Blockchain-based computing power resource dynamic scheduling system and method

Abstract

The blockchain-based computing power resource dynamic scheduling system and method of the present application, the system comprises a resource monitoring agent module that collects state resource signals in real time; a core trust layer module that registers and authenticates node identities through a blockchain smart contract, publishes task demand contracts, and defines scheduling strategy rules to form policy constraint signals; a scheduling coordinator module that receives dynamic resource signals and policy constraint signals, and outputs resource allocation instruction signals through an optimization decision algorithm; a task execution gateway module that receives resource allocation instruction signals and generates task state feedback signals; and a dynamic reputation database module that continuously receives policy constraint signals and task state feedback signals, updates node dynamic reputation scores, forms reputation weight signals, and feeds back to the scheduling coordinator module. The blockchain-based computing power resource dynamic scheduling system and method of the present application can solve the problems of trust deficiency and dynamic deficiency in computing power resource scheduling.

Description

Technical Field

[0001] The present invention relates to the interdisciplinary field of blockchain and cloud computing, and in particular to a system and method for dynamically scheduling computing resources based on blockchain. Background Art

[0002] Traditional computing power scheduling relies on centralized platforms to manage resource allocation, posing risks of single points of failure and data tampering. Users need to trust the fairness and security of the platform, but in practice, platforms may manipulate pricing, conceal resource status, or leak sensitive data. Furthermore, centralized architectures experience delayed resource status updates, making it difficult to respond promptly to sudden load fluctuations, resulting in delays for high-priority tasks and idle resources. Early decentralized solutions attempted to replace centralized platforms with blockchains, such as building computing power marketplaces based on smart contracts to enable direct transactions between resource providers and demanders. While these solutions improved transparency, they placed all scheduling decisions on-chain, making them limited by blockchain performance and unable to support high-frequency, real-time scheduling requirements. The high gas costs and network latency resulting from complex decision-making prevented the system from handling urgent computing tasks. Existing hybrid architectures attempt to move some functions off-chain, but lack effective trust mechanisms: off-chain schedulers can act maliciously and are difficult to hold accountable, and resource nodes lack the means to constrain the misreporting of performance metrics.

[0003] Furthermore, the coordinated scheduling of nodes across heterogeneous resources requires a comprehensive assessment of multi-dimensional dynamic factors such as real-time computing power, network topology, and historical reputation. Current solutions either rely on simple static weight models or employ unverified black-box algorithms, resulting in scheduling decisions that deviate from the optimal solution. In the trust verification phase, existing technologies often employ full-scale repeated computation or trusted hardware authentication. The former results in secondary consumption of computing power, while the latter increases node deployment costs and compatibility thresholds. More critically, the disconnect between resource dynamics and reputation evaluation means that high-performance but low-reputation nodes may be overscheduled, while high-reputation nodes cannot be downgraded in a timely manner when their performance suddenly degrades, reducing the overall reliability of the system. Therefore, there is an urgent need for a dynamic optimization mechanism that integrates the blockchain's trusted foundation with efficient off-chain scheduling capabilities and can couple resource status and reputation evolution in real time. Summary of the Invention

[0004] In view of the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a blockchain-based dynamic scheduling system and method for computing power resources, which is used to solve the problems of lack of trust and insufficient dynamism in computing power resource scheduling. The present invention constructs a core trust layer constraint strategy on the chain, collects multi-dimensional resource data in real time off the chain, and generates optimization instructions by integrating a dynamic scoring model with a reputation feedback mechanism by the scheduling coordinator. The task gateway drives distributed execution and feedbacks on-chain evidence, thus achieving a closed loop of accurate resource value assessment, dynamic matching of task requirements, and trusted traceability of the scheduling process in a decentralized environment.

[0005] The application provides a blockchain-based computing power resource dynamic scheduling system, comprising:

[0006] A resource monitoring agent module, which collects hardware performance indicators, real-time load states and network topology data of the distributed nodes in real time to form a dynamic resource signal;

[0007] A core trust layer module, which registers and authenticates node identities, publishes task demand contracts and defines scheduling strategy rules through a blockchain smart contract to form a strategy constraint signal;

[0008] A scheduling coordinator module, which receives the dynamic resource signal and the strategy constraint signal, generates a resource value evaluation result based on a multi-dimensional dynamic scoring model, fuses task service quality requirements and real-time market supply and demand states, and outputs a resource allocation instruction signal through an optimization decision algorithm;

[0009] A task execution gateway module, which receives the resource allocation instruction signal, triggers the distribution and collaborative execution of cross-node computing power tasks, and generates a task state feedback signal;

[0010] A dynamic reputation library module, which continuously receives the strategy constraint signal and the task state feedback signal, updates the dynamic reputation score of the node, forms a reputation weight signal, and feeds back to the resource value evaluation process of the scheduling coordinator module.

[0011] In an embodiment of the application, the resource monitoring agent module further comprises a distributed probe unit and a data fusion unit; characterized in that: the distributed probe unit is deployed locally on each computing power node and periodically collects underlying indicators including central processor core occupancy, graphics processor memory utilization, network round-trip delay and bandwidth availability; the data fusion unit receives raw indicator data from all probe units, eliminates transient noise interference through time series analysis, and integrates the network topology relationship between nodes to generate a structured dynamic resource signal; the dynamic resource signal is packaged as a standard data packet containing a timestamp and a unique node identifier, and is pushed to the data input interface of the scheduling coordinator module in real time.

[0012] In an embodiment of the application, the scheduling strategy rules defined in the core trust layer module include service quality grading strategies and market constraint strategies; characterized in that: the service quality grading strategies automatically generate differentiated scheduling priority labels according to the deadline threshold, minimum computing power guarantee level and fault tolerance redundancy requirements declared in the task demand contract; the market constraint strategies preset resource pricing floating range, transaction commission ratio and default penalty coefficient through the smart contract; both of them constitute executable clauses in the strategy constraint signal and are synchronously transmitted to the strategy analysis engine of the scheduling coordinator module in real time through the blockchain event listening mechanism.

[0013] In an embodiment of the present application, the multi-dimensional dynamic scoring model in the scheduling coordinator module includes a hardware benchmark score, a real-time performance score, and a network quality score; characterized in that: the hardware benchmark score calculates an initial static score according to the central processing unit model, graphics processor architecture, and persistent storage capacity provided when the node is registered; the real-time performance score adjusts the weight dynamically by analyzing the load fluctuation trend and resource idle window in the dynamic resource signal; the network quality score is generated according to the topology hop count and historical transmission stability from the node to the task data source; the three are fused into a resource value evaluation result according to a preset weighting formula, wherein the real-time performance score weight increases adaptively with the increase of task urgency.

[0014] In an embodiment of the present application, the optimization decision algorithm includes a demand matching stage and a resource allocation stage; characterized in that: in the demand matching stage, the available node pool is filtered according to the geographical affinity constraint in the task service quality requirement, and a preliminary sorting is performed based on the resource value evaluation result; in the resource allocation stage, a heuristic search algorithm is used to generate a node combination scheme to maximize global resource utilization under the conditions of meeting the deadline and budget double constraints; the finally output resource allocation instruction signal contains the target node list, task fragment mapping relationship, and execution timing control parameters.

[0015] In an embodiment of the present application, the task execution gateway module further includes a task fragment unit and a fault tolerance control unit; characterized in that: the task fragment unit splits the complex computing task into several independent sub-task packages according to the mapping relationship in the resource allocation instruction signal, and adds a data integrity check code; the fault tolerance control unit monitors the sub-task execution timeout or check failure event, and automatically triggers the task redistribution process based on the backup node list; the task state feedback signal contains the actual start time, execution time consumption, and verification result log of each sub-task package, which is sent to the core trust layer module after being digitally signed.

[0016] In an embodiment of the present application, the process of updating the node dynamic reputation score by the dynamic reputation library module includes behavior analysis and score reconstruction; characterized in that: the behavior analysis unit extracts the task completion timeliness deviation rate and result verification pass rate indicators from the task state feedback signal, and analyzes the node historical violation records from the policy constraint signal; the score reconstruction unit inputs the above indicators into the reputation decay model, wherein the recent behavior data is calculated by exponential weighted accumulation, and the reputation base is calibrated in combination with the total online duration of the node; the finally generated reputation weight signal contains the reputation level label and the scheduling recommendation weight value.

[0017] In an embodiment of the present application, the blockchain-based computing power resource dynamic scheduling system further comprises a cross-chain verification relay module; characterized in that: the module is deployed between the core trust layer module and the external blockchain network, and when a task demand contract involving cross-chain computing power resource scheduling is detected, a cross-chain resource query request is automatically generated; the node trustworthiness proof and resource bid data on the external chain are obtained through the relay bridge, and after format conversion, they are input into the scheduling coordinator module as extended strategy constraint signals; the settlement data generated by the task execution gateway module is forwarded to the external chain by the relay module for atomic cross-chain clearing.

[0018] In an embodiment of the present application, the reputation weight signal feedback mechanism of the dynamic reputation library module includes active push and passive response dual modes; characterized in that: in the active push mode, whenever the node reputation score changes by more than a set threshold, an incremental update signal is immediately sent to the scheduling coordinator module; in the passive response mode, the scheduling coordinator module sends a reputation query request before starting resource evaluation, and the dynamic reputation library module returns the latest full-amount reputation snapshot; the two modes are distinguished by signal type identifiers, and the scheduling coordinator module dynamically switches the call according to the real-time decision-making requirements.

[0019] The present application also includes a blockchain-based computing power resource dynamic scheduling method, comprising:

[0020] S1: Real-time collection of hardware performance indicators, real-time load status and network topology data of distributed nodes to form dynamic resource signals;

[0021] S2: Register and authenticate node identity through a blockchain smart contract, publish a task demand contract and define scheduling strategy rules to form strategy constraint signals;

[0022] S3: Receive dynamic resource signals and strategy constraint signals, generate resource value evaluation results based on a multi-dimensional dynamic scoring model, fuse task service quality requirements and real-time market supply and demand status, and output resource allocation instruction signals through an optimization decision algorithm;

[0023] S4: Receive resource allocation instruction signals, trigger cross-node computing power task distribution and collaborative execution, and generate task status feedback signals;

[0024] S5: Continuously receive strategy constraint signals and task status feedback signals, update node dynamic reputation scores, form reputation weight signals, and feed back to the resource value evaluation process of the scheduling coordinator module.

[0025] The application provides a blockchain-based computing power resource dynamic scheduling system and method, which constructs a core trust layer constraint strategy on the chain, collects multi-dimensional resource data off-chain in real time, generates optimization instructions by fusing a dynamic scoring model and a reputation feedback mechanism through a scheduling coordinator, drives distributed execution through a task gateway, and feeds back on-chain storage, so as to realize precise evaluation of resource value, dynamic matching of task demand, and traceable closed loop of the scheduling process in a decentralized environment. BRIEF DESCRIPTION OF DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0027] Figure 1 The system architecture diagram of the blockchain-based computing power resource dynamic scheduling system;

[0028] Figure 2 The schematic diagram of the running process of the computing power resource dynamic scheduling system displayed on the blockchain;

[0029] Figure 3 The method flowchart of the blockchain-based computing power resource dynamic scheduling method. DETAILED DESCRIPTION

[0030] The embodiments of the application are described below through specific, concrete examples, and those skilled in the art can easily understand other advantages and effects of the application from the content disclosed in the specification. The application can also be implemented or applied through other different specific embodiments, and each detail in the specification can be modified or changed based on different views and applications without departing from the spirit of the application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0031] It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the application in a schematic manner, and only the components related to the application are shown in the diagrams, not the number, shape and size of the components when actually implemented. The actual implementation of each component may be a random change in number, shape and size, and the layout pattern of the components may be more complex.

[0032] In the following description, numerous specific details are discussed in order to provide a thorough explanation of embodiments of the application. It will be apparent, however, to one skilled in the art, that embodiments of the application can be practiced without these specific details. In other instances, well-known structures and devices are not described in detail in order to avoid obscuring embodiments of the application.

[0033] Referring to Figures 1-3 , a blockchain-based computing power resource dynamic scheduling system and method of the present application are shown. The blockchain-based computing power resource dynamic scheduling system of the present application includes a resource monitoring agent module, which collects hardware performance indicators, real-time load states and network topology data of distributed nodes in real time to form dynamic resource signals; a core trust layer module, which registers and authenticates node identities through a blockchain smart contract, publishes task demand contracts and defines scheduling strategy rules to form policy constraint signals; a scheduling coordinator module, which receives dynamic resource signals and policy constraint signals, generates resource value evaluation results based on a multi-dimensional dynamic scoring model, fuses task service quality requirements and real-time market supply and demand states, and outputs resource allocation instruction signals through an optimization decision algorithm; a task execution gateway module, which receives resource allocation instruction signals, triggers distribution and collaborative execution of cross-node computing power tasks, and generates task state feedback signals; and a dynamic reputation library module, which continuously receives policy constraint signals and task state feedback signals, updates node dynamic reputation scores, forms reputation weight signals, and feeds back to the resource value evaluation process of the scheduling coordinator module.

[0034] As Figure 1As shown, the blockchain-based computing power resource dynamic scheduling system includes a collaborative operation mechanism of five core modules. The resource monitoring agent module serves as the system perception layer, and through the lightweight agent program implanted in each distributed computing power node, it continuously captures three types of key dynamic indicators: hardware performance indicators covering central processor benchmark frequency, graphics processor stream processor quantity, and persistent storage read / write rate, real-time load state including central processor core utilization rate instantaneous value, graphics processor memory occupation ratio, and memory exchange frequency, and network topology data obtained through inter-node probe communication transmission path hop count, round-trip delay, and packet loss rate, etc. network quality parameters. After these raw indicators are sampled at a millisecond level, the data is cleaned to remove sensor noise interference, and aggregated into structured dynamic resource signals according to the preset time window. The signal uses binary encoding to package the node unique identifier, timestamp sequence, and multi-dimensional indicator key-value pair, and pushes it to the message middleware through the publish / subscribe mode. The core trust layer module is built on the permissioned blockchain network, and its smart contract includes three types of core logic: node registration contract, task publishing contract, and strategy management contract. The node registration contract requires the resource provider to mortgage digital assets and submit a hardware authentication report to generate a trusted identity credential, the task publishing contract analyzes the task descriptor submitted by the demand side and generates a task demand contract containing the service quality level, budget upper limit, and data source location, and the strategy management contract allows authorized nodes to vote on the dynamic update of scheduling strategy rules. Finally, the identity credential, task contract, and strategy rule are compiled into a strategy constraint signal, which is broadcast to the entire network in the form of a blockchain transaction event. The scheduling coordinator module serves as the decision-making center, and sets up a dedicated message queue to receive dynamic resource signals and strategy constraint signals. The signal analysis engine first separates the node identity whitelist, task priority label, and market constraint parameters in the strategy constraint signal, and decodes the dynamic resource signal into a node real-time state matrix. The multi-dimensional dynamic scoring model runs the scoring process based on the matrix: the hardware benchmark score calls the static parameters stored in the registration contract to calculate the theoretical computing power value, the real-time efficiency score dynamically reduces the weight according to the degree of deviation of the load indicator from the benchmark line, and the network quality score calculates the transmission cost based on the task data source location and node network indicators. The three scoring inputs are weighted and fused to generate resource value evaluation results, with the weighting coefficients dynamically adjusted by the service quality requirements in the strategy constraint signal. The optimization decision algorithm is then started: in the demand matching stage, the node pool is filtered according to the task geographical affinity constraint, in the resource allocation stage, the tabu search algorithm is used to solve the global resource utilization rate maximization scheme under the constraints of time limit and budget, and finally the resource allocation instruction signal containing the target node address list, task fragment mapping table, and execution timeout threshold is generated.After the task execution gateway module receives the instruction signal, the task dispatcher splits the computing task into atomic subtask packages according to the mapping table and distributes them to the target nodes through a point-to-point encryption channel; the node local execution engine runs the subtasks and generates a local state report containing the execution timestamp and resource consumption log; the collaborative controller assembles all local reports into a global task state feedback signal after verifying data consistency, and sends it to the blockchain network with an additional digital signature. The dynamic reputation library module is deployed in an off-chain database, which listens to the node initial reputation value in the strategy constraint signal and the execution result in the task state feedback signal from the blockchain; the reputation update engine periodically scans new data, the behavior analyzer extracts the task completion timeliness deviation rate, the result verification pass rate and the resource data authenticity mark, inputs the reputation model based on the exponential decay function: the recent behavior data is assigned a higher weight, and the malicious behavior record triggers a punitive score reduction, and finally outputs the dynamic reputation score; the score is integrated with the node online duration and historical stability indicators into a reputation weight signal, which is called back to the score fusion unit of the scheduling coordinator module through the application programming interface, realizing the closed-loop coupling of resource value assessment and reputation feedback.

[0035] Further, the resource monitoring agent module internally adopts a double-layer architecture of distributed probe units and data fusion units. The distributed probe units are embedded in each computing node in the form of an operating system kernel module, containing three types of collectors: hardware performance probes, real-time load probes, and network topology probes. The hardware performance probes are activated when the node is registered, obtain the instruction set extension support flag by accessing the central processor model-specific register, call the graphics processor driver application programming interface to read the number of shader cores and memory bandwidth data, and scan the interface protocol and queue depth parameters of the persistent storage device. The real-time load probes poll system resource statistics information at a frequency of seconds, including central processor interrupt request rate, graphics processor computation unit active period ratio, memory page fault exception count, and network adapter buffer overflow events. The network topology probes actively send link layer discovery protocol packets to adjacent nodes after starting, construct an initial topology graph based on response latency and path trajectory, and dynamically maintain the topology relationship by listening to open shortest path first protocol routing update messages. The data fusion unit is deployed in the regional aggregation node, receives the raw index stream reported by the probe unit, and its data processing pipeline contains four levels of operations: the noise filtering layer uses the sliding window mean algorithm to eliminate transient jitter, such as applying the Hann window smoothing process to the central processor utilization rate; the data alignment layer corrects the time stamps of each node according to the network time protocol server, ensuring time synchronization of cross-node indicators; the correlation analysis layer detects the causal relationship between hardware performance and real-time load, such as identifying whether a sudden increase in memory occupancy rate is accompanied by a graphics processor computation unit idle; the topology integration layer injects network delay indicators between nodes into the initial topology graph to generate a weighted dynamic topology matrix. The structured dynamic resource signal generated finally adopts a protocol buffer encoding scheme: a nested data structure body is defined, the top layer contains a global sequence number and a data collection round number, the middle layer is indexed by a node unique identifier, and the bottom layer is divided into a static attribute segment, a dynamic load segment, and a network topology segment; the static attribute segment fixes the hardware benchmark parameters after registration and authentication, the dynamic load segment stores the index statistics in the recent time window, and the network topology segment records the path weight list from the node to the core switch node. The signal is multicast to the specified port of the dispatch coordinator cluster through the user datagram protocol, and the receiving end enables the cyclic redundancy check to ensure data transmission integrity.

[0036] Specifically, the scheduling strategy rule system in the core trust layer module is divided into two subsystems: quality of service grading strategy and market constraint strategy. The quality of service grading strategy acts on the task publishing contract analysis stage: when the demand side submits the computing task descriptor, the strategy engine scans the service quality declaration field in the descriptor, including the task latest completion timestamp, minimum computing power guarantee level and fault tolerance redundancy requirement; time-sensitive tasks are divided into emergency levels according to the difference between the deadline and the current blockchain time, and corresponding scheduling priority labels are generated; the computing power guarantee level is mapped to the minimum threshold combination of central processor core number, graphics processor floating point capability and memory capacity; the fault tolerance redundancy requirement is converted into the task replica quantity and the fault switching time window parameter, which together constitute a multi-dimensional vector of the task priority label. The market constraint strategy is configured through the on-chain governance contract: the resource pricing floating rule defines the benchmark computing power unit price curve, which is adjusted according to the historical market supply and demand ratio on a daily basis, and sets the upper limit of the single transaction price fluctuation; the transaction service charge ratio is designed according to the ladder model, large transactions enjoy a discount on the service charge ratio but need to pay additional on-chain settlement gas fee; the default penalty coefficient matrix includes two scenarios of provider default and demand default, the former covers node disconnection, insufficient computing power and result error, and the latter focuses on payment overtime and task description inaccuracy, and the penalty amount is calculated based on the proportion of the collateral and the severity of the default. The executable clauses of the strategy rule are generated as follows: the task priority label vector is input into the strategy compiler to generate a binary code segment containing the task scheduling weight coefficient; the market constraint parameters are compiled into smart contract code blocks of resource pricing function, service charge calculation function and default judgment function; after being verified by the blockchain node consensus, they are written into the new block as the payload of the strategy constraint signal. The blockchain event listening mechanism realizes strategy synchronization: the blockchain listener deployed by the scheduling coordinator module subscribes to the storage change events of the strategy management contract, and when a new block containing a strategy update transaction is detected, the event parser extracts the strategy operation code; after the operation code is simulated and executed in a safe sandbox environment, the incremental update package of the strategy constraint signal is output, and the local strategy cache library is updated through the remote procedure call interface; the strategy cache library adopts a version control mechanism to ensure that scheduling decisions are always based on the latest effective strategy version, while historical versions are retained for audit and traceability.

[0037] In an embodiment of the present application, the multi-dimensional dynamic scoring model in the scheduling coordinator module is composed of three core dimensions of hardware benchmark score, real-time performance score and network quality score. The hardware benchmark score calculation process calls the node registration information stored in the core trust layer module, extracts the central processing unit instruction set extension capability flag, combines the graphics processing unit unified computing device architecture core number and the persistent storage random read-write rate index, and generates a theoretical calculation capability index using a static weighting formula; among them, the instruction set extension capability is graded and valued according to key technologies such as support vector machine instruction set and matrix extension instruction set, the graphics processing unit parameters are converted into theoretical floating point performance values according to the stream processor number and texture unit proportion, and the storage performance is quantified through queue depth test data. The real-time performance score is dynamically calculated based on the second-level sampling sequence reported by the resource monitoring agent module: a time series model of the utilization rate of each core of the central processing unit is established, and the future short-term load trend is predicted through the autoregressive integrated moving average algorithm; the graphics processing unit performance evaluation introduces the product factor of the actual memory bandwidth occupancy rate and the computing unit active period; the memory subsystem performance is dynamically weighted according to the page fault exception frequency and the swap partition usage rate; the final performance score is the negative deviation function of the measured value from the benchmark theoretical value. The network quality score is generated according to the differences of specific task requirements: first, analyze the geographic location coordinates of the task data source in the strategy constraint signal, calculate the network topology hop number from the target node to the data source; combine the average value of the round-trip delay and the standard deviation of the packet loss rate of the path calculated from the historical transmission log; introduce the network congestion perception coefficient, which is dynamically adjusted by monitoring the recent transmission control protocol congestion window size change rate. The three scores input fusion engine performs weighted calculation: the hardware benchmark score weight is fixed as the benchmark value, the real-time performance score weight is positively correlated with the emergency level declared in the task demand contract, and the network quality score weight is automatically improved for data transmission intensive tasks; the built-in normalization processing unit of the fusion engine eliminates the dimension difference, applies a score attenuation factor to the ultra-low reputation node, and outputs the resource value evaluation result matrix containing the three-dimensional sub-score and the comprehensive score value of each node for the current task. The matrix is used as the core input parameter of the optimization decision algorithm.

[0038] As Figure 2As shown, the optimization decision algorithm is divided into two sequentially executed logical units: the demand matching stage and the resource allocation stage. The demand matching stage first constructs a node screening funnel: the geographic affinity constraint analyzes the location-sensitive tags in the task demand contract, and if this constraint is enabled, only nodes within the same autonomous system as the task data source are selected; the computing power threshold constraint filters low-performance nodes according to the minimum computing power guarantee level of the task; the reputation access constraint excludes blacklisted nodes marked by the dynamic reputation library module. The nodes passing through the funnel form an initial available node pool, and the pre-sorter arranges them in descending order of the comprehensive score in the resource value evaluation results. The resource allocation stage uses an improved tabu search algorithm to solve the multi-constraint optimization problem: the solution space is defined as the subset combination of the available node pool, and the objective function is to maximize the global resource utilization rate, which is defined as the ratio of the total computing amount of the task to the sum of the actual occupied node computing power; the constraint conditions include the task deadline threshold and the demand side budget upper limit, where the time constraint is estimated by the critical path method, and the budget constraint integrates the resource pricing function and the estimated execution time. The algorithm initialization stage generates a random feasible solution as the current solution, and the neighborhood operation uses the node replacement strategy: randomly select the node with the lowest score in the current solution, and find a replacement node with a higher score and meet the constraints in the unselected nodes. The tabu table records the identifiers of the recently replaced nodes, and prohibits them from re-entering the solution set within a certain number of iterations to avoid cyclic oscillation. The amnesty criterion is activated when the target function value of the candidate solution is better than the historical optimal solution, allowing the tabu restriction to be broken. The algorithm termination condition is to reach the maximum number of iterations or the number of consecutive iterations without improvement exceeds the limit, and the final output resource allocation instruction signal contains a four-tuple structure: the target node address list determines the physical nodes participating in the calculation; the task fragment mapping table defines the binding relationship between the subtasks after the complex task is decomposed and the nodes; the execution timing control parameter specifies the time window and synchronization barrier position of each subtask start; the dynamic pricing list lists the actual settlement unit price of each node computing power, which is generated within the floating range allowed by the strategy constraint signal according to the real-time market supply and demand status.

[0039] Further, the task execution gateway module guarantees the reliability of distributed execution through the cooperative mechanism of the task fragmentation unit and the fault tolerance control unit. After receiving the resource allocation instruction signal, the task fragmentation unit starts the fragmentation engine: for the computation-intensive task, it is divided into several homogeneous sub-task packages according to the data parallel principle, each package is allocated the same calculation logic but processes different data blocks; for the pipeline task, a function splitting strategy is adopted, and the processing flow is divided into sequential execution of heterogeneous sub-task stages. Each sub-task package is attached with an integrity check mechanism: the data block type task generates a Merkle tree root hash value and writes it into the package header; the algorithm logic type task injects a lightweight cyclic redundancy check code. The fragmentation engine generates a task dependency graph at the same time, marking the predecessor and successor constraints between sub-tasks. The sub-task package is distributed to the target node through a transmission encryption channel, and a quantum-resistant encryption algorithm is used to encapsulate the transmission payload. The fault tolerance control unit deploys a heartbeat monitor and a timeout timer for double protection: the heartbeat monitor receives the execution progress messages returned by each node at a frequency of seconds, the message contains the percentage of processed data and the resource consumption snapshot; the timeout timer initializes the maximum allowed execution time of each sub-task according to the execution timing control parameters. When detecting that the sub-task progress is stagnant or the timeout trigger, the exception processor starts a three-level response process: the primary response sends an execution status challenge request to the target node, and if it is confirmed as a temporary fault, the time limit is extended; the intermediate response selects a node of the same specification from the backup node list, redistributes the sub-task package and migrates the intermediate state data; the advanced response executes the same sub-task in multiple backup nodes in parallel, and uses the majority consensus mechanism to determine the correct result. After all sub-tasks are executed, the result aggregator verifies the integrity check code of each sub-task package and merges to generate the final calculation result. The task status feedback signal contains a structured log sequence: each sub-task package records the actual start timestamp, completion timestamp, consumed central processing unit clock cycles and graphics processing unit stream processor instruction number; the exception event is separately marked with type code and processing process; the overall task is attached with a digital signature and a result hash value, which is submitted to the core trust layer module through the blockchain application programming interface.

[0040] As Figure 3As shown, it is a blockchain-based computing power resource dynamic scheduling method of the application. S1: Real-time collection of hardware performance indicators, real-time load state and network topology data of distributed nodes to form dynamic resource signals; S2: Registering and authenticating node identity through a blockchain smart contract, publishing task demand contract and defining scheduling strategy rules to form policy constraint signals; S3: Receiving the dynamic resource signals and policy constraint signals, generating resource value evaluation results based on a multi-dimensional dynamic scoring model, fusing task service quality requirements and real-time market supply and demand state, and outputting resource allocation instruction signals through an optimized decision algorithm; S4: Receiving the resource allocation instruction signals, triggering the distribution and collaborative execution of cross-node computing power tasks, and generating task state feedback signals; S5: Continuously receiving the policy constraint signals and the task state feedback signals, updating the node dynamic reputation score, forming the reputation weight signal, and feeding back to the resource value evaluation process of the scheduling coordinator module.

[0041] Specifically, the node dynamic reputation score updating mechanism of the dynamic reputation library module is realized by a behavior analysis unit and a score reconstruction unit in series. The behavior analysis unit listens to two key data streams: the initial reputation seed value and the historical violation record summary when the node is registered are extracted from the policy constraint signal; three types of behavior evidence are analyzed from the task state feedback signal: the task completion timeliness deviation rate is calculated by comparing the actual completion timestamp with the planned completion time window in the resource allocation instruction signal; the result verification pass rate is the proportion of the number of failed task package verification to the total number of tasks; the resource data authenticity flag is added by the resource monitoring agent module when abnormal data fluctuation is detected. The score reconstruction unit runs a reputation model based on time decay: a reputation base value update formula is established, in which the initial reputation seed value weight exponentially decays with the increase of online time, and the recent behavior data is weighted and accumulated in a sliding window. The specific calculation process defines the behavior evidence weight coefficient: the task completion timeliness deviation rate exceeding the threshold is amplified by a quadratic function penalty factor; a result verification failure is considered as a serious violation, triggering a stepwise score reduction for a single event; a resource data false reporting behavior is immediately subject to a fixed penalty value once confirmed. The model introduces a positive incentive mechanism: nodes that continuously complete tasks successfully and have stable resource data obtain accelerated points, and the point growth rate is positively correlated with task complexity. The score output is normalized before being output: the original reputation score is mapped to a standardized interval of zero to one hundred percent, and a reputation upper limit is set for long-term online high-stability nodes. The final generated reputation weight signal contains three parts of data: the dynamic reputation score value retains two decimal places of precision; the reputation level label is divided into four levels of trusted, observed, restricted and blacklisted according to the score interval; the scheduling recommendation weight value is a linear function of the reputation score, used to adjust the node ranking priority in the resource value evaluation of the scheduling coordinator module. The signal is asynchronously pushed to the reputation input buffer area of the scheduling coordinator module through the message queue, and is written into the distributed database for audit query.

[0042] The blockchain-based computing power resource dynamic scheduling system and method of the present application, by constructing a chain core trust layer constraint strategy, collecting multi-dimensional resource data in real time off-chain, generating optimization instructions by the scheduling coordinator combining dynamic scoring model and reputation feedback mechanism, driving distributed execution by the task gateway and feeding back on-chain storage, realizes the closed loop of accurate evaluation of resource value, dynamic matching of task demand and credible traceability of scheduling process in a decentralized environment.

[0043] Therefore, by the blockchain-based computing power resource dynamic scheduling system and method of the present application, the problem of lack of trust and insufficient dynamic in computing power resource scheduling is solved.

[0044] The above embodiments only exemplarily illustrate the principles and effects of the present application, and are not used to limit the present application. Any person skilled in the art can modify or change the above embodiments without departing from the spirit and scope of the present application. Therefore, all equivalent modifications or changes completed by those skilled in the art without departing from the spirit and technical thought disclosed by the present application should be covered by the claims of the present application.

Claims

1. The dynamic scheduling system of computing power resources based on blockchain is characterized by: include: A resource monitoring agent module, which collects hardware performance indicators, real-time load status and network topology data of distributed nodes in real time to form dynamic resource signals; The core trust layer module registers and authenticates node identities through blockchain smart contracts, publishes task requirement contracts, and defines scheduling policy rules to form a policy constraint signal. The scheduling policy rules defined in the core trust layer module include a service quality grading strategy and a market constraint strategy. The service quality grading strategy automatically generates differentiated scheduling priority tags based on the deadline threshold, minimum computing power guarantee level, and fault tolerance redundancy requirements stated in the task requirement contract. The market constraint strategy presets the resource pricing floating range, transaction fee ratio, and default penalty coefficient through smart contracts. Together, the two constitute the executable terms in the policy constraint signal and are synchronized in real time to the policy parsing engine of the scheduling coordinator module through the blockchain event monitoring mechanism. A scheduling coordinator module receives the dynamic resource signal and the policy constraint signal, generates a resource value assessment result based on a multi-dimensional dynamic scoring model, integrates the task service quality requirements with the real-time market supply and demand status, and outputs a resource allocation instruction signal through an optimization decision algorithm; the optimization decision algorithm includes a demand matching stage and a resource allocation stage. In the demand matching stage, the available node pool is screened according to the geographical affinity constraints in the task service quality requirements, and a preliminary sorting is performed based on the resource value assessment result; in the resource allocation stage, a heuristic search algorithm is used to generate a node combination plan with the goal of maximizing global resource utilization under the dual constraints of deadline and budget; the resource allocation instruction signal finally output includes a target node list, a task shard mapping relationship, and an execution timing control parameter; A task execution gateway module receives resource allocation instruction signals, triggers the distribution and collaborative execution of cross-node computing tasks, and generates a task status feedback signal; the task status feedback signal contains the actual start time, execution time, and verification result log of each subtask package, and is sent to the core trust layer module after being digitally signed; A dynamic reputation library module continuously receives the policy constraint signal and the task status feedback signal, updates the node dynamic reputation score, forms a reputation weight signal, and feeds it back to the resource value assessment process of the scheduling coordinator module; the process of updating the node dynamic reputation score by the dynamic reputation library module includes behavior analysis and score reconstruction. The behavior analysis unit extracts the task completion time deviation rate and result verification pass rate indicators from the task status feedback signal, and analyzes the node's historical violation records from the policy constraint signal; the score reconstruction unit inputs the above indicators into the reputation decay model, in which the recent behavior data is calculated using exponential weighted accumulation, and the reputation base is calibrated in combination with the total online time of the node. The final generated reputation weight signal contains the reputation level label and the scheduling recommendation weight value.

2. The blockchain-based computing resource dynamic scheduling system according to claim 1 is characterized in that: The resource monitoring agent module further includes a distributed probe unit and a data fusion unit; it is characterized in that: the distributed probe unit is deployed locally at each computing power node, and periodically collects underlying indicators including central processing unit core occupancy, graphics processing unit memory utilization, network round-trip delay and bandwidth availability; the data fusion unit receives original indicator data from all probe units, eliminates instantaneous noise interference through time series analysis, and integrates the network topology relationship between nodes to generate a structured dynamic resource signal; the dynamic resource signal is encapsulated as a standard data packet containing a timestamp and a unique node identifier, and is pushed to the data input interface of the scheduling coordinator module in real time.

3. The blockchain-based computing resource dynamic scheduling system according to claim 1 is characterized in that: The multi-dimensional dynamic scoring model in the scheduling coordinator module includes hardware benchmark scores, real-time performance scores, and network quality scores. The hardware benchmark scores are calculated based on the CPU model, GPU architecture, and persistent storage capacity provided during node registration to calculate the initial static score. The real-time performance score is dynamically adjusted by analyzing the load fluctuation trend and resource idle window in the dynamic resource signal; the network quality score is generated based on the topological hop count and historical transmission stability from the node to the task data source; the three are fused into the resource value assessment result according to a preset weighting formula, among which the real-time performance score weight increases adaptively as the task urgency increases.

4. The blockchain-based computing resource dynamic scheduling system according to claim 1 is characterized in that: The task execution gateway module further includes a task slicing unit and a fault-tolerant control unit. The task slicing unit splits the complex computing task into several independent subtask packages according to the mapping relationship in the resource allocation instruction signal, and adds a data integrity check code; the fault-tolerant control unit monitors the subtask execution timeout or verification failure event, and automatically triggers the task redistribution process based on the backup node list.

5. The blockchain-based computing resource dynamic scheduling system according to claim 1 is characterized in that: The blockchain-based computing power resource dynamic scheduling system further includes a cross-chain verification relay module, which is deployed between the core trust layer module and the external blockchain network. When a task requirement contract involving cross-chain computing power resource scheduling is detected, it automatically generates a cross-chain resource query request; Obtain the credibility proof and resource quotation data of external chain nodes through the relay bridge, and input them into the scheduling coordinator module as the extended strategy constraint signal after format conversion; The settlement data generated by the task execution gateway module is forwarded by the relay module to the external chain for atomic cross-chain settlement.

6. The blockchain-based computing resource dynamic scheduling system according to claim 1 is characterized in that: The reputation weight signal feedback mechanism of the dynamic reputation library module includes active push and passive response modes. In the active push mode, whenever the change in the node reputation score exceeds the set threshold, an incremental update signal is immediately sent to the scheduling coordinator module; in the passive response mode, the scheduling coordinator module sends a reputation query request before starting resource evaluation, and the dynamic reputation library module returns the latest full reputation snapshot. The two modes are distinguished by a signal type identifier, and the scheduling coordinator module dynamically switches and calls according to the real-time decision-making requirements.

7. The method for dynamic scheduling of computing resources based on a blockchain computing resource dynamic scheduling system according to any one of claims 1 to 6, comprising: S1: Real-time collection of hardware performance indicators, real-time load status, and network topology data of distributed nodes to form dynamic resource signals; S2: Register and authenticate node identities through blockchain smart contracts, publish task requirement contracts, and define scheduling policy rules to form policy constraint signals; S3: Receive the dynamic resource signal and the policy constraint signal, generate a resource value assessment result based on a multi-dimensional dynamic scoring model, integrate the task service quality requirements with the real-time market supply and demand status, and output a resource allocation instruction signal through an optimized decision algorithm; S4: Receives resource allocation instruction signals, triggers the distribution and coordinated execution of cross-node computing tasks, and generates task status feedback signals; S5: Continuously receive the policy constraint signal and the task status feedback signal, update the node dynamic reputation score, form a reputation weight signal, and feed it back to the resource value evaluation process of the scheduling coordinator module.

Images