A computing power identification method for computing power internet

Abstract

The application discloses a kind of computing power internet-oriented computing power identification method, it is related to computing power internet technical field, including steps: S1, computing power resource information collection;S2, computing power identification coding;S3, computing power identification registration and report;S4, computing power identification analysis and statistics;S5, based on the scheduling of computing power identification.The computing power internet-oriented computing power identification method, by clear step and coding principle, carries out unified identification to computing power resource, solves the problem that there is no unified standard and specification in the prior art for the description, positioning of computing power resource, helps to realize the description, positioning and addressing of computing power resource, meets the deep development needs of computing power internet.And through the registration and report, analysis and statistics of computing power identification, provide unified resource description for the access, perception, scheduling, transaction settlement and other links of computing power resource, to realize the efficient management and scheduling of computing power resource, improve the utilization efficiency of computing power resource.

Description

Technical Field

[0001] This invention relates to the field of computing power internet technology, specifically to a computing power identification method for computing power internet. Background Technology

[0002] The Internet of Computing Power is built upon the Internet architecture, using a unified computing power identifier, a computing network cloud scheduling operating system, and high-performance transmission protocols to enhance computing and network capabilities and form a resilient network. It aims to achieve standardized interconnection of public computing power resources from different entities, types, and regions, providing queryable, interactive, and callable service capabilities, thereby enabling efficient supply and demand matching, flow, and migration of applications and data among computing power resources.

[0003] In the computing power internet, the role of computing power identifiers is crucial. It is a string that encodes the type, location, architecture, and other characteristics of computing power resources, possessing uniqueness, applicability, compatibility, scalability, and scientific validity. Computing power identifiers provide a unified resource description for all stages of computing power interconnection, including access, sensing, scheduling, and transaction settlement, and are key to achieving efficient management and scheduling of computing power resources.

[0004] However, in practical applications, while there have been some experimental explorations in the industry regarding the description and location of computing resources, a standardized approach has yet to be established that can fully realize the description, location, and addressing of computing resources, making it difficult to meet the needs of the in-depth development of the computing power internet. In particular, the coding rules of some computing power identification methods are not yet perfect, failing to comprehensively and accurately describe the various characteristics of computing resources. This will further lead to errors in resource matching and scheduling, affecting the utilization efficiency of computing resources. Summary of the Invention

[0005] The purpose of this invention is to provide a computing power identification method for the computing power Internet, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a computing power identification method for the computing power internet, comprising the following steps: S1. Computing Resource Information Collection: The computing resource provider collects comprehensive information on its computing resources, including enterprise information, industry affiliation, city and availability zone location, resource type, service type, computing volume, storage volume, network bandwidth, chip model and quantity, server type, CPU model, number of CPU cores, memory, storage, number of servers, computing power internet address, etc., to support subsequent computing power identification encoding and reporting. S2. Computing power identification coding: Based on the coding principles of computing power identification (uniqueness, compatibility, applicability, scalability, and scientific nature), the collected computing power resource information is encoded into computing power identification. S3. Computing power identification registration and reporting: Computing power providers register a unique registration code for computing power identification through the computing power market platform and obtain the qualification to report computing power resources. Then, using the computing power identification prefix and suffix, they report detailed computing power resource information to the computing power interconnection platform. S4. Computing Power Identifier Resolution and Statistics: After receiving the computing power identifier, the computing power interconnection platform verifies, resolves and statistically analyzes it to form a computing power identifier registry. Based on the resolution results, it statistically analyzes key information such as the number of connected computing power providers, resource type, service type, computing load, storage load, network type and bandwidth, computing power internet address, chip model and quantity. S5. Scheduling based on computing power identifier: Users select computing power resources through the computing power scheduling service platform, and activate and use appropriate computing power resources based on the resolution result of the computing power identifier and their own needs.

[0007] Furthermore, in step S1, the computing power provider obtains dynamic information on computing power resources in real time or periodically through automated data collection tools such as computing power identification gateways or management platforms such as cloud management platforms, and organizes it into a standardized data format.

[0008] Furthermore, in step S1, the specific content collected is as follows: Company Information: The company code corresponding to the company name is used to uniquely identify the computing power provider; Industry Classification: The industry category to which the computing power provider belongs, such as telecommunications, computer, finance, etc. City and Availability Zone Location: The city where the computing resources are located and the specific availability zone (data center area); Resource type: The type of computing power resource, which can be classified as supercomputing, intelligent computing, or general computing (multiple selections are allowed); Service type: Services that computing resources can provide, such as computing services, storage services, network services, etc. Computational load: The total computational load of computing resources and the total computational load of intelligent computing (FIN), supercomputing (FSU), and general computing (FGE), expressed in FLOPS (floating-point operations per second); Storage capacity: Total storage provided by computing resources, including block storage (BL), object storage (OB), file storage (FL), and supported storage architectures: distributed storage (DT) or centralized storage (CT). Network bandwidth: The network bandwidth within computing resources, including different types of bandwidth such as TCP, IB (unlimited bandwidth), and ROCE (RDMA based on converged Ethernet); Chip Model and Quantity: The chip models and corresponding quantities in the available areas of the computing power resources. Chip models include Ascend 910B, A100, etc. Server type: The server type of computing power resources at the server level, starting with Y, and each server type code is assigned by the computing power service provider. CPU Model: The CPU model of the server where the computing resources are located; CPU core count: The number of CPU cores on the server where the computing resources are located; Memory: The memory of the server where the computing resources are located; Storage: The hard disk storage of the server where the computing resources are located; Number of servers: The number of servers corresponding to the specified server model; Computing power internet address: The unique internet address of the server to which the computing power resources belong.

[0009] Furthermore, in step S2, the encoding principles for computing power identifiers include uniqueness (ensuring the uniqueness of each computing power node within the computing power interconnection system), compatibility (the encoding must be compatible with different types of computing power providers and resources, including data centers, cloud computing centers, private cloud computing centers, etc.), applicability (the encoding should meet the information needs in the process of matching supply and demand for computing power resources and be applicable to various resources such as general computing power, supercomputing power, and intelligent computing power), scalability (the encoding system should reserve sufficient expansion space to adapt to future growth and changes in computing power resources), and scientificity (the encoding structure should be concise and clear, and check bits and security codes should be set when necessary to ensure the correctness and security of the encoding. Once the encoding structure is determined, it should remain relatively stable).

[0010] Furthermore, in step S2, the overall computing power identification architecture can be divided into a unique identifier and a computing power path. The unique identifier is the prefix of the computing power identification, and the computing power path is the suffix of the computing power identification. The encoding content specifically includes: 1) Computing power identifier prefix: The unique registration code for computing power identification consists of an enterprise code, an industry code, a city code, and an availability zone code, and is used to identify the enterprise entity, industry, and region of the computing power resources. Availability Zone Resource Total Amount Identifier: This includes resource type, service type, total computing power, total storage, network bandwidth, chip model and quantity, etc., providing a comprehensive description of the computing power resource status of the availability zone; Server specification identifiers include server type, CPU model, number of CPU cores, memory, storage, and number of servers, which are used to identify the specific specifications and quantity of each server; 2) Computing power identifier suffix (i.e., computing power path identifier): Provide the specific path to computing resources, including the computing power internet address, chip type, chip model, and chip unique number, to ensure that the smallest computing unit can be accurately located.

[0011] Furthermore, in step S2, the specific operation of encoding is as follows: according to the encoding principles and content, a specific encoding length and data type are assigned to each field.

[0012] Furthermore, in step S3, the reporting of computing resource information is divided into: First-time full reporting: Computing power providers need to submit full computing power resource information, including all fields such as the unique registration code of computing power identifier, total resource quantity identifier of availability zone, server specification identifier, and computing power path identifier, to ensure that the computing power interconnection platform can obtain complete resource information; Subsequent real-time updates: The computing power provider updates the computing power resource information in real time or periodically according to the actual situation, mainly reporting key fields such as the unique registration code of the computing power identifier, the total amount of available zone resources, server type and number of servers, to ensure the real-time and accuracy of computing power resources.

[0013] Furthermore, step S4 includes the following sub-steps: S41. Receiving and Verifying Computing Power Identifiers: The platform receives computing power identifier data reported by computing power providers and verifies the received computing power identifiers, checking their format, the legality of field values, and whether they conform to encoding principles. The verification process is as follows: 1) Identifier preprocessing: Parse the computing power identifier, extract the corresponding enterprise code, total available zone resources, server specifications and chip information, and verify the format compliance; 2) Static verification: The blockchain ledger is queried through the enterprise code to compare whether the enterprise public key, business license hash value and registration information are consistent. If the same enterprise corresponds to different enterprise codes, a "duplicate registration" alarm is triggered, the identifier is locked and the regulatory agency is notified. 3) Dynamic verification: Obtain the real-time status of resources through the identifier gateway or gateway plugin API, and compare the identifier type with the reported one. If the identifier type is "available", the corresponding chip status is "online"; 4) Policy matching: Logical matching is performed based on the set of logical verification policies and the resource dictionary. For example, the chip model must match the "chip model and quantity" field in the total available zone resource identifier. The total number of chips must equal the number of servers × the number of chips per server. If the logical policy rules are not met, the verification will fail. 5) Status Flag: Output the flag verification result. If the computing power flag passes the full verification and the resource status is normal, the computing power flag is marked as valid and scheduling is allowed; if the computing power flag fails the verification, the computing power flag is marked as invalid and scheduling is prohibited. S42. Computing power identifier resolution: Using an identifier resolution algorithm, the verified computing power identifier is resolved. The resolution process includes extracting various fields from the computing power identifier, such as the unique registration code of the computing power identifier, the total amount of available zone resources identifier, the server specification identifier, and the computing power path identifier. S43. Computing resource statistics: Statistical calculations are performed based on the parsed computing resource information; S44. Generate visual reports and statistical data: Utilize statistical tools to transform statistical results into visual reports and charts, intuitively displaying key information such as the distribution, quantity, and type of computing resources, making it easier for administrators and the computing power scheduling platform to quickly understand the overall status of computing resources; S45. Data Application and Services: The generated reports and statistical data are provided to administrators and the computing power scheduling platform. Specifically, administrators use the reports and statistical data to understand the overall status of computing power resources, and to carry out resource planning, optimization and management. The computing power scheduling platform uses this information to intelligently schedule and allocate computing power resources, thereby improving resource utilization efficiency and user satisfaction.

[0014] Furthermore, in step S43, the statistical content includes the number of connected computing power providers, resource types (such as general computing power, supercomputing power, intelligent computing power, etc.), service types (such as computing services, storage services, network services, etc.), computing volume (such as total computing volume, intelligent computing total computing volume, supercomputing total computing volume, general computing total computing volume), storage volume (such as total storage capacity), network type (such as TCP, IB, ROCE, etc.) and bandwidth, computing power internet address, chip model and quantity, and other key information, and the statistical results are stored in a structured manner.

[0015] Furthermore, in step S4, the computing power identifier resolution scheme is generally divided into four levels: computing power resource level, resource pool-side computing power identifier gateway level, regional computing power identifier server level, and root computing power identifier server level, and the computing power identifier range of the four levels of nodes increases sequentially, as detailed below: Computing resources refer to various nodes that provide computing services, such as cloud pools of computing resources or bare metal servers that directly provide computing power. The resource pool-side computing power identification gateway is designed to uniformly manage all computing nodes belonging to the same computing power provider. In order to shield GPU servers and computing power resource pools with different architectures, all computing power resources of the computing power provider are mapped to computing power identification and then uniformly managed by a single computing power identification gateway. The regional computing power identification server consists of an identification resolution system, an identification verification system, and an identification storage system. The identification verification system is responsible for performing format and logical verification on the computing power identifications input by each computing power identification gateway. The identification resolution system provides the ability to map computing power identifications to computing power resource information. The computing power storage system is responsible for persistently storing the verified computing power identifications. The regional computing power identification server is manually assigned a jurisdiction and is responsible for receiving and managing all computing power identifications within that jurisdiction. The root computing power identification server has a similar architecture to the regional server, but it additionally mounts an identification verification system responsible for verifying the computing power identification passed from the regional computing power identification server; its verification algorithm is as follows: ; Where ΔC represents the change ratio of computing power identifiers in the same availability zone, C_new represents newly added identifiers, and C_old represents existing identifiers in the database; If the ratio of the change between the old and new identifiers is less than the threshold, the old identifier data will be overwritten with the new identifier data; if the error is greater than the threshold, it will be classified as abnormal data. The specific hierarchical mapping steps for the computing resources are as follows: 1) After the computing power resources of each computing power provider and different architectures are mapped to computing power identifiers, the computing power identifiers are sent to the computing power identifier gateways on the resource pool side to which they belong, and the computing power identifier gateways on the resource pool side manage the computing power resources in a unified manner. 2) In order to connect computing resources to the computing network, each computing power identification gateway needs to send authentication information and the computing power identification under its jurisdiction to the regional computing power identification server, which will then store and manage all computing power identification data for the region. 3) In order to coordinate the computing power identification data across the country, the regional computing power identification servers need to synchronize their data to the root computing power identification server on a regular basis. To alleviate the processing bottleneck of the root computing power identification server, this section adopts an incremental push method to push the newly added data from the regional computing power identification servers to the root computing power identification server on a regular basis. The root computing power identification server verifies and receives the computing power identification through its identification verification system. Furthermore, in step S5, the user filters suitable computing resources through computing power identification fields (such as computing power provider, resource type, service type, etc.) and queries specific specification information through API. The computing power scheduling service platform opens specific computing power resources for the user based on the parsing results of the computing power identification, ensuring efficient utilization of resources. In practical applications, users will recursively query computing power identification data, starting from the regional computing power identification server. If the regional computing power identification server's database does not contain relevant computing power identification information, the regional computing power identification server will query the root computing power identification server for computing power identification information and return the final result to the user. The specific query latency is shown in the formula: ; in, To query the total latency, To query root server latency, To query the regional server latency, This represents the probability of a region server miss.

[0016] By adopting a tiered query scheme, the query traffic pressure on the root computing power identification server is significantly reduced. Since the communication latency between the regional computing power identification server and the user is lower, the user's query latency can be greatly optimized when the regional query is successful.

[0017] Furthermore, in step S5, the scheduling based on computing power identifiers is implemented through a computing power identifier matching algorithm based on computing power demand. The mechanism of the computing power identifier matching algorithm based on computing power demand is as follows: The user's resource demand is broken down into sub-items of each field of the identifier. One demand is mapped to the sub-items of chip, computing power, storage, memory, and network. The demand is compared with each set of identifiers in the computing power resource pool. The TOPSI algorithm is used to find the set of computing power identifiers that best matches the demand. The user's demand for chips, computing power, storage, memory, and network bandwidth is used as hard constraints. Resource groups that do not meet any of the constraints are removed from the computing power resource pool, and resource groups that fully meet all constraints are retained. The number of chips, computing power, storage scale, memory size, and network bandwidth of the retained resource groups are normalized using the following formula: in, Let i be the value of the i-th indicator in the resource group. For the number of chips, For computing power, For storage scale, For memory size, For network bandwidth requirements, Let i be the user's desired value for the i-th metric. For normalized results; Weighted standardization: Using the average cost of computing resources as weights, a weighted standardization decision matrix is ​​constructed. ,in, The representative weight is the average cost of computing resources. For normalized results; calculation of ideal and negative ideal solutions: determining the optimal solution for each index. and worst solution To form an ideal solution and negative ideal solution ,in, For a specific resource in a set of resources; Distance calculation: Calculate the Euclidean distance from the resource set to the ideal solution. Euclidean distance to the negative ideal solution Relative proximity ranking: by Calculate relative proximity, select The largest resource group is selected as the optimal match, and its computing power is activated and used.

[0018] This invention provides a computing power identification method for the computing power internet, which has the following beneficial effects: This invention provides a unified identifier for computing resources through clear steps and coding principles, solving the problem of the lack of unified standards and specifications for the description and location of computing resources in existing technologies. This facilitates the description, location, and addressing of computing resources, meeting the needs of the in-depth development of the computing power internet. Furthermore, through the registration, reporting, resolution, and statistics of computing power identifiers, a unified resource description is provided for the access, sensing, scheduling, and transaction settlement of computing resources, thereby achieving efficient management and scheduling of computing resources and improving their utilization efficiency. In the computing power resource information collection phase, this method requires the collection of comprehensive information on the computing power resources of computing power providers, ensuring that the computing power identifier can comprehensively and accurately describe the various characteristics of computing power resources, avoiding errors in resource matching and scheduling. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the steps of a computing power identification method for the computing power internet according to the present invention. Figure 2 This is a schematic diagram of step S4 of a computing power identification method for computing power Internet according to the present invention; Figure 3 This is a flowchart of a resource activation scheme for a computing power identification method for the computing power Internet according to the present invention; Figure 4 This is a flowchart of the identifier resolution system for a computing power identifier method for the computing power Internet according to the present invention. Detailed Implementation

[0020] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and should not be construed as limiting the scope of the invention.

[0021] like Figures 1-4 As shown, a computing power identification method for the computing power internet includes the following steps: S1. Collection of computing resource information: Computing power providers collect comprehensive information about their computing power resources. Specifically, they obtain dynamic information about computing power resources in real time or periodically through automated data collection tools such as computing power identification gateways or management platforms such as cloud management platforms, and organize it into a standardized data format. The specific content collected is as follows: Company Information: The company code corresponding to the company name is used to uniquely identify the computing power provider; Industry Classification: The industry category to which the computing power provider belongs, such as telecommunications, computer, finance, etc. City and Availability Zone Location: The city where the computing resources are located and the specific availability zone (data center area); Resource type: The type of computing power resource, which can be classified as supercomputing, intelligent computing, or general computing (multiple selections are allowed); Service type: Services that computing resources can provide, such as computing services, storage services, network services, etc. Computational load: The total computational load of computing resources and the total computational load of intelligent computing (FIN), supercomputing (FSU), and general computing (FGE), expressed in FLOPS (floating-point operations per second); Storage capacity: Total storage provided by computing resources, including block storage (BL), object storage (OB), file storage (FL), and supported storage architectures: distributed storage (DT) or centralized storage (CT). Network bandwidth: The network bandwidth within computing resources, including different types of bandwidth such as TCP, IB (unlimited bandwidth), and ROCE (RDMA based on converged Ethernet); Chip Model and Quantity: The chip models and corresponding quantities used in the availability zones of computing resources. Chip models include Ascend 910B, A100, etc. Server type: The server type of computing power resources at the server level, starting with Y, and each server type code is assigned by the computing power service provider. CPU Model: The CPU model of the server where the computing resources are located; CPU core count: The number of CPU cores on the server where the computing resources are located; Memory: The memory of the server where the computing resources are located; Storage: The hard disk storage of the server where the computing resources are located; Number of servers: The number of servers corresponding to the specified server model; Computing power internet address: The unique internet address of the server to which the computing power resources belong.

[0022] S2, Computing Power Identifier Code: According to the coding principles of computing power identification, the collected computing power resource information is encoded into computing power identification. The coding principles for computing power identification include: uniqueness (ensuring the uniqueness of each computing power node within the computing power interconnection system), compatibility (the code must be compatible with different types of computing power providers and resources, including data centers, cloud computing centers, private cloud computing centers, etc.), applicability (the code should meet the information needs in the process of matching supply and demand for computing power resources and be applicable to various resources such as general computing power, supercomputing power, and intelligent computing power), scalability (the coding system should reserve sufficient expansion space to adapt to future growth and changes in computing power resources), and scientific rigor (the coding structure should be concise and clear, and check bits and security codes should be set when necessary to ensure the correctness and security of the code. Once the coding structure is determined, it should remain relatively stable).

[0023] In this embodiment, a specific encoding length and data type are assigned to each field according to the encoding principles and content. The encoded content specifically includes: 1) Computing power identifier prefix: The unique registration code for computing power identification consists of an enterprise code, an industry code, a city code, and an availability zone code, and is used to identify the enterprise entity, industry, and region of the computing power resources. Availability Zone Resource Total Amount Identifier: This includes resource type, service type, total computing power, total storage, network bandwidth, chip model and quantity, etc., providing a comprehensive description of the computing power resource status of the availability zone; Server specification identifiers include server type, CPU model, number of CPU cores, memory, storage, and number of servers, which are used to identify the specific specifications and quantity of each server; 2) Computing power identifier suffix (i.e., computing power path identifier): Provide the specific path for computing resources, including the computing power internet address, chip type (CPU / GPU / NPU, etc.), chip model (Ascend 910b / A100 / H100, etc.), and chip unique serial number, to ensure that the smallest computing unit can be accurately located.

[0024] S3, Computing Power Identification Registration and Reporting: Computing power providers register a unique computing power identifier code through the computing power marketplace platform and obtain the qualification to report computing power resources. They then use the computing power identifier prefix and suffix to report detailed computing power resource information to the computing power interconnection platform.

[0025] In this embodiment, the reporting of computing resource information is divided into: First-time full reporting: Computing power providers need to submit full computing power resource information, including all fields such as the unique registration code of computing power identifier, total resource quantity identifier of availability zone, server specification identifier, and computing power path identifier, to ensure that the computing power interconnection platform can obtain complete resource information; Subsequent real-time updates: The computing power provider updates the computing power resource information in real time or periodically according to the actual situation, mainly reporting key fields such as the unique registration code of the computing power identifier, the total amount of available zone resources, server type and number of servers, to ensure the real-time and accuracy of computing power resources.

[0026] S4. Computing Power Identification Resolution and Statistics: After receiving computing power identifiers, the computing power interconnection platform verifies, parses, and statistically analyzes them to form a computing power identifier registry. Based on the parsing results, it statistically analyzes key information such as the number of connected computing power providers, resource types, service types, computing load, storage load, network type and bandwidth, computing power internet address, and chip model and quantity. In this embodiment, the computing power interconnection platform uses identifier resolution algorithms and statistical tools to automatically process the reported computing power identifiers, generating visualized reports and statistical data for use by administrators and the computing power scheduling platform.

[0027] S41. Computing Power Identifier Reception and Verification: The platform receives computing power identifier data reported by computing power providers and verifies the received computing power identifiers, checking their format, the legality of field values, and whether they conform to encoding principles to ensure the accuracy and integrity of the data. The verification process is as follows: 1) Identifier preprocessing: Parse the computing power identifier, extract the corresponding enterprise code, total available zone resources, server specifications and chip information, and verify the format compliance; 2) Static verification: The blockchain ledger is queried through the enterprise code to compare whether the enterprise public key, business license hash value and registration information are consistent. If the same enterprise corresponds to different enterprise codes, a "duplicate registration" alarm is triggered, the identifier is locked and the regulatory agency is notified. 3) Dynamic verification: Obtain the real-time status of resources through the identifier gateway or gateway plugin API, and compare the identifier type with the reported one. If the identifier type is "available", the corresponding chip status is "online"; 4) Policy matching: Logical matching is performed based on the set of logical verification policies and the resource dictionary. For example, the chip model must match the "chip model and quantity" field in the total available zone resource identifier. The total number of chips must equal the number of servers × the number of chips per server. If the logical policy rules are not met, the verification will fail. 5) Status Flag: Output the flag verification result. If the computing power flag passes the full verification and the resource status is normal, the computing power flag is marked as valid and scheduling is allowed; if the computing power flag fails the verification, the computing power flag is marked as invalid and scheduling is prohibited. Assume that the Computing Power Identifier (CRID) of different availability zones is a tuple of multi-dimensional attributes: ID: Static global enterprise identifier, i.e., a unique registration code for computing power identification; S(t): State function, i.e., identifier type; : A set of resource attributes, which can be used to identify the total amount of resources in the area and the server specifications; Each chip is a unique identifier.

[0028] Define the verification function V as follows: Where P is the set of strategies. This is a resource dictionary. For validation to pass, the following conditions must be met: P includes, but is not limited to: Where F represents the total computational load, , , These represent the computing power of intelligent computing, supercomputing, and general computing within the current available area, respectively.

[0029] Where S represents the total storage capacity. , , These represent the storage capacity of block storage, object storage, and file storage within the current available area, respectively.

[0030] Wherein, SN represents the number of servers of a certain type. This refers to the number of Internet address codes associated with this type of server.

[0031] in, This indicates the total computing power within the availability zone. This indicates the available computing power within the availability zone.

[0032] Where F represents the total computational cost. This indicates the computational cost per chip within the available area. Indicates the total number of chips in the available area; S42. Computing power identifier resolution: Using an identifier resolution algorithm, the verified computing power identifier is resolved. The resolution process includes extracting various fields from the computing power identifier, such as the unique registration code of the computing power identifier, the total amount of available zone resources identifier, the server specification identifier, and the computing power path identifier. S43. Computing Resource Statistics: Based on the parsed computing resource information, statistical calculations are performed. The statistical content includes the number of connected computing power providers, resource types (such as general computing power, supercomputing power, intelligent computing power, etc.), service types (such as computing services, storage services, network services, etc.), computing volume (such as total computing volume, intelligent computing total computing volume, supercomputing total computing volume, general computing total computing volume), storage volume (such as total storage capacity), network type (such as TCP, IB, ROCE, etc.) and bandwidth, computing power internet address, chip model and quantity, and other key information. The statistical results are stored in a structured manner. S44. Generate visual reports and statistical data: Utilize statistical tools to transform statistical results into visual reports and charts, intuitively displaying key information such as the distribution, quantity, and type of computing resources, making it easy for administrators and computing scheduling platforms to quickly understand the overall status of computing resources. These reports and statistical data can be generated periodically or updated in real time as needed. S45. Data Application and Services: The generated reports and statistical data are provided to administrators and the computing power scheduling platform. Specifically, administrators use the reports and statistical data to understand the overall status of computing power resources, and to carry out resource planning, optimization and management. The computing power scheduling platform uses this information to intelligently schedule and allocate computing power resources, thereby improving resource utilization efficiency and user satisfaction.

[0033] In this embodiment, the computing power identifier resolution scheme is generally divided into four levels: computing power resource level, resource pool-side computing power identifier gateway level, regional computing power identifier server level, and root computing power identifier server level. The scope of computing power identifiers under the jurisdiction of the four levels of nodes increases sequentially, as detailed below: Computing resources refer to various nodes that provide computing services, such as cloud pools of computing resources or bare metal servers that directly provide computing power. The resource pool-side computing power identification gateway is designed to uniformly manage all computing nodes belonging to the same computing power provider. In order to shield GPU servers and computing power resource pools with different architectures, all computing power resources of the computing power provider are mapped to computing power identification and then uniformly managed by a single computing power identification gateway. The regional computing power identification server consists of an identification resolution system, an identification verification system, and an identification storage system. The identification verification system is responsible for performing format and logical verification on the computing power identifications input by each computing power identification gateway. The identification resolution system provides the ability to map computing power identifications to computing power resource information. The computing power storage system is responsible for persistently storing the verified computing power identifications. The regional computing power identification server is manually assigned a jurisdiction and is responsible for receiving and managing all computing power identifications within that jurisdiction. The root computing power identification server has a similar architecture to the regional server, but it additionally mounts an identification verification system responsible for verifying the computing power identification passed from the regional computing power identification server; its verification algorithm is as follows: ; Where ΔC represents the change ratio of computing power identifiers in the same availability zone, C_new represents newly added identifiers, and C_old represents existing identifiers in the database; If the ratio of the change between the old and new identifiers is less than the threshold, the old identifier data will be overwritten with the new identifier data; if the error is greater than the threshold, it will be classified as abnormal data. The specific hierarchical mapping steps for the computing resources are as follows: 1) After the computing power resources of each computing power provider and different architectures are mapped to computing power identifiers, the computing power identifiers are sent to the computing power identifier gateways on the resource pool side to which they belong, and the computing power identifier gateways on the resource pool side manage the computing power resources in a unified manner. 2) In order to connect computing resources to the computing network, each computing power identification gateway needs to send authentication information and the computing power identification under its jurisdiction to the regional computing power identification server, which will then store and manage all computing power identification data for the region. 3) In order to coordinate the computing power identification data across the country, the regional computing power identification servers need to synchronize their data to the root computing power identification server on a regular basis. To alleviate the processing bottleneck of the root computing power identification server, this section adopts an incremental push method to push the newly added data from the regional computing power identification servers to the root computing power identification server on a regular basis. The root computing power identification server verifies and receives the computing power identification through its identification verification system. In step S5, scheduling based on computing power identifiers is implemented through a computing power identifier matching algorithm based on computing power demand. The mechanism of the computing power identifier matching algorithm based on computing power demand is as follows: The user's resource demand is broken down into sub-items of each field of the identifier. One demand is mapped to chip, computing power, storage, memory, and network sub-items. The demand is compared with each set of identifiers in the computing power resource pool. The TOPSI algorithm is used to find the set of computing power identifiers that best matches the demand. The user's demand for chips, computing power, storage, memory, and network bandwidth is used as hard constraints. Resource groups that do not meet any of the constraints are removed from the computing power resource pool, and resource groups that fully meet all constraints are retained. The number of chips, computing power, storage scale, memory size, and network bandwidth of the retained resource groups are normalized using the following formula: in, Let i be the value of the i-th indicator in the resource group. For the number of chips, For computing power, For storage scale, For memory size, For network bandwidth requirements, Let i be the user's desired value for the i-th metric. For normalized results; Weighted standardization: Using the average cost of computing resources as weights, a weighted standardization decision matrix is ​​constructed. ,in, The representative weight is the average cost of computing resources. For normalized results; calculation of ideal and negative ideal solutions: determining the optimal solution for each index. and worst solution To form an ideal solution and negative ideal solution ,in, For a specific resource in a set of resources; Distance calculation: Calculate the Euclidean distance from the resource set to the ideal solution. Euclidean distance to the negative ideal solution Relative proximity ranking: by Calculate relative proximity, select The largest resource group is selected as the optimal matching result, and its computing power resource is activated and used. In this embodiment, users filter suitable computing resources through computing power identification fields (such as computing power provider, resource type, service type, etc.) and query specific specification information through API. The computing power scheduling service platform opens specific computing power resources for users based on the parsing results of the computing power identification, ensuring efficient utilization of resources. In practical applications, users will recursively query computing power identification data, starting from the regional computing power identification server. If the regional computing power identification server's database does not contain relevant computing power identification information, the regional computing power identification server will query the root computing power identification server for computing power identification information and return the final result to the user. The specific query latency is shown in the formula: ; in, To query the total latency, To query root server latency, To query the regional server latency, This represents the probability of a region server miss.

[0034] By adopting a tiered query scheme, the query traffic pressure on the root computing power identification server is significantly reduced. Since the communication latency between the regional computing power identification server and the user is lower, the user's query latency can be greatly optimized when the regional query is successful.

[0035] The embodiments of the present invention are given for illustrative and descriptive purposes only, and are not intended to be exhaustive or to limit the invention to the forms disclosed. Many modifications and variations will be apparent to those skilled in the art. The embodiments were chosen and described in order to better illustrate the principles and practical application of the invention, and to enable those skilled in the art to understand the invention and to design various embodiments with various modifications suitable for a particular purpose.

Claims

1. A method for identifying computing power for the Internet of Computing Power, characterized in that, Includes the following steps: S1. Computing Resource Information Collection: The computing resource provider collects comprehensive information on its computing resources, including enterprise information, industry affiliation, city and availability zone location, resource type, service type, computing volume, storage volume, network bandwidth, chip model and quantity, server type, CPU model, number of CPU cores, memory, storage, number of servers, and computing power internet address; S2. Computing Power Identifier Encoding: Based on the encoding principles and structural requirements of computing power identifiers, the collected computing power resource information is encoded into computing power identifiers. In step S2, the overall computing power identifier architecture is divided into a unique identifier and a computing power path. The unique identifier is the prefix of the computing power identifier, and the computing power path is the suffix of the computing power identifier. The encoding content specifically includes: 1) Computing power identifier prefix: The unique registration code for computing power identification consists of an enterprise code, an industry code, a city code, and an availability zone code, and is used to identify the enterprise entity, industry, and region of the computing power resources. Availability Zone Resource Total Amount Identifier: Covers resource type, service type, computing power, storage capacity, network bandwidth, chip model and quantity, comprehensively describing the computing power resource status of the availability zone; Server specification identifiers include server type, CPU model, number of CPU cores, memory, storage, and number of servers, which are used to identify the specific specifications and quantity of each server; 2) Computing power identifier suffix: Provide the specific path to computing resources, including the computing power internet address, chip type, chip model, and chip unique number, to ensure that the smallest computing unit can be accurately located; S3. Computing power identification registration and reporting: Computing power providers register a unique registration code for computing power identification through the computing power market platform and obtain the qualification to report computing power resources. Then, using the computing power identification prefix and suffix, they report detailed computing power resource information to the computing power interconnection platform. S4. Computing Power Identifier Resolution and Statistics: After receiving the computing power identifier, the computing power interconnection platform verifies, parses, and statistically analyzes it to form a computing power identifier registry. Based on the parsing results, it also statistically analyzes the key access information. Step S4 includes the following sub-steps: S41. Receiving and Verifying Computing Power Identifiers: The computing power interconnection platform receives computing power identifier data reported by computing power providers and verifies the received computing power identifiers, checking their format, the legality of field values, and whether they conform to encoding principles. The verification process is as follows: 1) Identifier preprocessing: Parse the computing power identifier, extract the corresponding enterprise code, total available zone resources, server specifications and chip information, and verify the format compliance; 2) Static verification: Query the blockchain ledger through the enterprise code to compare whether the enterprise public key, business license hash value and registration information are consistent; 3) Dynamic verification: Obtain the real-time status of resources through the identifier gateway or gateway plugin API, and compare whether the identifier type is consistent with the reported one; 4) Policy matching: Logical matching is performed based on the logical verification policy set and the resource dictionary; 5) Status flag: Outputs the verification result; The computing power identifier resolution scheme is generally divided into four levels: computing power resource level, resource pool side computing power identifier gateway level, regional computing power identifier server level, and root computing power identifier server level, and the computing power identifier range of the four level nodes increases sequentially. S5. Scheduling based on computing power identifier: Users select computing power resources through the computing power scheduling service platform, and activate and use appropriate computing power resources based on the resolution result of the computing power identifier and their own needs.

2. The computing power identification method for the computing power Internet according to claim 1, characterized in that, In step S1, the computing power provider obtains dynamic information about computing power resources in real time or periodically through the computing power identification gateway or cloud management platform, and organizes it into a standardized data format.

3. The computing power identification method for the computing power Internet according to claim 1, characterized in that, In step S1, the specific content collected is as follows: Company Information: The company code corresponding to the company name is used to uniquely identify the computing power provider; Industry Classification: The industry category to which the computing power provider belongs; City and Availability Zone Location: The city and specific availability zone where the computing resources are located; Resource type: The type of computing power resource, which is classified as supercomputing, intelligent computing, or general computing. Service type: The types of services that computing resources can provide; Computational load: The total computational load of computing resources and the total computational load of intelligent computing, supercomputing, and general computing, expressed in FLOPS; Storage capacity: Total storage provided by computing resources, including block storage, object storage, file storage, and supported storage architecture: distributed storage or centralized storage; Network bandwidth: The network bandwidth within computing resources, including TCP, IB, or ROCE bandwidth; Chip Model and Quantity: The chip models and corresponding quantities included in the available area; Server type: The server type of computing power resources at the server level, starting with Y, and each server type code is assigned by the computing power service provider. CPU Model: The CPU model of the server where the computing resources are located; CPU core count: The number of CPU cores on the server where the computing resources are located; Memory: The memory of the server where the computing resources are located; Storage: The hard disk storage of the server where the computing resources are located; Number of servers: The number of servers corresponding to the specified server model; Computing power internet address: The unique internet address of the server to which the computing power resources belong.

4. The computing power identification method for the computing power Internet according to claim 1, characterized in that, In step S2, the encoding principles for computing power identifiers include uniqueness, compatibility, applicability, scalability, and scientific rigor.

5. A computing power identification method for a computing power internet according to claim 1, characterized in that, In step S2, the specific operation of encoding is as follows: according to the encoding principles and content, a specific encoding length and data type are assigned to each field.

6. A computing power identification method for a computing power internet according to claim 1, characterized in that, In step S3, the reporting of computing resource information is divided into: For the first full report: The computing power provider needs to submit all computing power resource information, including all fields such as the unique registration code of the computing power identifier, the total amount of resources in the availability zone, the server specification identifier, and the computing power path identifier. Subsequent real-time updates: The computing power provider updates the computing power resource information in real time or periodically according to the actual situation, and reports the unique registration code of the computing power identifier, the total amount of available zone resources, server type and number of servers.

7. A computing power identification method for a computing power internet according to claim 1, characterized in that, The specific hierarchical mapping steps for the computing resources are as follows: 1) After the computing power resources of each computing power provider and different architectures are mapped to computing power identifiers, the computing power identifiers are sent to the computing power identifier gateways on the resource pool side to which they belong, and the computing power identifier gateways on the resource pool side manage the computing power resources in a unified manner. 2) In order to connect computing resources to the computing network, each computing power identification gateway needs to send authentication information and the computing power identification under its jurisdiction to the regional computing power identification server, which will then store and manage all computing power identification data for the region. 3) In order to coordinate the computing power identification data across the country, the regional computing power identification servers need to synchronize their data to the root computing power identification server on a regular basis. In order to alleviate the processing bottleneck of the root computing power identification server, an incremental push method is adopted to push the newly added data of the regional computing power identification servers to the root computing power identification server on a regular basis. The root computing power identification server verifies and receives the computing power identification through its identification verification system.

8. A computing power identification method for a computing power internet according to claim 1, characterized in that, Step S4 further includes the following sub-steps: S42. Computing power identifier resolution: Using an identifier resolution algorithm, the verified computing power identifier is resolved. The resolution process includes extracting each field from the computing power identifier. S43. Computing resource statistics: Based on the parsed computing resource information, statistical calculations are performed. The statistical content includes the number of connected computing providers, resource type, service type, computing volume, storage volume, network type and bandwidth, computing power Internet address, chip model and quantity, and the statistical results are stored in a structured manner. S44. Generate visual reports and statistical data: Use statistical tools to transform statistical results into visual reports and charts; S45. Data Application and Services: The generated reports and statistical data are provided to administrators and the computing power scheduling platform. Specifically, administrators use the reports and statistical data to understand the overall status of computing power resources, and to carry out resource planning, optimization and management. The computing power scheduling platform uses the reports and statistical data to intelligently schedule and allocate computing power resources.

Images