Computing power information notification method and device, computing power node, system, and storage medium

Abstract

The disclosure provides a computing power information announcement method and device, a computing power node, a system and a storage medium. The method applied to a first computing power node comprises: obtaining first link quality information between the first computing power node and a second computing power node in a computing power network, first computing power information of the first computing power node; obtaining second link quality information between the second computing power node and other second computing power nodes and the first computing power node, second computing power information of the second computing power node, and generating first computing power network graph data; generating at least one announcement path based on the first link quality information and the second link quality information, and performing transmission processing on the first computing power network graph data. The disclosure can reduce the synchronization time of the computing power node for the computing power network graph data and make all network computing power nodes reach a computing power information synchronization state under the condition of low announcement overhead, can meet the demand of the computing power network for real-time synchronization of resource information, and improves the use experience of users.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a method, apparatus, computing node, computing network system, and storage medium for disseminating computing power information. Background Technology

[0002] In a computing power network, computing nodes need to obtain resource information such as their own computing power resources and network resources, and then announce this information to other computing nodes to achieve collaborative awareness of resources across the entire computing power network. Currently, the computing power announcement technology used in computing power networks uses existing routing protocol messages to carry resource information, synchronizing resource information by transmitting these messages between computing nodes during route convergence. However, synchronizing resource information in a computing power network using existing routing protocol messages has problems. For example, the route convergence time of existing routing protocols is relatively long, which cannot meet the real-time synchronization requirements of computing power networks for resource information. Increasing the sending frequency of routing protocol messages will increase network overhead and may lead to problems such as route oscillations, affecting network quality. Summary of the Invention

[0003] In view of this, one technical problem to be solved by the present invention is to provide a method, apparatus, computing node, computing network system and storage medium for disseminating computing power information.

[0004] According to a first aspect of this disclosure, a method for announcing computing power information is provided, applied to a first computing power node in a computing power network, comprising: acquiring first link quality information between the first computing power node and a second computing power node in the computing power network, and first computing power information of the first computing power node; acquiring second link quality information between the second computing power node and other second computing power nodes, and between the second computing power node and the first computing power node, and second computing power information of the second computing power node; generating first computing power network graph data based on the first link quality information, the first computing power information, and the second computing power information; generating at least one announcement path based on the first link quality information and the second link quality information; and performing transmission processing on the first computing power network graph data according to the at least one announcement path.

[0005] Optionally, the first link quality information includes: first latency information, and the second link quality information includes: second latency information; generating at least one announcement path based on the first link quality information and the second link quality information includes: determining the at least one announcement path according to the first latency information and the second latency information; generating at least one first computing power data packet corresponding to the at least one announcement path based on the first computing power network graph data, and adding an announcement path in the header of each first computing power data packet.

[0006] Optionally, determining the at least one notification path based on the first latency information and the second latency information includes: constructing coordinate information of the first computing power node and the second computing power node based on the first latency information and the second latency information; constructing a self-organizing neural network model, determining the winning neurons of the first computing power node and the second computing power node in the self-organizing neural network model based on the coordinate information; generating multiple notification paths based on the winning neurons; wherein the notification path is a circular path with the first computing power node as the starting point and the ending point, and the circular path passes through all the second computing power nodes.

[0007] Optionally, the coordinate information includes: a two-dimensional matrix of computing node coordinates; the step of constructing the coordinate information of the first computing node and the second computing node based on the first delay information and the second delay information includes: constructing a two-dimensional delay matrix based on the first delay information and the second delay information; wherein, the row vector of the two-dimensional delay matrix is ​​the delay information vector of the first computing node or the second computing node; and the two-dimensional coordinate matrix of the computing node is constructed based on the two-dimensional delay matrix.

[0008] Optionally, the number of rows in the two-dimensional matrix of computing power node coordinates is the sum of the number of the first computing power node and the second computing power node; the number of columns in the two-dimensional matrix of computing power node coordinates is the dimension of the neuron vector in the self-organizing neural network model; determining the winning neuron of the first computing power node and the second computing power node in the self-organizing neural network model based on the coordinate information includes: determining the neuron in the self-organizing neural network model that is closest to the coordinate vector of the first computing power node or the second computing power node in a candidate circular path, and using it as the winning neuron of the first computing power node or the second computing power node in this candidate circular path, and performing vector update processing on the remaining neurons in the self-organizing neural network model after determining the winning neuron; wherein, the coordinate vector of the first computing power node or the second computing power node is the row vector of the two-dimensional matrix of computing power node coordinates; the candidate circular path starts and ends at the first computing power node and passes through all the second computing power nodes.

[0009] Optionally, generating multiple notification paths based on the winning neuron includes: obtaining the winning neuron of the first computing power node or the second computing power node in multiple candidate circular paths; sorting the first computing power node and the second computing power node in each notification path according to the index order of the winning neurons of the first computing power node and the second computing power node in each candidate circular path; and generating multiple notification paths based on the multiple candidate circular paths.

[0010] Optionally, generating multiple notification paths based on the multiple candidate ring paths includes: sorting and adjusting at least two second computing power nodes in at least one candidate ring path, and / or reversing the sorting of the second computing power nodes in at least one candidate ring path to obtain candidate ring conversion paths; and deduplicating the multiple candidate ring paths and the candidate ring conversion paths to obtain multiple notification paths.

[0011] Optionally, the transmission processing of the first computing power network graph data according to the at least one announcement path includes: forwarding the first computing power data packet according to the announcement path in the header of the first computing power data packet.

[0012] Optionally, the forwarding process of the first computing power data packet according to the announcement path in the header of the first computing power data packet includes: determining the next-hop computing power node according to the announcement path in the header of the first computing power data packet; deleting the self node in the announcement path in the header of the first computing power data packet, generating a new announcement path and replacing the announcement path in the header of the first computing power data packet; and sending the first computing power data packet to the next-hop computing power node.

[0013] Optionally, obtaining the second link quality information between the second computing power node and other second computing power nodes, the first computing power node, and the second computing power information of the second computing power node includes: receiving second computing power network graph data sent by the second computing power node; and extracting the second computing power information and the second link quality information from the second computing power network graph data.

[0014] According to a second aspect of this disclosure, a computing power information notification device is provided, applied to a first computing power node in a computing power network, comprising: a computing power notification module, configured to acquire second link quality information between the second computing power node and other second computing power nodes, and between the first computing power node and second computing power information of the second computing power node; a computing power sensing module, configured to acquire first link quality information between the first computing power node and second computing power nodes in the computing power network, and first computing power information of the first computing power node; and generate first computing power network graph data based on the first link quality information, the first computing power information, the first computing power information, and the second computing power information; the computing power notification module is further configured to generate at least one notification path based on the first link quality information and the second link quality information, and perform transmission processing on the first computing power network graph data according to the at least one notification path.

[0015] Optionally, the first link quality information includes: first latency information, and the second link quality information includes: second latency information; the computing power announcement module includes: a path selection module, used to determine the at least one announcement path based on the first latency information and the second latency information; generate at least one first computing power data packet corresponding to the at least one announcement path based on the first computing power network graph data, and add an announcement path in the header of each first computing power data packet.

[0016] Optionally, the path selection module includes: a coordinate determination unit, configured to construct coordinate information of the first computing power node and the second computing power node based on the first delay information and the second delay information; a coordinate processing unit, configured to construct a self-organizing neural network model and determine the winning neuron of the first computing power node and the second computing power node in the self-organizing neural network model based on the coordinate information; and a path determination unit, configured to generate multiple notification paths based on the winning neuron; wherein the notification path is a circular path with the first computing power node as the starting point and the ending point, and the circular path passes through all the second computing power nodes.

[0017] Optionally, the coordinate information includes: a two-dimensional matrix of computing node coordinates; the coordinate determination unit is specifically used to construct a two-dimensional delay matrix based on the first delay information and the second delay information; wherein, the row vector of the two-dimensional delay matrix is ​​the delay information vector of the first computing node or the second computing node; and the two-dimensional matrix of computing node coordinates is constructed based on the two-dimensional delay matrix.

[0018] Optionally, the number of rows in the two-dimensional matrix of computing power node coordinates is the sum of the number of the first computing power node and the second computing power node; the number of columns in the two-dimensional matrix of computing power node coordinates is the dimension of the neuron vector in the self-organizing neural network model; the coordinate processing unit is used to determine the neuron in the self-organizing neural network model that is closest to the coordinate vector of the first computing power node or the second computing power node in a candidate circular path, and to select the winning neuron of the first computing power node or the second computing power node in this candidate circular path, and to perform vector update processing on the remaining neurons in the self-organizing neural network model after determining the winning neuron; wherein, the coordinate vector of the first computing power node or the second computing power node is the row vector of the two-dimensional matrix of computing power node coordinates; the candidate circular path starts and ends at the first computing power node and passes through all the second computing power nodes.

[0019] Optionally, the path determination unit is used to obtain the winning neurons of the first computing power node or the second computing power node in a plurality of candidate ring paths, and sort the first computing power node and the second computing power node in each announcement path according to the index order of the winning neurons of the first computing power node and the second computing power node in each candidate ring path; and generate a plurality of announcement paths based on the plurality of candidate ring paths.

[0020] Optionally, the path determination unit is used to sort and adjust at least two second computing power nodes in at least one candidate ring path, and / or reverse sort the second computing power nodes in at least one candidate ring path to obtain a candidate ring conversion path; and to deduplicate the multiple candidate ring paths and the candidate ring conversion path to obtain multiple notification paths.

[0021] Optionally, the computing power notification module includes: a message transmission module, used to forward the first computing power data packet according to the notification path in the header of the first computing power data packet.

[0022] Optionally, the message transmission module is specifically configured to determine the next-hop computing power node based on the announcement path in the header of the first computing power data packet; delete its own node in the announcement path in the header of the first computing power data packet, generate a new announcement path and replace the announcement path in the header of the first computing power data packet; and send the first computing power data packet to the next-hop computing power node.

[0023] Optionally, the message transmission module is configured to receive the second computing power network graph data sent by the second computing power node, and extract the second computing power information and the second link quality information from the second computing power network graph data.

[0024] Optionally, the computing power perception module includes: a link quality detection module, used to detect the link quality between the first computing power node and the second computing power node, to obtain the first link quality information; a computing power resource monitoring module, used to obtain the first computing power information; and a resource collection module, used to generate first computing power network graph data based on the first link quality information, the first link quality information, the first computing power information and the second computing power information.

[0025] According to a third aspect of this disclosure, a computing power information notification device is provided, comprising: a memory; and a processor coupled to the memory, the processor being configured to execute the method described above based on instructions stored in the memory.

[0026] According to a fourth aspect of this disclosure, a first computing power node is provided, comprising: a computing power information notification device as described above.

[0027] According to a fifth aspect of this disclosure, a computing power network system is provided, comprising: a second computing power node and a first computing power node as described above.

[0028] According to a sixth aspect of this disclosure, a computer-readable storage medium is provided that stores computer instructions which are executed by a processor using the method described above.

[0029] The computing power information notification method, apparatus, computing power node, computing power network system, and storage medium disclosed herein can reduce the synchronization time of computing power nodes for computing power network graph data and enable all computing power nodes in the network to achieve a state of computing power information synchronization while keeping notification overhead low. It can ensure the availability and scalability of the computing power network system, meet the real-time synchronization requirements of the computing power network for resource information, save network resources, ensure network quality, and improve the user experience. Attached Figure Description

[0030] The above and other objects, features, and advantages of this disclosure will become more apparent from the more detailed description of the embodiments thereof in conjunction with the accompanying drawings. The accompanying drawings are provided to further illustrate the embodiments of this disclosure and form part of the specification. They are used together with the embodiments of this disclosure to explain the disclosure and do not constitute a limitation thereof. The above and other objects and advantages of this disclosure will be further described below with reference to specific embodiments and the accompanying drawings. In the drawings, the same or corresponding technical features or components will be represented by the same or corresponding reference numerals.

[0031] Figure 1 This is a flowchart illustrating an embodiment of the computing power information notification method according to the present disclosure;

[0032] Figure 2 This is a schematic diagram illustrating an application scenario of an embodiment of the computing power information notification method according to the present disclosure;

[0033] Figure 3 This is a schematic diagram of the process of forwarding computing power data packets according to an embodiment of the computing power information notification method of this disclosure;

[0034] Figure 4 This is a flowchart illustrating the generation of an announcement path according to an embodiment of the computing power information announcement method of this disclosure;

[0035] Figure 5 This is a flowchart illustrating the generation of a notification path according to another embodiment of the computing power information notification method of this disclosure;

[0036] Figure 6This is a schematic diagram of the process of forwarding computing power data packets according to another embodiment of the computing power information notification method of this disclosure;

[0037] Figure 7A This diagram illustrates a comparison of the notification overhead of the computing power information notification method disclosed herein with that of greedy routing algorithms and random routing algorithms.

[0038] Figure 7B This diagram illustrates a comparison of the convergence time of the computing power information notification method disclosed herein with that of the greedy routing algorithm and the random routing algorithm.

[0039] Figure 8 This is a schematic diagram of a module according to an embodiment of the computing power information notification device of this disclosure;

[0040] Figure 9 This is a schematic diagram of a path selection module in one embodiment of the computing power information notification device according to the present disclosure;

[0041] Figure 10 This is a schematic diagram of a module according to another embodiment of the computing power information notification device according to the present disclosure. Detailed Implementation

[0042] Exemplary embodiments of the present disclosure will be described below with reference to the accompanying drawings. For clarity and brevity, not all features of the embodiments are described in the specification. However, it should be understood that many implementation-specific settings must be made in carrying out the embodiments to achieve the developer's specific goals, such as complying with constraints related to the device and business, and these constraints may vary depending on the implementation. Furthermore, it should be understood that while development work can be very complex and time-consuming, such development work is merely a routine task for those skilled in the art who benefit from the present disclosure.

[0043] It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of this disclosure.

[0044] Those skilled in the art will understand that the terms "first," "second," etc., in the embodiments of this disclosure are only used to distinguish different steps, devices, or modules, and do not represent any specific technical meaning, nor do they indicate a necessary logical order between them.

[0045] It should also be understood that in the embodiments disclosed herein, "a plurality of" may refer to two or more, and "at least one" may refer to one, two or more.

[0046] It should also be understood that any component, data or structure mentioned in the embodiments of this disclosure can generally be understood as one or more unless expressly defined or given to the contrary in the context.

[0047] Furthermore, the term "and / or" in this disclosure is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this disclosure generally indicates that the preceding and following related objects have an "or" relationship.

[0048] It should also be understood that the description of the various embodiments in this disclosure emphasizes the differences between the various embodiments, and the similarities or similarities can be referred to each other. For the sake of brevity, they will not be described in detail.

[0049] At the same time, it should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn according to actual scale.

[0050] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit this disclosure or its application or use.

[0051] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0052] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0053] Furthermore, to avoid obscuring this disclosure with unnecessary detail, only processing steps and / or apparatus structures closely related to at least the solutions according to this disclosure are shown in the accompanying drawings, while other details less relevant to this disclosure are omitted. It should also be noted that similar reference numerals and letters in the drawings indicate similar items, and therefore once an item is defined in one drawing, it need not be discussed again in subsequent drawings.

[0054] Among the related technologies known to the inventors, with the rapid development of hardware devices such as smart terminals, computing resources are gradually showing a trend of distributed and ubiquitous deployment. Furthermore, people's demand for computing power-intensive and latency-sensitive businesses such as autonomous driving, telemedicine, and the industrial internet is constantly increasing. Therefore, how to efficiently and collaboratively utilize resources and reduce usage costs is a current research hotspot.

[0055] To address the aforementioned key issues, the concept of a computing power network is proposed. A computing power network is a novel network architecture that connects computing devices through a network, coordinating distributed computing and storage resources to achieve joint configuration of computing, storage, and network resources. This reduces resource usage costs and improves user experience. A computing power network system requires the rational allocation of resources and the awareness and sharing of resources.

[0056] In a computing power network, computing nodes providing computing power first need to be self-aware, obtaining information about their own computing resources, network resources, storage resources, and other resources. Then, each computing node announces this resource information to other computing nodes, achieving collaborative awareness of network-wide resources. When a user's task arrives at a computing node, the computing node makes a computation offloading decision based on the network-wide resource information, achieving collaborative resource allocation.

[0057] Currently, the computing power announcement technology used in computing power networks employs existing routing protocol messages to carry resource information, synchronizing this information by transmitting these messages between computing power nodes during route convergence. However, synchronizing resource information in computing power networks using existing routing protocol messages presents several problems. For instance, the route convergence time of existing protocols (on the order of minutes) is relatively long, failing to meet the real-time synchronization requirements of computing power networks. Furthermore, increasing the sending frequency of routing protocol messages increases network overhead and may lead to issues such as route oscillations, impacting network quality.

[0058] To address the aforementioned issues, this disclosure provides a technical solution for disseminating computing power information.

[0059] Figure 1 This is a flowchart illustrating an embodiment of the computing power information notification method according to the present disclosure. This computing power information notification method is applied to a first computing power node, such as... Figure 1 As shown:

[0060] Step 101: Obtain the first link quality information between the first computing power node and the second computing power node in the computing power network, and the first computing power information of the first computing power node.

[0061] In one embodiment, the computing power network includes multiple computing power nodes with strong computing power and superior link quality. These computing power nodes include a first computing power node and a second computing power node. This disclosure describes a computing power information notification method applied to the first computing power node. The computing power information notification method of this disclosure can also be applied to the second computing power node; the computing power information notification methods applied to the first and second computing power nodes are substantially the same.

[0062] The first computing node can use various existing link quality testing tools to detect the first link quality information between itself and all second computing nodes. These tools include existing tools such as Prometheus and Pinpoint. The first link quality information includes round-trip latency, link bandwidth, jitter, and other information between the first and second computing nodes. The first computing node can obtain its own first computing power information by calling operating system instructions. This first computing power information includes CPU frequency, CPU utilization, cache capacity, cache utilization, and other information.

[0063] Step 102: Obtain the second link quality information between the second computing power node and other second computing power nodes, the first computing power node, and the second computing power information of the second computing power node.

[0064] In one embodiment, various methods can be used to obtain second link quality information, second computing power information, etc. The second computing power node can use various existing link quality detection tools to obtain second link quality information, including round-trip time, link bandwidth, jitter, etc., between the second computing power node and all other second computing power nodes, as well as round-trip time, link bandwidth, jitter, etc., between the second computing power node and the first computing power node. The second computing power node can obtain its own second computing power information by calling operating system instructions. This second computing power information includes CPU frequency, CPU utilization, cache capacity, cache utilization, etc.

[0065] Step 103: Generate first computing power network graph data based on first link quality information, first computing power information, and second computing power information.

[0066] In one embodiment, the first computing power network graph includes information such as computing power resource information and network resource topology graph of the computing power network; for example, the first computing power network graph data may include first link quality information, first computing power information and second computing power information, as well as network resource topology graph information of the computing power network.

[0067] Step 104: Generate at least one notification path based on the first link quality information and the second link quality information. The number of notification paths can be one or more.

[0068] Step 105: Transmit the first computing power network graph data according to at least one notification path.

[0069] In one embodiment, the first computing power node updates the first computing power network graph data based on the first link quality information, the first computing power information, and the second computing power information. The first computing power node determines one or more announcement paths, places the announcement path in the header of a first computing power data packet carrying the latest first computing power network graph data, and forwards the first computing power data packet according to the announcement path in the header. The first computing power node forwards the first computing power network graph data to all second computing power nodes in the computing power network according to the announcement path and returns it to the first computing power node.

[0070] In one embodiment, such as Figure 2 As shown, the computing power network system includes a first computing power node 21 and second computing power nodes 22, 23, 24, etc., all of which utilize the computing power information notification method disclosed herein. The computing power nodes include edge server nodes and cloud nodes, etc., and are connected to the computing power network system through a computing power network router.

[0071] Compared to ordinary routers, computing network routers add the functions of billing authentication and computation offloading decision-making. The computing network router obtains the computing network graph of its neighboring computing nodes. When a computing task initiated by a terminal device arrives, the computing network router performs task scheduling based on the computing network graph data, determining the computing node to handle the task or forwarding the task to other idle computing nodes.

[0072] Figure 3 This is a schematic diagram illustrating the process of forwarding computing power data packets according to an embodiment of the computing power information notification method of this disclosure. The first link quality information includes first latency information and other information, and the second link quality information includes second latency information and other information, such as... Figure 3 As shown,

[0073] Step 301: Determine at least one notification path based on the first delay information and the second delay information.

[0074] Step 302: Generate at least one first computing power data packet corresponding to at least one announcement path based on the first computing power network graph data, and add an announcement path to the header of each first computing power data packet.

[0075] Step 303: Forward the first computing power data packet according to the announcement path in the header of the first computing power data packet.

[0076] At least one notification path can be determined based on the first delay information and the second delay information using various methods. Figure 4 This is a flowchart illustrating the generation of an announcement path according to an embodiment of the computing power information announcement method of this disclosure, as shown below. Figure 4 As shown:

[0077] Step 401: Construct the coordinate information of the first computing node and the second computing node based on the first delay information and the second delay information.

[0078] Step 402: Construct a self-organizing neural network model and determine the winning neurons of the first and second computing power nodes in the self-organizing neural network model based on coordinate information.

[0079] In one embodiment, a self-organizing neural network (SNN) is a type of neural network model that employs a teacherless learning approach. It does not require a desired output; instead, it learns from data samples and adjusts its weights to achieve the training objective. The SNN model can be an existing SNN model, or a pre-trained SNN model can be used, with pre-defined neuron vectors. The winning neurons of the first and second computing nodes are then obtained using the SNN model.

[0080] Step 403: Generate multiple notification paths based on the winning neuron; wherein, the notification path is a circular path with the first computing power node as the starting point and the ending point, and the circular path passes through all the second computing power nodes.

[0081] Figure 5 This is a flowchart illustrating the generation of a notification path according to another embodiment of the computing power information notification method of this disclosure. The coordinate information includes a two-dimensional matrix of computing power node coordinates, such as... Figure 5 As shown:

[0082] Step 501: Construct a two-dimensional delay matrix based on the first delay information and the second delay information; wherein, the row vectors of the two-dimensional delay matrix are the delay information vectors of the first computing node or the second computing node.

[0083] Step 502: Construct a two-dimensional matrix of computing node coordinates based on the two-dimensional delay matrix.

[0084] In one embodiment, the number of rows in the two-dimensional matrix of computing node coordinates is the sum of the number of the first computing node and the second computing node, and the number of columns in the two-dimensional matrix of computing node coordinates is the dimension of the neuron vector in the self-organizing neural network model.

[0085] Step 503: In the self-organizing neural network model, determine the neuron whose coordinate vector distance is closest to the first or second computing power node in a candidate circular path, and use it as the winning neuron of the first or second computing power node in this candidate circular path. After determining the winning neuron, perform vector update processing on the remaining neurons in the self-organizing neural network model.

[0086] In one embodiment, the computing power network system has N computing power nodes (including a first computing power node and a second computing power node). The first computing power node constructs a two-dimensional delay matrix D based on the first delay information and the second delay information, which can be represented as:

[0087]

[0088] In equation (1), d i,j Let represent the round-trip delay (first delay information or second delay information) between computing power node i and node j, where node 1 can be the first computing power node and the remaining nodes are the second computing power nodes; Equation (1) satisfies:

[0089]

[0090] The row vectors of the two-dimensional delay matrix are the delay information vectors of the first or second computing node. For example, the delay information vector of the first computing node (node ​​1) is D1 = (d 1,1 d 1,2 , ...d 1,N The latency information vector of a second computing node (node ​​2) is D2 = (d 2,1 d 2,2 , ...d 2,N The latency information vector of a second computing node (node ​​N) is D. N =(d N,1 d N,2 , ...d N,N ) etc. Among them, d N,1 With d 1,N They can be the same or different.

[0091] The two-dimensional matrix P of the computing node coordinates is:

[0092]

[0093] In this matrix, the number of rows in the two-dimensional matrix P of computing node coordinates is the sum of the number of the first and second computing nodes, N, and the number of columns in the two-dimensional matrix P of computing node coordinates is the dimension dim of the neuron vector in the self-organizing neural network model.

[0094] Elements p in the two-dimensional matrix P of computing node coordinates i,j The Euclidean distance between any two nodes i and j can be equal, and can be calculated using formula (4):

[0095]

[0096] Here, ED() represents the Euclidean distance function, and D can be determined based on the row vectors of the two-dimensional time-delay matrix. i =(d i,1 di,2 , ...d i,dim ),D j =(d j,1 d j,2 , ...d j,dim ), i = 1, 2, ..., N, j = 1, 2, ..., dim, dim <= N, dim and N are constants; in the two-dimensional matrix P of computing node coordinates, p 1,1 p 2,2 The value is 0.

[0097] Elements p in the two-dimensional matrix P of computing node coordinates i,j It can also be set as an element d in the two-dimensional delay matrix D. i,j .

[0098] The proposed circular path starts and ends at the first computing power node and passes through all the second computing power nodes. The coordinate vector of the first or second computing power node is the row vector of the two-dimensional matrix P of computing power node coordinates. The i-th row vector of the two-dimensional matrix P of computing power node coordinates is:

[0099] P i =[p i,1 p i,2 …p i,dim ]……(5);

[0100] Among them, P i The dim-dimensional spatial coordinate vector representing computing node i (the first computing node or the second computing node).

[0101] Construct a (ring-shaped) self-organizing neural network with the number of neurons being 8 or 9 times the number of computing nodes (the total number of the first and second computing nodes). The dimension of each neuron vector is the same as the dimension (dim) of the coordinate vector of the computing node, and the neuron vectors are pre-configured.

[0102] In a self-organizing neural network model, the neuron whose coordinate vector is closest to the first or second computing power node in a candidate circular path is identified as the winning neuron in that path. The coordinate vector of the first or second computing power node is a row vector of a two-dimensional matrix of computing power node coordinates.

[0103] For example, the two-dimensional matrix P of the computing node coordinates generated by the first computing node is input into the self-organizing neural network, and the winning neuron m of the first or second computing node in a candidate circular path is calculated by formula (6):

[0104]

[0105] Where ED() represents the Euclidean distance function, Wi p is the vector of the i-th neuron in a self-organizing neural network. n Let p be the coordinate vector of the first or second computing node, and let m be the coordinate vector of the neuron in a candidate circular path. n The nearest neuron.

[0106] After determining the winning neuron, the remaining neurons in the self-organizing neural network model are updated using formulas (7) and (8):

[0107]

[0108] W j =W j +T j,m ·(p n -W j )……(8);

[0109] Among them, S j,m Let δ be the distance between neuron j and the winning neuron m, and p be a parameter. n The coordinate vector of the first or second computing node mentioned above.

[0110] Step 504: Obtain the winning neuron of the first or second computing power node among multiple candidate circular paths.

[0111] In one embodiment, step 503 is performed until the number of candidate ring paths reaches 0.3, 0.5, or other times the total number of computing power nodes N.

[0112] Step 505: Sort the first and second computing power nodes in each notification path according to the index order of the winning neurons of the first and second computing power nodes in each candidate ring path.

[0113] In one embodiment, the first computing node obtains the winning neurons of the first computing node or the second computing node in multiple candidate ring paths, and sorts the first computing node and the second computing node in each announcement path according to the index order of the winning neurons of the first computing node and the second computing node in each candidate ring path.

[0114] For example, in a self-organizing neural network model, all neurons form a ring. Each neuron is indexed, and the neurons are arranged in descending order of their indices. In a candidate ring path, the winning neuron of the first computing power node has an index of 1, and the winning neurons of the second computing power nodes A, B, and C have indices of 4, 5, and 2 respectively. After sorting the first and second computing power nodes A, B, and C in this notification path according to the index order of the winning neurons (based on ascending order of indices), the candidate ring path becomes: First computing power node - Second computing power node C - Second computing power node A - Second computing power node B - First computing power node. This candidate ring path starts and ends at the first computing power node and passes through all the second computing power nodes.

[0115] Step 506: Generate multiple notification paths based on multiple candidate ring paths.

[0116] In one embodiment, multiple methods can be used to generate multiple announcement paths based on multiple candidate ring paths. This involves sorting and adjusting at least two second computing power nodes in at least one candidate ring path, and / or reversing the sorting of the second computing power nodes in at least one candidate ring path to obtain candidate ring conversion paths; and then deduplicating the multiple candidate ring paths and candidate ring conversion paths to obtain multiple announcement paths.

[0117] For example, the two second computing nodes in the candidate ring path "First computing node - Second computing node C - Second computing node A - Second computing node B - First computing node" can be sorted and adjusted to obtain candidate ring transformation paths such as "First computing node - Second computing node A - Second computing node C - Second computing node B - First computing node" and "First computing node - Second computing node C - Second computing node B - Second computing node A - First computing node". Alternatively, the candidate ring path can be cyclically shifted several times to obtain the candidate ring transformation path.

[0118] The candidate ring path "First computing power node - Second computing power node C - Second computing power node A - Second computing power node B - First computing power node" can be reverse-sorted to obtain candidate ring transformation paths such as "First computing power node - Second computing power node B - Second computing power node A - Second computing power node C - First computing power node". Multiple candidate ring paths and candidate ring transformation paths are then deduplicated to remove duplicate paths, resulting in multiple announcement paths.

[0119] Figure 6 This is a schematic diagram of the process of forwarding computing power data packets according to another embodiment of the computing power information notification method of this disclosure;

[0120] Step 601: Determine the next-hop computing node based on the announcement path in the header of the first computing power data packet.

[0121] Step 602: Delete the self-node in the announcement path in the header of the first computing power data packet, generate a new announcement path and replace the announcement path in the header of the first computing power data packet.

[0122] Step 603: Send the first computing power data packet to the next-hop computing power node.

[0123] The first computing node receives the second computing network graph data sent by the second computing node, and extracts the second computing information and the second link quality information from the second computing network graph data.

[0124] In one embodiment, the first computing power node generates multiple first computing power data packets corresponding to multiple announcement paths based on the first computing power network graph data, and adds an announcement path to the header of each first computing power data packet.

[0125] The first and second computing power nodes determine the next-hop computing power node based on the announcement path in the header of the first computing power data packet, delete their own nodes in the announcement path in the header of the first computing power data packet, generate a new announcement path, and replace the announcement path in the header of the first computing power data packet.

[0126] The first computing power node and the second computing power node can send the first computing power data packet to the next-hop computing power node based on routing information through the computing power network router. The computing power network router can use existing methods to send the first computing power data packet to the next-hop computing power node based on routing information.

[0127] The method by which the second computing power node generates and forwards the second computing power network graph data is the same as the method by which the first computing power node generates and forwards the second computing power network graph data. The first computing power node receives the second computing power network graph data sent by the second computing power node and extracts the second computing power information and the second link quality information from the second computing power network graph data.

[0128] In one embodiment, Figure 7A and 7B The figures show a comparison of simulation results between the proposed computing power information notification method and existing greedy routing algorithms and random routing algorithms regarding notification cost and convergence time. The proposed method is a multipath routing algorithm. The greedy routing algorithm always selects the path with the shortest total latency when generating notification paths, while the random routing algorithm randomly selects paths. Notification cost refers to the total number of messages sent to achieve synchronization of the computing power network graph data across all computing power nodes. Convergence time refers to the time required for the computing power network graph data across all computing power nodes to achieve synchronization.

[0129] like Figure 7A and 7BAs shown, the computing power information notification method disclosed herein can reduce the synchronization time of computing power nodes for computing power network graph data while keeping notification overhead low. Therefore, adopting the computing power information notification method disclosed herein can better support the upper-layer technologies of computing power networks while saving network resources.

[0130] The computing power information notification method disclosed herein ensures high availability and scalability of the computing power network system by using distributed computing power notification technology; it can adaptively adjust the number of notification paths according to the number of devices, reduce the synchronization time of computing power nodes for computing power network graph data with low notification overhead; it can enable all computing power nodes in the network to reach a computing power information synchronization state in a short time with low notification overhead; and it can meet the real-time synchronization requirements of computing power networks for resource information.

[0131] In one embodiment, such as Figure 8 As shown, this disclosure provides a computing power information notification device applied to a first computing power node in a computing power network, including a computing power sensing module 81 and a computing power notification module 86. The computing power information notification device of this disclosure can also be applied to a second computing power node in a computing power network. The computing power notification module 86 acquires second link quality information between the second computing power node and other second computing power nodes, the first computing power node, and second computing power information of the second computing power node.

[0132] The computing power perception module 81 acquires the first link quality information between the first computing power node and the second computing power node in the computing power network, and the first computing power information of the first computing power node; the computing power perception module 81 generates first computing power network graph data based on the first link quality information, the first computing power information, and the second computing power information; the computing power notification module 81 generates at least one notification path based on the first link quality information and the second link quality information, and performs transmission processing on the first computing power network graph data according to the at least one notification path.

[0133] In one embodiment, the computing power notification module 86 includes a path selection module 87 and a message transmission module 88. The path selection module 87 determines at least one notification path based on first delay information and second delay information; the path selection module 87 generates at least one first computing power data packet corresponding to the at least one notification path based on the first computing power network graph data, and adds a notification path to the header of each first computing power data packet. The message transmission module 88 forwards the first computing power data packets according to the notification path in the header of the first computing power data packets.

[0134] In one embodiment, such as Figure 9As shown, the path selection module 87 includes a coordinate determination unit 871, a coordinate processing unit 872, and a path determination unit 873. The coordinate determination unit 871 constructs coordinate information for the first and second computing power nodes based on the first and second time delay information. The coordinate processing unit 872 constructs a self-organizing neural network model and determines the winning neurons for the first and second computing power nodes based on the coordinate information within the self-organizing neural network model. The path determination unit 873 generates multiple notification paths based on the winning neurons. Each notification path is a circular path starting and ending at the first computing power node, and the circular path passes through all the second computing power nodes.

[0135] In one embodiment, the coordinate information includes a two-dimensional matrix of computing node coordinates; the coordinate determination unit 871 constructs a two-dimensional delay matrix based on the first delay information and the second delay information, wherein the row vectors of the delay matrix are the delay information vectors of the first computing node or the second computing node. The coordinate determination unit 871 constructs a two-dimensional matrix of computing node coordinates based on the delay matrix, wherein the number of rows in the two-dimensional matrix of computing node coordinates is the sum of the number of the first computing node and the second computing node, and the number of columns in the two-dimensional matrix of computing node coordinates is the dimension of the neurons in the self-organizing neural network model.

[0136] The coordinate processing unit 872 determines the neuron in the self-organizing neural network model that has the closest coordinate vector distance to the first or second computing power node in a candidate circular path. This neuron is then designated as the winning neuron for that candidate circular path. After determining the winning neuron, the unit updates the vectors of the remaining neurons in the self-organizing neural network model. The coordinate vector of the first or second computing power node is a row vector of the two-dimensional coordinate matrix of the computing power nodes. The candidate circular path starts and ends at the first computing power node and passes through all the second computing power nodes.

[0137] The path determination unit 873 obtains the winning neurons of the first computing power node or the second computing power node in multiple candidate ring paths, and sorts the first computing power node and the second computing power node in each announcement path according to the index order of the winning neurons of the first computing power node and the second computing power node in each candidate ring path; the path determination unit 873 generates multiple announcement paths based on the multiple candidate ring paths.

[0138] For example, the path determination unit 873 sorts and adjusts at least two second computing power nodes in at least one candidate ring path, and / or reverses the sorting of the second computing power nodes in at least one candidate ring path to obtain a candidate ring conversion path; the path determination unit 873 performs deduplication processing on multiple candidate ring paths and candidate ring conversion paths to obtain multiple announcement paths.

[0139] In one embodiment, the path selection module 87 generates a first computing power data packet and sends it to the message transmission module 88. The message transmission module 88 determines the next-hop computing power node based on the advertised path in the header of the first computing power data packet; the message transmission module 88 deletes its own node from the advertised path in the header of the first computing power data packet, generates a new advertised path, and replaces the advertised path in the header of the first computing power data packet; the message transmission module 88 sends the first computing power data packet to the next-hop computing power node. The message transmission module 88 receives second computing power network graph data sent by the second computing power node and extracts second computing power information and second link quality information from the second computing power network graph data.

[0140] In one embodiment, the computing power perception module 81 includes a resource collection module 82, a link quality detection module 83, and a computing power resource monitoring module 84. The link quality detection module 83 detects the link quality between the first computing power node and the second computing power node to obtain first link quality information; the computing power resource monitoring module 84 obtains first computing power information; and the resource collection module 82 generates first computing power network graph data based on the first link quality information, the first computing power information, and the second computing power information. The resource collection module 82 can aggregate the computing power resources and network resource information of its own computing power node and other computing power nodes, and provide this information externally.

[0141] The link quality detection module 83 can periodically detect the link quality information between this computing power node and other computing power nodes by deploying link quality detection tools such as Prometheus and Pinpoint, including indicators such as round-trip latency between computing power nodes, link bandwidth, and jitter, and send the detected data to the resource collection module 82 using the publisher (pub) mode of the open-source messaging library ZeroMQ (ZMQ).

[0142] The computing power resource monitoring module 84 periodically obtains computing power information of its local computing power node by calling terminal instructions of the operating system, including CPU frequency, CPU utilization, cache capacity, cache utilization, etc., and sends the detected data to the resource collection module 82 using the ZMQ protocol in pub mode. The resource collection module 82 is set to subscribe to the computing power information and inter-link quality information sent from the computing power resource monitoring module 84, the link quality detection module 83, and the message transmission module 88, in order to update the local computing power network graph data.

[0143] Resource collection module 82 opens an HTTP service port to transmit the computing power network graph of this node to path selection module 87 and computing offloading module 89. Message transmission module 88 extracts the latest computing power information and network information of all computing power nodes in the network from the received computing power network graph data packet and sends it to resource collection module 82 via the ZMQ protocol's pub mode.

[0144] In one embodiment, such as Figure 10 As shown, the computing power information notification device may include a memory 1001, a processor 1002, a communication interface 1003, and a bus 1004. The memory 1001 is used to store instructions, and the processor 1002 is coupled to the memory 1001. The processor 1002 is configured to execute the above-described computing power information notification method based on the instructions stored in the memory 1001.

[0145] The memory 1001 can be high-speed RAM, non-volatile memory, or a memory array. The memory 1001 may also be divided into blocks, and these blocks can be combined into virtual volumes according to certain rules. The processor 1002 can be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement a computing power information notification method.

[0146] In one embodiment, this disclosure provides a first computing node, including the computing power information notification device as described in any of the above embodiments.

[0147] In one embodiment, this disclosure provides a computing power network system, including a second computing power node and a first computing power node as described in any of the above embodiments; wherein, the computing power information notification device of this disclosure can be deployed on both the second computing power node and the first computing power node.

[0148] In one embodiment, this disclosure provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the method as described in any of the above embodiments.

[0149] Computer-readable storage media may take the form of any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples (not an exhaustive list) of readable storage media may include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0150] Embodiments of this disclosure may also be computer program products comprising computer program instructions that, when executed by a processor, cause the processor to perform the steps of the methods according to various embodiments of this disclosure as described in the "Exemplary Methods" section above.

[0151] The computing power information notification method, apparatus, computing power nodes, computing power network system, and storage medium in the above embodiments, by using distributed computing power notification technology, ensure the high availability and scalability of the computing power network system; can adaptively adjust the number of notification paths according to the number of devices; can reduce the synchronization time of computing power nodes for computing power network graph data with low notification overhead; can enable all computing power nodes in the network to reach a computing power information synchronization state in a short time with low notification overhead; can meet the real-time synchronization requirements of computing power networks for resource information; can save network resources and ensure network quality; can guarantee service quality and improve user experience.

[0152] The basic principles of this disclosure have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this disclosure are merely examples and not limitations, and should not be considered as essential features of each embodiment of this disclosure. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the scope of this disclosure to the necessity of employing the aforementioned specific details for implementation.

[0153] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For system embodiments, since they largely correspond to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0154] The block diagrams of devices, apparatuses, devices, and systems disclosed herein are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0155] It should also be noted that in the apparatus, devices, and methods of this disclosure, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions to this disclosure.

[0156] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this disclosure. Therefore, this disclosure is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0157] The above description has been given for illustrative and descriptive purposes. Furthermore, this description is not intended to limit the embodiments of this disclosure to the forms disclosed herein. Although several exemplary aspects and embodiments have been discussed above, those skilled in the art will understand that the above embodiments are illustrative only and do not limit the scope of this disclosure. Those skilled in the art will understand that the above embodiments can be combined, modified, or replaced without departing from the scope and spirit of this disclosure.

Claims

1. A method for announcing computing power information, applied to the first computing power node of a computing power network, comprising: Obtain the first link quality information between the first computing power node and the second computing power node in the computing power network, and the first computing power information of the first computing power node; Obtain the second link quality information between the second computing power node and other second computing power nodes, the first computing power node, and the second computing power information of the second computing power node; First computing power network graph data is generated based on the first link quality information, the second link quality information, the first computing power information, and the second computing power information; The first link quality information includes first delay information, and the second link quality information includes second delay information; Generate at least one notification path, including: The coordinate information of the first computing power node and the second computing power node is constructed based on the first latency information and the second latency information; the coordinate information includes a two-dimensional matrix of computing power node coordinates; the number of rows of the two-dimensional matrix of computing power node coordinates is the sum of the number of the first computing power node and the second computing power node, and the number of columns is the dimension of the neuron vector in the constructed self-organizing neural network model; the coordinate vector of the first computing power node or the second computing power node is the row vector of the two-dimensional matrix of computing power node coordinates; In the self-organizing neural network model, the neuron whose coordinate vector distance is closest to the first computing power node or the second computing power node in a candidate circular path is determined as the winning neuron, and vector update processing is performed on the remaining neurons; the candidate circular path starts and ends at the first computing power node and passes through all the second computing power nodes; multiple announcement paths are generated based on the winning neuron, and the announcement paths are circular paths that start and end at the first computing power node and pass through all the second computing power nodes; Based on the first computing power network graph data, at least one first computing power data packet corresponding to the at least one announcement path is generated, and an announcement path is added to the header of each first computing power data packet. The first computing power network graph data is transmitted and processed according to the at least one notification path.

2. The method as described in claim 1, wherein, The step of constructing the coordinate information of the first computing node and the second computing node based on the first delay information and the second delay information includes: Construct a two-dimensional delay matrix based on the first delay information and the second delay information; Wherein, the row vector of the delay two-dimensional matrix is ​​the delay information vector of the first computing power node or the second computing power node; The two-dimensional matrix of computing node coordinates is constructed based on the aforementioned two-dimensional delay matrix.

3. The method of claim 1, wherein generating multiple notification paths based on the winning neuron comprises: Obtain the winning neuron of the first computing power node or the second computing power node among multiple candidate circular paths; The first computing power node and the second computing power node in each announcement path are sorted according to the index order of the winning neurons of the first computing power node and the second computing power node in each candidate ring path. Multiple notification paths are generated based on the multiple candidate ring paths.

4. The method as described in claim 3, wherein generating multiple notification paths based on the multiple candidate ring paths comprises: The second computing power nodes in at least one of the candidate ring paths are sorted and adjusted, and / or the second computing power nodes in at least one of the candidate ring paths are sorted in reverse order to obtain the candidate ring conversion path. The multiple candidate ring paths and the candidate ring transformation paths are deduplicated to obtain multiple notification paths.

5. The method as described in claim 1, wherein the transmission processing of the first computing power network graph data according to the at least one notification path includes: The first computing power data packet is forwarded according to the announcement path in the header of the first computing power data packet.

6. The method as described in claim 5, wherein forwarding the first computing power data packet according to the announcement path in the header of the first computing power data packet includes: The next-hop computing node is determined based on the announcement path in the header of the first computing power data packet. Delete the self node in the announcement path in the header of the first computing power data packet, generate a new announcement path and replace the announcement path in the header of the first computing power data packet; The first computing power data packet is sent to the next-hop computing power node.

7. The method as described in claim 6, wherein obtaining the second link quality information between the second computing power node and other second computing power nodes, the first computing power node, and the second computing power information of the second computing power node includes: Receive the second computing power network graph data sent by the second computing power node; Extract the second computing power information and the second link quality information from the second computing power network graph data.

8. A computing power information notification device, applied to a first computing power node in a computing power network, comprising: The computing power notification module is used to obtain the second link quality information between the second computing power node and other second computing power nodes, the first computing power node, and the second computing power information of the second computing power node. The computing power awareness module is used to obtain the first link quality information between the first computing power node and the second computing power node in the computing power network, and the first computing power information of the first computing power node. First computing power network graph data is generated based on the first link quality information, the second link quality information, the first computing power information, and the second computing power information; The computing power notification module is also used to generate at least one notification path and to perform transmission processing on the first computing power network graph data according to the at least one notification path; The first link quality information includes first latency information, and the second link quality information includes second latency information; the computing power notification module includes a path selection module, which includes: A coordinate determination unit is used to construct coordinate information of the first computing power node and the second computing power node based on the first delay information and the second delay information; the coordinate information includes a two-dimensional matrix of computing power node coordinates; the number of rows of the two-dimensional matrix of computing power node coordinates is the sum of the number of the first computing power node and the second computing power node, and the number of columns is the dimension of the neuron vector in the constructed self-organizing neural network model; the coordinate vector of the first computing power node or the second computing power node is the row vector of the two-dimensional matrix of computing power node coordinates; A coordinate processing unit is used in the self-organizing neural network model to determine the neuron whose coordinate vector distance is closest to the first computing power node or the second computing power node in a candidate circular path, and to select the winning neuron, and to perform vector update processing on the remaining neurons; the candidate circular path starts and ends at the first computing power node and passes through all the second computing power nodes. A path determination unit is used to generate multiple notification paths based on the winning neuron; the notification path is a circular path with the first computing power node as the starting point and the ending point, and the circular path passes through all the second computing power nodes; The path selection module is further configured to generate at least one first computing power data packet corresponding to the at least one announcement path based on the first computing power network graph data, and to add an announcement path in the header of each first computing power data packet.

9. The apparatus of claim 8, wherein, The coordinate determination unit is specifically used to construct a two-dimensional delay matrix based on the first delay information and the second delay information; wherein, the row vector of the two-dimensional delay matrix is ​​the delay information vector of the first computing node or the second computing node; and to construct the two-dimensional coordinate matrix of the computing node based on the two-dimensional delay matrix.

10. The apparatus of claim 9, wherein, The path determination unit is used to obtain the winning neurons of the first computing power node or the second computing power node in multiple candidate circular paths, and sort the first computing power node and the second computing power node in each announcement path according to the index order of the winning neurons of the first computing power node and the second computing power node in each candidate circular path. Multiple notification paths are generated based on the multiple candidate ring paths.

11. The apparatus of claim 10, wherein, The path determination unit is used to sort and adjust at least two second computing power nodes in at least one candidate ring path, and / or reverse sort the second computing power nodes in at least one candidate ring path to obtain a candidate ring conversion path; and to deduplicate the multiple candidate ring paths and the candidate ring conversion path to obtain multiple notification paths.

12. The apparatus of claim 8, wherein, The computing power notification module includes: The message transmission module is used to forward the first computing power data packet according to the announcement path in the header of the first computing power data packet.

13. The apparatus of claim 12, wherein, The message transmission module is specifically used to determine the next-hop computing power node based on the announcement path in the header of the first computing power data packet; delete its own node in the announcement path in the header of the first computing power data packet, generate a new announcement path and replace the announcement path in the header of the first computing power data packet; and send the first computing power data packet to the next-hop computing power node.

14. The apparatus of claim 13, wherein, The message transmission module is used to receive the second computing power network graph data sent by the second computing power node, and extract the second computing power information and the second link quality information from the second computing power network graph data.

15. The apparatus of claim 8, wherein, The computing power sensing module includes: The link quality detection module is used to detect the link quality between the first computing power node and the second computing power node, in order to obtain the first link quality information. The computing power resource monitoring module is used to obtain the first computing power information; The resource collection module is used to generate first computing power network graph data based on the first link quality information, the first computing power information, the first computing power information, and the second computing power information.

16. A computing power information notification device, comprising: Memory; And a processor coupled to the memory, the processor being configured to perform the method as described in any one of claims 1 to 8 based on instructions stored in the memory.

17. A first computing node, comprising: The computing power information notification device as described in any one of claims 8 to 16.

18. A computing power network system, comprising: The second computing node and the first computing node as described in claim 17.

19. A computer-readable storage medium storing computer instructions that are executed by a processor according to any one of claims 1 to 7.

Images