0036] Repeat the above steps until the task is completed, or all heterogeneous computing devices in the resource model cannot meet the task requirements and the task processing fails.

[0037] In a second aspect, the present invention provides a distributed resource perception fusion device, comprising:

[0038] The resource modeling module is used to: perform resource modeling according to the computing resources, storage resources, and communication resources of the network heterogeneous computing devices to obtain a resource model;

[0039] The task modeling module is used to: perform task modeling using a directed acyclic graph according to the basic properties of network-dividable tasks, dependencies between tasks, and execution constraints of tasks to obtain a task model;

[0040] A resource-driven module is used to: dynamically sense and adjust the usage status of computing resources, storage resources and communication resources in the resource model using a bottom-up clustering fusion method to achieve resource-driven decentralized resource perception fusion;

[0041] The task-driven module is used to: dynamically perceive resources during task execution using a local multi-hop fusion method according to the task requirements in the task model and the usage status of computing resources, storage resources and communication resources in the resource model, and fuse the perceived resources to achieve task-driven decentralized resource perception fusion.

[0042] In a third aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the distributed resource awareness fusion method described in the first aspect.

[0043] In a fourth aspect, the present invention provides a computer device, comprising:

[0044] Memory, for storing computer instructions;

[0045] A processor is used to execute the computer instructions to implement the steps of the decentralized resource awareness fusion method described in the first aspect.

[0046] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions, characterized in that when the computer instructions are executed by a processor, the steps of the distributed resource awareness fusion method described in the first aspect are implemented.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. The decentralized resource perception fusion method provided by the present invention models device resources and tasks so that independent resources and tasks can be associated. When decomposing and allocating tasks, it is only necessary to find resource vectors that meet the node task requirements, and the resource status and task allocation situation can be intuitively understood in the form of embedded graphs. In addition, the decentralized resource perception fusion method based on resource and task drive can flexibly adapt to different situations. In the case of task requirements, the resource-driven perception fusion method can effectively manage the resources of computing devices and form a virtual resource pool. When the task requirements arrive, the task-driven resource perception fusion method can respond quickly to the task and meet the high sensitivity requirements of tasks to latency in decentralized environments.

[0049] 2. The distributed resource perception fusion device provided by the present invention realizes the perception fusion of distributed resources by setting a resource modeling module, a task modeling module, a resource driving module and a task driving module, which can solve the information fragmentation and redundancy problems in distributed resources and improve the task processing efficiency, and has practical significance and good application prospects;

[0050] 3. The computer-readable storage medium, computer device / equipment / system and computer program product provided by the present invention can execute the steps of the distributed resource perception fusion method provided by the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A flow chart of a decentralized resource awareness fusion method provided according to an embodiment of the present invention;

[0052] Figure 2 A flow chart of resource-driven decentralized resource perception fusion provided according to an embodiment of the present invention;

[0053] Figure 3A flowchart of task-driven decentralized resource perception fusion provided according to an embodiment of the present invention;

[0054] Figure 4 A schematic diagram of a clustering example provided according to an embodiment of the present invention;

[0055] Figure 5 A schematic diagram of an example of task-driven decentralized resource-aware fusion provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical solution of the present application, rather than limitations on the technical solution of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0057] It should be noted that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0058] Embodiment 1:

[0059] The embodiment of the present invention discloses a distributed resource perception fusion method, referring to Figure 1 As shown, the specific steps include:

[0060] S1, performing resource modeling according to computing resources, storage resources and communication resources of network heterogeneous computing devices to obtain a resource model;

[0061] S2, based on the basic properties of network separable tasks, the dependencies between tasks and the execution constraints of tasks, the task modeling is performed using a directed acyclic graph to obtain a task model;

[0062] S3, using a bottom-up clustering fusion method to dynamically perceive and adjust the usage status of computing resources, storage resources and communication resources in the resource model, to achieve resource-driven decentralized resource perception fusion;

[0063] S4, according to the task requirements in the task model and the usage status of computing resources, storage resources and communication resources in the resource model, a local multi-hop fusion method is used to dynamically perceive resources during task execution, and the perceived resources are integrated to achieve task-driven decentralized resource perception fusion.

[0064] Specifically,

[0065] In step S1, this embodiment characterizes the resources of a single heterogeneous computing device in the form of a tuple feature vector, and forms a large resource network graph consisting of a distributed computing network composed of several heterogeneous computing devices. ; The definition of resource structure includes the basic attributes of resources, the dynamic characteristics of resources and the constraints of resources; the basic attributes of resources include the following aspects: resource UID represents the unique identifier of resources, which is used to distinguish different resources; resource type represents that there are multiple resources, and only computing resources, storage resources and communication resources are considered in this embodiment; computing resources are the core resources required for processing tasks and provide computing power support; storage resources determine the data storage capacity of the device; communication resources determine the data transmission capacity of the device; in summary, the resource modeling includes:

[0066] Each heterogeneous computing device resource is represented by a tuple feature vector to obtain the tuple definition of the resource model. , a single heterogeneous computing device The resource representation is ;

[0067] in, represents a vertex set, where vertices represent heterogeneous computing devices. represents an edge set, where an edge represents the connection between two heterogeneous computing devices. represents the set of computing speeds of heterogeneous computing devices processing tasks, Represents the maximum capacity set that heterogeneous computing devices can store. represents the set of communication speeds of heterogeneous computing devices, Indicates Heterogeneous computing devices, Indicates The computing speed of tasks processed by heterogeneous computing devices, Indicates The maximum capacity that a heterogeneous computing device can store, Indicates The communication speed of heterogeneous computing devices.

[0068] In step S2, the definition of the task structure includes the basic attributes of the task, the dependencies between tasks, and the execution constraints of the task; the basic attributes of the task include the following aspects: the task ID represents the unique identifier of the task, which is used to distinguish different tasks; the resource requirement represents the resources required for task execution, which are consistent with the basic attributes described in the resource model; the execution time represents the expected execution time of the task; the priority is used to determine the importance and urgency of the task; in summary, the task modeling includes:

[0069] (a1) Based on the basic properties of network separable tasks, the following definitions are made:

[0070]

[0071] Among them, each task can be divided into a finite number of node tasks. Indicates the program The size of the data stored and processed by the task is a collection of Indicates the program The set of task workloads, Indicates the program The maximum processing time set that the tasks can accept, Indicates the program The number of node tasks, Indicates the program No. The amount of data stored and processed by each node task, Indicates the program No. The task workload of each node task, Indicates the program No. The maximum processing time that a node task can accept;

[0072] (b1) Define the amount of data to be transferred , indicating that the communication between adjacent node tasks with sequential dependencies will carry a small amount of data information, that is, the input information of the latter will be obtained from the output information of the previous one;

[0073] (c1) Use a directed acyclic graph (DAG) to describe the task set. Each task can have several node tasks. A single task and its node tasks are decomposed into a DAG graph, and the following expression is obtained;

[0074]

[0075] in, represents the task graph, Indicates the program No. Node tasks, Representing a task graph The set of finite node tasks in , Representing a task graph The set of directed edges in Indicates the program From the first Node tasks to directed edges of node tasks, and tasks for The parent task of the task The execution priority is higher than the task ;

[0076] Combining the definitions and descriptions in (a1) to (c1), we get the tuple definition of the task model. .

[0077] In step S3, this embodiment takes decentralized resources as the core and adopts a bottom-up clustering and fusion method; heterogeneous computing devices regularly broadcast heartbeat packets every 60 seconds to associate computing devices with other computing devices in the neighborhood; the heartbeat packets contain resource representations of the resources of each heterogeneous computing device; a number of computing devices are clustered according to the sensing range, and cluster heads are selected within the cluster to fuse each other's resources; refer to Figure 2 The resource-driven decentralized resource perception fusion flow chart shown in the figure, the bottom-up clustering fusion method is used to dynamically perceive and adjust the usage status of resources in the resource model, including:

[0078] The heterogeneous computing devices are associated by periodically broadcasting heartbeat packets, wherein the heartbeat packets contain resource representations of the heterogeneous computing devices;

[0079] According to the sensing range of each heterogeneous computing device, each heterogeneous computing device is divided into several clusters; wherein the sensing ranges of any two heterogeneous computing devices in a cluster overlap and can sense each other;

[0080] The computing resources, storage resources and communication resources of each heterogeneous computing device in each cluster are comprehensively scored according to different weights, and the heterogeneous computing device with the highest comprehensive score is used as the cluster head device of the cluster;

[0081] Generate or update the resource adjacency table of the heterogeneous computing devices of all clusters until all heterogeneous computing devices are sensed to form a global virtual resource pool; wherein the resource adjacency table is a one-dimensional array, and the elements in the group are the numbers of other heterogeneous computing devices accessible to the current heterogeneous computing device, representing the set of other heterogeneous computing devices accessible to the current heterogeneous computing device, and the resource adjacency table of the cluster head device stores the resource representation of all heterogeneous computing devices in the cluster.

[0082] In this embodiment, the comprehensive score is calculated by the following formula:

[0083]

[0084] in, Indicates The comprehensive score of heterogeneous computing devices, Indicates the weight of computing resources. Indicates the weight of storage resources. Indicates the weight of communication resources.

[0085] In computing-intensive scenarios, computing resources, as the core resources of the device, account for the largest proportion, followed by the need to save corresponding information within the node task. Therefore, the weight of storage resources is greater than that of communication resources. Therefore, the weight distribution can be referred to The above weight distribution is only for reference. Different weights should be allocated according to different scenario requirements. For example, if you encounter a scenario that is sensitive to latency, you can consider increasing the proportion of communication resources.

[0086] In step S4, this embodiment takes task requirements as the core and adopts a local multi-hop fusion method; dynamically optimizes and utilizes according to task requirements and resource status, emphasizing dynamic perception during task execution; when a heterogeneous computing device receives a task processing request, it first decides whether to process the task based on the computing power of the node task itself; if there is a node task that has all the resources to process the task and meets the task constraints, it responds to the task quickly; otherwise, it will be fused according to other node tasks within the range to collaboratively process the task; refer to Figure 3 As shown in the figure, a local multi-hop fusion method is used to dynamically sense resources during task execution and fuse the sensed resources, including:

[0087] According to the task requirements in the task model, it is determined whether the computing resources, storage resources, and communication resources of the heterogeneous computing device currently receiving the task meet the task requirements. If they meet, the task is processed directly; otherwise, recursive fusion within the one-hop perception range is triggered:

[0088] According to the resource adjacency table of the heterogeneous computing device currently receiving the task, obtain the first-level perception device of the heterogeneous computing device currently receiving the task; perform first-level fusion of the computing resources, storage resources and communication resources of the heterogeneous computing device currently receiving the task and the first-level perception device of the heterogeneous computing device currently receiving the task to obtain the first-level fused resources; determine whether the first-level fused resources meet the task requirements, and directly process them if they meet; otherwise, trigger recursive fusion within the two-hop perception range:

[0089] According to the resource adjacency list of other heterogeneous computing devices accessible to the heterogeneous computing device currently receiving the task, obtain the secondary perception device of the heterogeneous computing device currently receiving the task; perform secondary fusion of the computing resources, storage resources and communication resources of the secondary perception device of the heterogeneous computing device currently receiving the task with the primary fusion resources to obtain the secondary fusion resources; determine whether the secondary fusion resources meet the task requirements, and directly process them if they do; otherwise, assign the current task to any heterogeneous computing device in the resource model except the heterogeneous computing device currently receiving the task and its primary perception device and secondary perception device;

[0090] Repeat the above steps until the task is completed, or all heterogeneous computing devices in the resource model cannot meet the task requirements and the task processing fails.

[0091] To help understand the specific operation method of this embodiment, the following operation examples are given:

[0092] refer to Figure 4 The clustering example diagram shown here assumes that a decentralized environment exists. , , , , , Six heterogeneous computing devices, each with different resource attributes and perception ranges.

[0093] Heterogeneous computing devices The perception range has heterogeneous computing devices ;

[0094] Heterogeneous computing devices The perception range has heterogeneous computing devices , heterogeneous computing devices and heterogeneous computing devices ;

[0095] Heterogeneous computing devices The perception range has heterogeneous computing devices and heterogeneous computing devices ;

[0096] Heterogeneous computing devices The perception range has heterogeneous computing devices and heterogeneous computing devices ;

[0097] Heterogeneous computing devices The perception range has heterogeneous computing devices and heterogeneous computing devices ;

[0098] Heterogeneous computing devices Not within the perception of any node task.

[0099] When performing perception fusion, heterogeneous computing devices and The fusion is performed and the fusion resource adjacency table is saved in each node task, which connects heterogeneous computing devices. and Fusion, while on heterogeneous computing devices and The fused resource adjacency table is saved in the internal storage; thus heterogeneous computing devices There are two fused resource adjacency tables in the system. You can consider updating them during fusion. The resource adjacency table stored in the system contains all the heterogeneous computing devices. , and The resource attributes of the resource adjacency table are then updated to the heterogeneous computing device during the next heartbeat detection. and Achieve local unification. Then, in the same way, through the intermediate heterogeneous computing device Heterogeneous computing devices and heterogeneous computing devices Make an association.

[0100] This fusion method starts from the most marginal distributed devices and integrates resources in the form of clusters, like a tree structure, and finally achieves global resource unification; however, the problem is that there will eventually be a super central node task in the node task network, and this node task stores a large number of resource attributes of distributed devices. If the node task fails, data loss will occur. The reason for this problem is that the resource adjacency table is only stored in a single node task. After fusion, the local resource adjacency table can be updated to other node tasks within the range during the next heartbeat detection to achieve redundant backup and improve performance by sacrificing some storage resources; because the storage of the adjacency table only takes up a small amount of memory attributes, this method is feasible.

[0101] Reference Figure 5 The figure shows a schematic diagram of a task-driven decentralized resource-aware fusion example. , , , and The current task is decomposed into five node tasks that are executed sequentially. , , , , and For six distributed heterogeneous computing devices, the distributed resource awareness fusion process includes:

[0102] First, according to heterogeneous computing devices If the resource cannot be processed, the heterogeneous computing device within the sensing range will be used to judge Fusion, after fusion, heterogeneous computing devices and heterogeneous computing devices Collaborative task processing; because tasks arrive at each computing device in parallel, all idle computing devices in the environment can process the task request in parallel, and finally select a node task that meets the constraints and has the best response time for processing.

[0103] Similarly, when heterogeneous computing devices When receiving a task request, if it cannot be processed, a fusion of the one-hop perception range is performed through heterogeneous computing devices. and heterogeneous computing devices If the minimum resource requirements for processing the task cannot be met through collaborative processing, then consider increasing the number of collaborative node tasks through secondary perception, so that heterogeneous computing devices can be used. , , Co-processing, which obviously goes beyond heterogeneous computing devices Although the one-hop perception range can bring more resource attributes, the complexity of the system will increase exponentially and the delay will be greater.

[0104] To sum up, the distributed resource perception fusion method proposed in this embodiment models device resources and tasks so that independent resources and tasks can be associated. When decomposing and allocating tasks, it is only necessary to find resource vectors that meet the node task requirements. The resource status and task allocation situation can be intuitively understood in the form of embedded graphs. In addition, the distributed resource perception fusion method based on resource and task driving can flexibly adapt to different situations. In the case of task requirements, the resource-driven perception fusion method can effectively manage the resources of the computing device to form a virtual resource pool. When the task requirements arrive, the task-driven resource perception fusion method can respond quickly to the task and meet the high sensitivity of tasks to latency in a distributed environment.

[0105] Embodiment 2:

[0106] Based on the same inventive concept as the first embodiment, the embodiment of the present invention discloses a distributed resource perception fusion device, including:

[0107] The resource modeling module is used to: perform resource modeling according to the computing resources, storage resources, and communication resources of the network heterogeneous computing devices to obtain a resource model;

[0108] The task modeling module is used to: perform task modeling using a directed acyclic graph according to the basic properties of network-dividable tasks, dependencies between tasks, and execution constraints of tasks to obtain a task model;

[0109] A resource-driven module is used to: dynamically sense and adjust the usage status of computing resources, storage resources and communication resources in the resource model using a bottom-up clustering fusion method to achieve resource-driven decentralized resource perception fusion;

[0110] The task-driven module is used to: dynamically perceive resources during task execution using a local multi-hop fusion method according to the task requirements in the task model and the usage status of computing resources, storage resources and communication resources in the resource model, and fuse the perceived resources to achieve task-driven decentralized resource perception fusion.

[0111] The specific functional implementation of each of the above modules can be found in the relevant content of the method in Example 1 and will not be elaborated here.

[0112] Embodiment three:

[0113] This embodiment provides a computer-readable storage medium on which computer instructions are stored. When the computer instructions are executed by a processor, the steps of the decentralized resource awareness fusion method described in the first embodiment are implemented.

[0114] Embodiment 4:

[0115] This embodiment provides a computer device, including:

[0116] Memory, for storing computer instructions;

[0117] The processor is used to execute the computer instructions to implement the steps of the decentralized resource awareness fusion method described in the first embodiment.

[0118] Embodiment five:

[0119] This embodiment provides a computer program product, including computer instructions, characterized in that when the computer instructions are executed by a processor, the steps of the distributed resource awareness fusion method described in the first embodiment are implemented.

[0120] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0121] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0122] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0123] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0124] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A distributed resource perception fusion method, characterized in that: include: Perform resource modeling based on computing resources, storage resources and communication resources of network heterogeneous computing devices to obtain a resource model; According to the basic properties of network separable tasks, the dependencies between tasks and the execution constraints of tasks, the task modeling is carried out using directed acyclic graph to obtain the task model; A bottom-up clustering fusion method is used to dynamically sense and adjust the usage status of computing resources, storage resources and communication resources in the resource model, so as to realize resource-driven decentralized resource perception fusion; According to the task requirements in the task model and the usage status of computing resources, storage resources and communication resources in the resource model, a local multi-hop fusion method is used to dynamically perceive resources during task execution, and the perceived resources are fused to achieve task-driven decentralized resource perception fusion.

2. The distributed resource perception fusion method according to claim 1 is characterized in that: The resource modeling includes: The computing resources, storage resources, and communication resources of each heterogeneous computing device are represented in the form of tuple feature vectors to obtain the tuple definition of the resource model. , a single heterogeneous computing device The resource representation is ; in, represents a vertex set, where vertices represent heterogeneous computing devices. represents an edge set, where an edge represents the connection between two heterogeneous computing devices. represents the set of computing speeds of heterogeneous computing devices processing tasks, Represents the maximum capacity set that heterogeneous computing devices can store. represents the set of communication speeds of heterogeneous computing devices, Indicates Heterogeneous computing devices, Indicates The computing speed of tasks processed by heterogeneous computing devices, Indicates The maximum capacity that a heterogeneous computing device can store, Indicates The communication speed of heterogeneous computing devices.

3. The distributed resource perception fusion method according to claim 1 is characterized in that: The task modeling includes: (a1) Based on the basic properties of network separable tasks, the following definitions are made: ， Among them, each task is divided into a finite number of node tasks. Indicates the program The size of the data stored and processed by the task is a collection of Indicates the program The set of task workloads, Indicates the program The maximum processing time set that the tasks can accept, Indicates the program The number of node tasks, Indicates the program No. The amount of data stored and processed by each node task, Indicates the program No. The task workload of each node task, Indicates the program No. The maximum processing time that a node task can accept; (b1) Define the amount of data to be transferred , represents the data information carried by the communication between adjacent node tasks with sequential dependencies; (c1) Use directed acyclic graph DAG to describe the task set. Each task includes several node tasks, and the following expression is obtained; ， in, represents the task graph, Representing a task graph The set of finite node tasks in , Representing a task graph The set of directed edges in Indicates the program No. Node tasks, Indicates the program From the Node tasks to There are directed edges between node tasks, and node tasks For node tasks Parent task, node task The execution priority is greater than the node task ; Combining the definitions and descriptions in (a1) to (c1), we get the tuple definition of the task model. .

4. The distributed resource perception fusion method according to claim 2 is characterized in that: The method of dynamically sensing and adjusting the usage status of resources in the resource model by using a bottom-up clustering fusion method includes: The heterogeneous computing devices are associated by periodically broadcasting heartbeat packets, wherein the heartbeat packets contain resource representations of the heterogeneous computing devices; According to the sensing range of each heterogeneous computing device, each heterogeneous computing device is divided into several clusters; wherein the sensing ranges of any two heterogeneous computing devices in a cluster overlap and can sense each other; The computing resources, storage resources and communication resources of each heterogeneous computing device in each cluster are comprehensively scored according to different weights, and the heterogeneous computing device with the highest comprehensive score is used as the cluster head device of the cluster; Generate or update the resource adjacency table of the heterogeneous computing devices of all clusters until all heterogeneous computing devices are sensed to form a global virtual resource pool; wherein the resource adjacency table is a one-dimensional array, and the elements in the group are the numbers of other heterogeneous computing devices accessible to the current heterogeneous computing device, representing the set of other heterogeneous computing devices accessible to the current heterogeneous computing device, and the resource adjacency table of the cluster head device stores the resource representation of all heterogeneous computing devices in the cluster.

5. The distributed resource perception fusion method according to claim 4 is characterized in that: The comprehensive score is calculated by the following formula: ， in, Indicates The comprehensive score of heterogeneous computing devices, Indicates the weight of computing resources. Indicates the weight of storage resources. Indicates the weight of communication resources.

6. The distributed resource perception fusion method according to claim 4 is characterized in that: The local multi-hop fusion method is used to dynamically sense resources during task execution and fuse the sensed resources, including: According to the task requirements in the task model, it is determined whether the computing resources, storage resources, and communication resources of the heterogeneous computing device currently receiving the task meet the task requirements. If they meet, the task is processed directly; otherwise, recursive fusion within the one-hop perception range is triggered: According to the resource adjacency table of the heterogeneous computing device currently receiving the task, obtain the first-level perception device of the heterogeneous computing device currently receiving the task; perform first-level fusion of the computing resources, storage resources and communication resources of the heterogeneous computing device currently receiving the task and the first-level perception device of the heterogeneous computing device currently receiving the task to obtain the first-level fused resources; determine whether the first-level fused resources meet the task requirements, and directly process them if they meet; otherwise, trigger recursive fusion within the two-hop perception range: According to the resource adjacency list of other heterogeneous computing devices accessible to the heterogeneous computing device currently receiving the task, obtain the secondary perception device of the heterogeneous computing device currently receiving the task; perform secondary fusion of the computing resources, storage resources and communication resources of the secondary perception device of the heterogeneous computing device currently receiving the task with the primary fusion resources to obtain the secondary fusion resources; determine whether the secondary fusion resources meet the task requirements, and directly process them if they do; otherwise, assign the current task to any heterogeneous computing device in the resource model except the heterogeneous computing device currently receiving the task and its primary perception device and secondary perception device; Repeat the above steps until the task is completed, or all heterogeneous computing devices in the resource model cannot meet the task requirements and the task processing fails.

7. A distributed resource perception fusion device, characterized in that: include: The resource modeling module is used to: perform resource modeling according to the computing resources, storage resources, and communication resources of the network heterogeneous computing devices to obtain a resource model; The task modeling module is used to: perform task modeling using a directed acyclic graph according to the basic properties of network-dividable tasks, dependencies between tasks, and execution constraints of tasks to obtain a task model; A resource-driven module is used to: dynamically sense and adjust the usage status of computing resources, storage resources and communication resources in the resource model using a bottom-up clustering fusion method to achieve resource-driven decentralized resource perception fusion; The task-driven module is used to: dynamically perceive resources during task execution using a local multi-hop fusion method according to the task requirements in the task model and the usage status of computing resources, storage resources and communication resources in the resource model, and fuse the perceived resources to achieve task-driven decentralized resource perception fusion.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the decentralized resource awareness fusion method described in any one of claims 1 to 6 are implemented.

9. A computer device, characterized in that: include: Memory, for storing computer instructions; A processor is used to execute the computer instructions to implement the steps of the decentralized resource awareness fusion method according to any one of claims 1 to 6.

10. A computer program product comprising computer instructions, characterized in that: When the computer instructions are executed by a processor, the steps of the decentralized resource awareness fusion method described in any one of claims 1 to 6 are implemented.

Images