A charging and settlement method of computing power and content distribution fusion service

Abstract

The present application relates to the technical field of computing power service and content distribution network charging, in particular to a kind of computing power and content distribution fusion service billing settlement method, the method first normalizes the fusion service request and is decomposed into minimum independent semantic chain element, constructs the directed graph with ring of containing forward, control and feedback dependence;The logical degree of linkage of each chain element and resource consumption density double-dimensional score are calculated, joint routing distribution is completed based on the score, time cross is executed in the form of intersection to closed loop subgraph, and convergence state is judged;After execution is completed, chain element logic fingerprint is generated to complete on-chain evidence, based on chain element dependence relationship automatic and account, contribution coefficient is determined by multi-agent negotiation, and the account is completed by triggering smart contract, while realizing bill traceability and abnormal rectification.This method can improve the fineness, fairness and transparency of fusion service charging, and adapt to the business scenarios of deep coupling of computing power and content distribution.

Description

Technical Field

[0001] This invention relates to the field of computing power services and content delivery network billing technology, and more specifically, to a billing and settlement method for integrated computing power and content delivery services. Background Technology

[0002] This invention relates to the field of billing and settlement technology for integrated computing power services and content delivery networks (CDNs). With the deep integration of edge computing, distributed computing power networks and content distribution services, integrated computing power scheduling and content distribution services have become the core form of digital content services and distributed application deployment. Correspondingly, reliable and refined billing and settlement technology is the core foundation for ensuring the orderly operation of services.

[0003] Currently, billing for computing power services primarily uses hardware resource usage time and API call count as the core measurement metrics, while billing for content distribution focuses on bandwidth usage and traffic consumption. This disconnect in measurement metrics creates a problem in converged service scenarios, failing to adapt to closed-loop task scenarios where computing power processing and content distribution are deeply coupled. Existing solutions struggle to jointly quantify the semantic value of tasks and resource consumption in converged services, easily leading to billing discrepancies between resource input and task output. For closed-loop execution tasks with feedback dependencies, duplicate billing in intermediate rounds is prone to occur. Furthermore, inconsistent measurement standards between heterogeneous computing power nodes and distribution nodes result in cumbersome reconciliation processes across participants, insufficient settlement transparency, and a lack of traceable and quickly correctable reliable settlement mechanisms.

[0004] To address the numerous shortcomings of the existing technologies, this invention proposes a billing and settlement method for integrated computing power and content distribution services. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a billing and settlement method for integrated computing power and content distribution services.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A billing and settlement method for a service that integrates computing power and content distribution specifically includes the following steps: S1. Normalization of fused service requests and generation of directed cyclic design graph: Receive fused service requests and complete normalization processing. Decompose the normalized requests into several minimum independent semantic chain elements and construct a directed cyclic design graph containing forward dependencies, control dependencies and feedback dependencies. S2. Logical constraint score calculation: For each chain element in the directed cyclic graph, calculate the logical constraint score that represents the semantic complexity of the task and the strength of the constraints. S3. Resource consumption density score calculation: Perform heterogeneous hardware normalization processing on each chain element, calculate the corresponding resource consumption density score, and form a two-dimensional score result. S4. Joint Routing and Closed-Loop Execution: Based on the two-dimensional scoring results, complete the joint routing allocation and distribution path planning of the chain elements, execute the closed-loop subgraph in the graph using the time-interval method, and determine the closed-loop convergence status. S5. Logical fingerprint generation and on-chain evidence storage: Generate a unique logical fingerprint for the completed chain element and complete the on-chain evidence storage of the minimum evidence chain. S6. Automatic merging and smart contract revenue sharing: Automatic merging is completed based on the dependency relationship between chain elements to form a settlement package. The contribution coefficient of each participant is determined through multi-agent negotiation, and the smart contract is triggered to complete the revenue sharing operation. S7. Bill Output and Anomaly Correction: Output standardized and traceable bills, and trigger corresponding correction processes for settlement anomaly scenarios.

[0007] Furthermore, when constructing the directed and cyclic fixed graph, the target elements in the fusion service request are first extracted to form the target skeleton, and then expanded into the processing action sequence and the distribution action sequence. The disassembly and reassembly are completed in units of the smallest independent semantic chain element. The corresponding feedback edges are retained for the feedback dependencies that meet the preset conditions, forming a directed graph that allows closed loops. Each node in the graph is bound to a fixed metric window.

[0008] Furthermore, based on the independence of the input and output semantic slots of the chain element, the feasibility of separate billing of the execution result, and the subsequent consumption attributes of the output content, the chain element is split or merged; for feedback edges that do not meet the retention conditions, boundary cache nodes are inserted and rewritten as cross-time window dependencies to avoid scheduling loops.

[0009] Furthermore, when calculating the logical constraint score, multi-dimensional logical features corresponding to the chain element are extracted from the current session, and the execution uncertainty obtained from the real-time running status of the session is combined to quantify the logical value and constraint strength of the chain element; when calculating the resource consumption density score, the original resource measurement collected during the execution of the chain element is mapped to a unified scale based on the normalized parameters obtained from the node standard reference test, thus completing the normalized quantification of resource consumption.

[0010] Furthermore, in the process of calculating the resource consumption density score, the bandwidth usage of chain elements in the content distribution scenario is measured based on the request frequency, cross-node replication times, and back-to-origin depth of content fragments within the same session to obtain the effective distribution volume, which is used to complete the bandwidth-related measurement in the resource consumption density score.

[0011] Furthermore, for strongly connected closed-loop subgraphs in a directed cyclic graph, all chain elements within the subgraph are grouped into the same settlement package. The closed-loop execution adopts a time-interleaved approach, with forward chain elements and feedback chain elements being executed in adjacent micro-time windows and sharing the same settlement context. Only the first round of forward output and the final convergence output of the closed loop are subject to main billing, while intermediate rounds only accumulate correction costs without generating separate entries.

[0012] Furthermore, the convergence of the closed loop is determined based on the semantic change rate of the output content before and after the closed loop execution. The convergence threshold is determined by the upper limit of latency and the lower limit of quality in the service level constraints. When performing joint route allocation, candidate nodes that meet the constraints are first screened, and then the comprehensive execution cost is calculated based on the two-dimensional scoring results to select the optimal node allocation result.

[0013] Furthermore, after the chain element is executed, a unique logical fingerprint is generated based on the chain element's input semantic description, execution context information, output content summary, and bill count snapshot. Only the logical fingerprint and its corresponding associated identification information are written into the evidence ledger, while the original execution data is stored locally on each participant's local machine. For the closed-loop settlement package, a unique root fingerprint is generated based on the fingerprints of each chain element within the package as the external index of the settlement package.

[0014] Furthermore, based on the semantic slot dependency strength between chain elements, automatic consolidation determination is completed, and related chain elements that meet the conditions are merged into the same settlement package. After the settlement package is formed, the agents of each participating party complete a limited number of asynchronous negotiations based on the evidence stored in the settlement package to determine the effective contribution coefficient and settlement amount of each participating party, triggering the smart contract to complete the fund transfer and distribution. In case of abnormal scenarios in the settlement process, the corresponding settlement package is frozen and the correction process is triggered to complete the recalculation and update of the bill.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention breaks through the limitations of traditional billing schemes that measure by resource call counts and data bytes by decomposing fusion service requests into the smallest independent semantic chain elements and constructing a directed cyclic metering graph adapted to closed-loop tasks. This achieves precise matching between billing objects and task semantic units. Simultaneously, it quantifies the semantic complexity and constraint strength of tasks through logical constraint scoring, and combines this with resource consumption density scoring normalized by heterogeneous hardware to construct a two-dimensional joint metering system. This system can adapt to the metering needs of scenarios integrating computing power and content distribution, helping to alleviate billing deviations caused by mismatches between resource input and task output, and improving the rationality and precision of billing. 2. This invention employs a time-interleaved execution method for closed-loop subgraphs with feedback dependencies. It only performs main billing on the first round of forward output and the final convergence output of the closed loop; intermediate rounds only accumulate and correct costs without generating separate entries, effectively avoiding duplicate billing issues in closed-loop tasks. Simultaneously, it achieves on-chain notarization by generating unique logical fingerprints for chain elements, combined with automatic merging of chain element dependencies, multi-agent negotiation to determine contribution coefficients, and automatic revenue sharing via smart contracts. This automates and makes the settlement process traceable, helps unify the measurement standards of heterogeneous nodes, simplifies cross-participant reconciliation processes, and improves settlement transparency and dispute resolution efficiency. Attached Figure Description

[0016] Figure 1 A flowchart of a billing and settlement method for a service that integrates computing power and content distribution; Figure 2 This is a flowchart of the logic constraint scoring steps based on a logic probability dual-drive network of the present invention. Figure 3 This is a flowchart of the heterogeneous hardware normalization scoring device and resource consumption density calculation steps of the present invention. Detailed Implementation

[0017] Example, refer to Figure 1 This embodiment discloses a billing and settlement method for a converged computing power and content distribution service. The invention is deployed on a converged service orchestration platform, which includes at least: a service access gateway, a logic decomposer, a directed and cyclical calculation graph generator, a logic probability dual-driven scoring network, a heterogeneous hardware normalization scoring device, a joint routing executor, an execution metric collector, a logic fingerprint generator, an on-chain evidence storage interface, a multi-agent negotiation and settlement network, and a smart contract executor. The modules are processed in the order of request description flow, chain metagraph flow, scoring flow, execution flow, evidence flow, and settlement flow, and are associated through a unified session identifier, chain meta identifier, and settlement package identifier. Specifically, the method includes the following steps: Step 1: Standardize the integration service request and generate a directed cyclic design graph.

[0018] External requests are transformed into quantifiable chain metagraphs, providing a unified measurement object for subsequent scoring, routing, and settlement, and outputting a traceable and closed-loop billing semantic chain metagraph.

[0019] In this embodiment, the service access gateway receives a converged service request. The request includes at least: content object identifier, processing target, distribution target, service level constraint, latency limit, privacy level, participating settlement entity identifier, and budget boundary. The access gateway performs protocol unpacking, field normalization, and time window numbering on the request to form a standardized session description. The logic decomposer hierarchically decomposes the standardized session description. The decomposition process does not use the number of model calls or the number of bytes transmitted as the smallest unit, but rather the smallest independent semantic chain element as the decomposition object. Each semantic chain element includes at least the following fields: chain element identifier, input semantic slot set, output semantic slot set, execution type, hard constraint set, time window, privacy level, candidate resource domain, and expected write-back target. The logic decomposer employs the following processing sequence: S101. Extract action words, object words, limiting words, and deliverable words from the user's goal and form a goal skeleton; S102, Further expand the target skeleton into a processing action sequence and a distribution action sequence; S103. Based on whether the input and output semantic slots are independent, whether the execution result can be billed separately, and whether the output will be directly consumed by subsequent steps, the chain elements are split or merged. S104. If the execution result of the subsequent chain element will reverse the parameters, cache state, fragment selection or reprocessing strategy of the preceding chain element, then a feedback edge is retained between the preceding chain element and the subsequent chain element, thereby forming a directed graph that allows closed loops. Feedback edges in the graph are retained instead of being cut off when the following conditions are met: First, the feedback information can be returned within the maximum tolerable latency of the current session; second, the feedback only changes the execution parameters, fragment subsets, or scheduling weights of the preceding chain elements, without changing the main goal of the session; and third, the resource consumption corresponding to the feedback edge can be grouped into the same settlement package. For feedback edges that do not meet the above conditions, the system inserts boundary cache nodes in the graph and rewrites them as cross-time window dependencies, thereby avoiding infinite loop scheduling. In the resulting directed and cyclic fixed graph, nodes represent the smallest independent semantic chain element, and edges represent forward dependencies, control dependencies, or feedback dependencies. Each node is bound to a fixed metric window, and all subsequent resource sampling, logical scoring, and settlement are performed around this node.

[0020] Step 2: Logical traction scoring based on a dual-drive network of logical probabilities.

[0021] like Figure 2 As shown, the logical value and constraint strength of the quantified chain element in the task output are used to output a logical constraint score. This score is not a repetitive expression of resource usage, but a quantitative representation of the semantic complexity of the task and the constraint of the goal.

[0022] For any chain element The system extracts four types of features from the current session in real time: topic layer features, background layer features, control layer features, and target layer features; Among them, the topic layer features are used to represent the number of branch choices and the strength of mutual exclusion relationships involved in the chain element; the background layer features are used to represent the breadth of the chain element's dependence on external states, fragment contexts, and associated objects; the control layer features are used to represent the density of key semantic slots that still need to be retained after context compression; and the target layer features are used to represent the rigidity of accuracy, latency, format, and compliance boundaries. The logical branch performs typed message propagation on the directed cyclic deterministic graph. The predecessor edge, successor edge, and feedback edge within the loop update the hidden state of the node with different gating parameters to preserve the causal constraint information in the loop. The probabilistic branch does not build an external knowledge base, but directly reads the queuing length, link jitter, node temperature fluctuation, number of retries, and fragment retransmission rate in the adjacent micro-time window of the current session, and outputs the execution uncertainty of the link element. The logical constraint score is calculated using the following formula: ; in, For chain element The logical constraint scoring system has a value range of 0 to 1. This is a compressed activation function used to map scores to bounded intervals; It is a four-dimensional logical feature vector of the chain element, whose elements correspond to the features of the question layer, background layer, control layer and target layer in turn; This is a logical feature weight vector, which is written into the parameter file by the system during deployment through a standard calibration process; The execution uncertainty of the probability branch output is determined based on the real-time observation results within the most recent micro-time windows of the current session, rather than the offline knowledge base retrieval results. In order to allow the scores to be directly involved in scheduling, the system also... Stabilization is performed so that when the closed-loop subgraph where the chain element is located experiences continuous oscillation, the logical branch prioritizes retaining the characteristics of the target layer and control layer, so that highly rigid nodes have higher priority in subsequent routing.

[0023] Step 3: Calculation of heterogeneous hardware normalization scoring device and resource consumption density.

[0024] like Figure 3 As shown, in this step, each candidate execution node periodically runs a standard reference test task; the standard reference test task includes fixed fragment transcoding, fixed length inference, fixed packet distribution, and fixed memory usage maintenance; the system obtains a normalized parameter set for each node based on the test results, including processing throughput, distribution throughput, peak latency, energy consumption level, and memory carrying capacity; When a chain element is actually executed, the execution metric collector synchronously collects the following raw metrics from kernel counters, driver interface, and distribution protocol statistics: number of central processing cycles, graphics processing usage, number of ingress bytes, number of egress bytes, peak memory, number of fragment retransmissions, and execution duration. Subsequently, the heterogeneous hardware normalization scoring device maps the above original measurements to a unified scale to obtain a resource consumption density score: ; in, For chain element Resource consumption density score; This is the normalized resource vector; These represent the central processing usage, graphics processing usage, distribution bandwidth usage, and peak memory usage collected during the execution of the chain element, respectively. Representing nodes respectively The corresponding normalized parameters obtained under the standard reference test; This is a weight vector for resource dimensions related to service type, and its value is determined by the configuration template when the service is accessed. In content distribution scenarios, bandwidth metrics use the converted effective distribution volume, rather than the raw number of bytes transmitted. Effective distribution volume is a standardized metric that eliminates invalid bandwidth usage and matches the actual business value of the content with transmission costs. The specific conversion rules, statistical definitions, and operational procedures are as follows: First dimension: Request frequency conversion (distinguishing the business value of popular and unpopular content) Statistical criteria: The statistics include the cumulative number of user requests for the current content segment within the measurement window, and the average number of requests for all content segments within the same session measurement window; Calculation rules: If the number of requests for the segment is less than or equal to the average number of requests in the session, it will be counted as 100% of the effective measurement; if the number of requests exceeds the average number of requests in the session, it will be reduced in a stepwise manner according to the multiple of the excess. For each order of magnitude increase in the multiple of the excess, the calculation ratio will be reduced by 30%, with a minimum calculation ratio of no less than 20%, to avoid excessive discounting of popular content. Second dimension: Cross-node replication count calculation (eliminating invalid bandwidth usage from redundant replication). Statistical definition: The total number of cross-node replication and forwarding steps that the content fragment takes from the original storage node to the edge node that finally delivers it to the user; Calculation rules: No cross-node replication (content is sent directly from the storage node to the user) will be counted as 100% of the effective metering; for each additional cross-node replication, the calculation ratio will be reduced by 15%, with a minimum calculation ratio of no less than 30%, to avoid meaningless multi-node repeated transmissions being fully charged. The third dimension: Origin back depth calculation (matching the actual bandwidth cost of cross-level origin back connections). Statistical definition: The number of network layers between the edge node where the user request is received and the node where the content is finally stored is counted. The edge node's local cache hit is layer 0. The layer number is increased by 1 for each request to the origin server sent to the next higher level node, with the maximum layer being the one that reaches the original origin server. Calculation rules: Local cache hits at edge nodes are counted as valid metering at a rate of 100%; for each additional layer of origin pull depth, the bandwidth metering weight increases by 50%, in order to match the additional bandwidth costs and scheduling overhead incurred by cross-layer origin pulls; The final determination of the effective distribution volume is based on the number of original transmission bytes generated during the execution of the chain element. First, it is adjusted by attenuation according to the conversion ratio of request frequency and cross-node replication number, and then adjusted by cost upward according to the weight of back-to-origin depth. The final value is the effective distribution volume corresponding to the chain element, which is directly used for bandwidth dimension measurement of resource consumption density scoring.

[0025] Step 4: Perform joint routing and time-based cross-processing based on dual-dimensional scoring.

[0026] Based on the logical traction system score and resource consumption density score, determine where the chain element should be executed, through which path it should be distributed, and how the closed-loop subgraph should converge, and output an execution plan with a settlement package number.

[0027] The joint routing executor first filters candidate nodes, retaining only those that simultaneously meet privacy level, protocol compatibility, content format support, geographical domain constraints, and current load constraints; then, it calculates the comprehensive execution cost of each chain element on different candidate nodes and selects the optimal allocation result. The joint route target is determined by the following formula: ; in, Assign a mapping to the optimal node; Represents chain element The assigned candidate node; The chain element weight vector must contain at least the logical constraint score. Resource consumption density score and target rigidity coefficient; For chain element At the node The cost vector on the surface includes at least the inverse of semantic processing capability, the inverse of distribution capability, and end-to-end latency estimation. For a directed cyclic deterministic graph, there is a set of chain elements. For non-closed-loop parts, the system executes directly in the order of dependencies; for closed-loop parts, the system first identifies strongly connected subgraphs and assigns chain elements in the same strongly connected subgraph to the same settlement package. The closed-loop execution adopts a time-interleaved method: forward chain elements are executed in the current micro-time window, and feedback chain elements are executed in the next adjacent micro-time window. The two share the same settlement package identifier and the same convergence context. To avoid double billing for the closed loop at each feedback, only the first round of forward output and the final convergence output of the closed loop are billed as the main billing objects. Intermediate rounds only accumulate correction costs and do not form new accounts separately. Whether the closed loop converges is determined by the following ratio: ; in, For the closed loop in the first... The rate of semantic change after round feedback; For the first The semantic summary vector output by the round; To avoid a protection constant where the denominator is zero; when When the latency falls below a preset threshold, the closed loop is considered to have converged, and further feedback is stopped. The threshold is determined jointly based on the upper limit of latency and the lower limit of quality in the service level constraints. After this step is completed, the allocation relationship from chain elements to physical nodes, cross-node distribution paths, round information of closed-loop subgraphs, and settlement package identifiers are obtained. The original resource metering and output summary generated during the execution process are written to the execution meter collector in real time.

[0028] Step 5: Logical fingerprint generation and storage on the minimum chain of evidence.

[0029] The purpose of converting the execution result into verifiable evidence that it was indeed generated by the chain element and has not been tampered with is to form a low-overhead, verifiable logical fingerprint.

[0030] After any chain element completes execution, the logical fingerprint generator extracts the following fields: source semantic description summary, session identifier, time window identifier, execution node identifier, output content summary, and execution bill count snapshot; Logical fingerprints are then generated in the following manner: ; in, For chain element Logical fingerprint; It is a one-way hash function; Input a summary value of the semantic description into the chain element; To set the execution window identifier; For session identification; For execution node identification; Output a summary value of the content for the chain element; A snapshot of the bill count at the end of the chain element execution; This is a field concatenation symbol, representing the concatenation of the preceding and following fields in a fixed order into a complete string before inputting it into a hash function; The on-chain evidence storage interface only Chain element identifier, settlement package identifier, predecessor fingerprint index and timestamp are written to the consortium chain or evidence ledger, while the original content, original logs and original resource samples are retained locally by each participant. When a settlement dispute exists, the disputing parties only need to submit the local original record and recalculate the logical fingerprint to compare it with the on-chain fingerprint, thereby confirming whether the content source, execution node and bill count are consistent. For closed-loop settlement packages, the system will also generate a root fingerprint from the fingerprints of each chain element within the same settlement package in a fixed order, and use the root fingerprint as the unique external index of the settlement package. This will maintain the traceability of the entire package without exposing the intermediate content.

[0031] Step Six: Automatic merging of related tasks, negotiation and settlement among multiple agents, and revenue sharing through smart contracts.

[0032] The final bill is generated based on verified chain element relationships, execution contributions, and distribution contributions, and the cost allocation is automatically completed among multiple participants. First, it is determined which chain elements should be merged into the same billing object, and then the share of each party in the billing object is calculated.

[0033] First, automatic merging is performed on chain elements in the graph that have strong input-output dependencies, and then the preceding chain elements are merged. With subsequent chain elements Calculate its dependency strength: ; in, For chain element chain element The strength of dependence; For chain element The set of key semantic slots in the output; For chain element The set of input semantic slots necessary for execution; when When the logical fingerprint continuity verification of the two chain elements is not less than the preset threshold, the two are merged into the same settlement package. This process is used to avoid splitting the preceding computing power fee and the subsequent distribution fee into two independent bills when they are essentially from the same output chain. After the settlement package is formed, the multi-agent negotiation settlement network begins operation. The content provider agent, computing power provider agent, content distribution provider agent, and platform agent each read the following evidence from the same settlement package: logical fingerprint, execution node proof, normalized resource density, path occupancy ratio, service quality completion rate, and guarantee responsibility ratio. The negotiation controller adopts a limited round asynchronous negotiation method. In each round, each agent submits the bid boundary and the acceptable share range. The controller removes bids with incomplete evidence or discontinuous fingerprints, and then recalculates the sharing coefficient of each party according to the contribution. If the sharing coefficient changes less than the set threshold for two consecutive rounds, or reaches the maximum number of rounds, the settlement result of the settlement package is fixed. The final settlement amount for each participating party is obtained using the following formula: ; in, As a participant The final settlement amount; This is the collection of all settlement packages; For settlement package The set of entities participating in the settlement; For settlement package Main body The effective contribution coefficient is calculated by the multi-agent negotiation controller based on verified computing power contribution, distribution contribution, guarantee responsibility, and path occupancy ratio, as follows: The effective contribution coefficient is determined based on the comprehensive contribution of each participant in the settlement package to four core categories. The default weights for the four categories of contributions are: computing power contribution 40%, distribution contribution 40%, service guarantee responsibility 10%, and path scheduling contribution 10%, with a total weight of 100%. This weight can be fine-tuned according to the service type. The specific quantification rules are as follows: Computing power contribution: determined by the proportion of the sum of the cumulative resource consumption density scores of all chain elements executed by the participant within the settlement package to the total resource consumption density scores of all chain elements within the settlement package; Distribution contribution: Determined by the proportion of the cumulative effective distribution amount of all distribution chain elements completed by the participant in the settlement package to the total effective distribution amount in the settlement package; Service Guarantee Responsibility: Quantified according to the compliance rate of the corresponding service level SLA of the participating party. A compliance rate of 100% is calculated as full marks. For each SLA breach, the corresponding contribution percentage will be reduced proportionally to the impact of the breach. Path scheduling contribution: Quantified according to the level and scheduling role of participating nodes in the distribution path. The contribution ratio of core source nodes and back-to-source nodes is higher than that of edge cache nodes, and the contribution ratio of cross-region scheduling nodes is higher than that of nodes within the same region. Once the settlement package is formed, the multi-agent negotiation and settlement network uses the on-chain stored logical fingerprint and execution evidence as the sole immutable basis, and executes according to the following fixed process: 1. Negotiation Initialization: The negotiation controller synchronizes all evidence stored in the settlement package with the agents of all participants in the settlement package (content providers, computing power providers, distribution providers, and platform providers), including chain element logical fingerprints, two-dimensional scoring results, effective distribution volume, SLA compliance status, and path occupancy records. 2. Asynchronous negotiation execution: The asynchronous negotiation mode with a maximum of 5 rounds is adopted. In each round, each participating party's agent submits its own acceptable contribution coefficient range and corresponding supporting evidence based on the synchronously stored evidence. 3. Invalid proposal removal: The negotiation controller verifies the proposals in each round and removes invalid proposals that lack corresponding evidence or exceed the total contribution boundary of the settlement package. 4. Negotiation convergence determination: Based on valid proposals and the above-mentioned quantitative rules, the controller calculates the baseline contribution coefficient of each participant; when the fluctuation range of the coefficient calculated in two consecutive rounds does not exceed 1%, or when the maximum number of negotiation rounds of 5 is reached, the negotiation terminates and the final valid contribution coefficient is solidified. For settlement package The settlement value is jointly mapped by the logical entanglement mechanism, resource consumption density, convergence rounds, and service quality completion of the chain elements within the settlement package; After receiving the settlement package identifier, logical fingerprint root, effective contribution coefficient and settlement amount, the smart contract executor automatically completes the freezing, allocation, accounting and billing on the blockchain. If supplementary evidence or dispute recalculation occurs later, the smart contract will transfer the original bill to the pending verification state and generate a new bill version with the original logical fingerprint as the index, without overwriting the existing bill record.

[0034] Step 7: Bill Output and Anomaly Correction.

[0035] The execution evidence and settlement results are integrated into an auditable invoice, and corrections are made when there are evidence conflicts, non-convergence of the closed loop, or node replacement, so as to ensure that the final invoice is both readable and traceable.

[0036] When outputting the bill, at least the following items should be generated: session identifier, chain element identifier, settlement package identifier, logical control score, resource consumption density score, execution node identifier, distribution path summary, logical fingerprint, amount allocated to each participant, and bill version number. If any of the following situations are detected, the anomaly correction process will be triggered: discontinuous chain element fingerprints, dependency strength reaching the consolidation threshold but not being consolidated, closed-loop rounds exceeding the upper limit, execution nodes being replaced midway, and inconsistencies between the bill count snapshot and resource sampling. The anomaly correction process includes: freezing the current settlement package, retrieving the local original records of the relevant chain elements, recalculating the logical fingerprint, re-executing the dependency strength determination, and, if necessary, regrouping the settlement package and triggering multi-agent negotiation again. During the correction period, the bill status remains queryable but cannot be transferred until the new version of the bill is confirmed by the smart contract.

[0037] Through the detailed description of the above embodiments, the billing and settlement method for the integrated computing power and content distribution service of the present invention transforms the billing object from traditional resource input items into the smallest independent semantic chain element and its settlement package, and expresses the real closed-loop relationship in the integrated service through a directed cyclic billing graph; furthermore, through joint modeling of logical constraint scoring and resource consumption density scoring, it realizes the transformation from simply calculating how many resources were used to whether the resources support effective task output; at the same time, by leveraging logical fingerprints, on-chain notarization, automatic reconciliation of associated tasks, and multi-agent negotiation settlement, it completes automatic allocation and reliable settlement among computing power service providers, content distribution providers, and platform providers. Therefore, this solution not only improves the granularity, fairness, and transparency of billing, but also reduces problems such as duplicate billing of closed-loop tasks, difficulties in cross-entity reconciliation, and inconsistencies in the metering standards of heterogeneous nodes.

[0038] The preset parameters in the above formulas shall be set by those skilled in the art according to the actual situation.

[0039] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0040] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0041] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0042] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0043] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0044] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0045] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A billing and settlement method for a service integrating computing power and content distribution, characterized in that, The method flow is as follows: S1. Normalization of fused service requests and generation of directed cyclic design graph: Receive fused service requests and complete normalization processing. Decompose the normalized requests into several minimum independent semantic chain elements and construct a directed cyclic design graph containing forward dependencies, control dependencies and feedback dependencies. S2. Logical constraint score calculation: For each chain element in the directed cyclic graph, calculate the logical constraint score that represents the semantic complexity of the task and the strength of the constraints. S3. Resource consumption density score calculation: Perform heterogeneous hardware normalization processing on each chain element, calculate the corresponding resource consumption density score, and form a two-dimensional score result. S4. Joint Routing and Closed-Loop Execution: Based on the two-dimensional scoring results, complete the joint routing allocation and distribution path planning of the chain elements, execute the closed-loop subgraph in the graph using the time-interval method, and determine the closed-loop convergence status. S5. Logical fingerprint generation and on-chain evidence storage: Generate a unique logical fingerprint for the completed chain element and complete the on-chain evidence storage of the minimum evidence chain. S6. Automatic merging and smart contract revenue sharing: Automatic merging is completed based on the dependency relationship between chain elements to form a settlement package. The contribution coefficient of each participant is determined through multi-agent negotiation, and the smart contract is triggered to complete the revenue sharing operation. S7. Bill Output and Anomaly Correction: Output standardized and traceable bills, and trigger corresponding correction processes for settlement anomaly scenarios.

2. The billing and settlement method for a computing power and content distribution integrated service according to claim 1, characterized in that, When constructing a directed and cyclic fixed graph, the target elements in the fusion service request are first extracted to form a target skeleton, and then expanded into a sequence of processing actions and a sequence of distributing actions. The disassembly and reassembly are completed in units of the smallest independent semantic chain element. Feedback edges are retained for feedback dependencies that meet preset conditions, forming a directed graph that allows closed loops. Each node in the graph is bound to a fixed metric window.

3. The billing and settlement method for a computing power and content distribution integrated service according to claim 2, characterized in that, Based on the independence of the input and output semantic slots of the chain element, the feasibility of separate billing of the execution result, and the subsequent consumption attributes of the output content, the chain element is split or merged; for feedback edges that do not meet the retention conditions, boundary cache nodes are inserted and rewritten as cross-time window dependencies to avoid scheduling loops.

4. The billing and settlement method for a computing power and content distribution integrated service according to claim 1, characterized in that, When calculating the logical constraint score, multi-dimensional logical features corresponding to the chain element are extracted from the current session, and the execution uncertainty obtained from the real-time running status of the session is combined to quantify the logical value and constraint strength of the chain element. When calculating the resource consumption density score, the normalized parameters obtained from the node standard reference test are used to map the original resource measurement collected during the execution of the chain element to a unified scale, thereby completing the normalization and quantification of resource consumption.

5. The billing and settlement method for a computing power and content distribution integrated service according to claim 4, characterized in that, In the process of calculating the resource consumption density score, the bandwidth usage of chain elements in the content distribution scenario is measured based on the request frequency, cross-node replication times, and origin-following depth of content fragments within the same session to obtain the effective distribution volume, which is used to complete the bandwidth-related measurement in the resource consumption density score.

6. The billing and settlement method for a computing power and content distribution integrated service according to claim 1, characterized in that, For strongly connected closed-loop subgraphs in a directed cyclic fixed-calculation graph, all chain elements within the subgraph are assigned to the same settlement package. The closed-loop execution adopts a time-interleaved method, with forward chain elements and feedback chain elements being executed in adjacent micro-time windows and sharing the same settlement context. Only the first round of forward output and the final convergence output of the closed loop are subject to main billing, while intermediate rounds only accumulate correction costs without generating separate accounts.

7. The billing and settlement method for a computing power and content distribution integrated service according to claim 6, characterized in that, The convergence of the closed loop is determined by the semantic change rate of the output content before and after the closed loop execution. The convergence threshold is determined by the upper limit of latency and the lower limit of quality in the service level constraints. When performing joint route allocation, candidate nodes that meet the constraints are first screened, and then the comprehensive execution cost is calculated based on the two-dimensional scoring results to select the optimal node allocation result.

8. The billing and settlement method for a computing power and content distribution integrated service according to claim 1, characterized in that, After the chain element is executed, a unique logical fingerprint is generated based on the chain element's input semantic description, execution context information, output content summary, and bill count snapshot. Only the logical fingerprint and its corresponding associated identification information are written to the evidence ledger, while the original execution data is stored locally on each participant's local machine. For the closed-loop settlement package, a unique root fingerprint is generated based on the fingerprints of each chain element within the package as the external index of the settlement package.

9. The billing and settlement method for a computing power and content distribution integrated service according to claim 1, characterized in that, Automatic consolidation determination is completed based on the semantic slot dependency strength between chain elements, and related chain elements that meet the conditions are merged into the same settlement package. After the settlement package is formed, the agents of each participating party complete a limited number of asynchronous negotiations based on the evidence stored in the settlement package to determine the effective contribution coefficient and settlement amount of each participating party, and trigger the smart contract to complete the fund transfer and distribution. In response to abnormal scenarios during the settlement process, the corresponding settlement package is frozen and a correction process is triggered to complete the recalculation and update of the bill.

Images