Method and system for source configuration and price decoupling based on dynamic pricing mechanism

Abstract

The application relates to the field of electric digital data processing, and discloses a method and system for decoupling of a signal source configuration and a price based on a dynamic pricing mechanism, which comprises the following steps: defining a time recovery entropy for a signal source, the time recovery entropy representing a time required for recovery to an ideal idle state; constructing a virtual opportunity cost field based on the entropy value; abstracting a service request as a logical particle propagating in the field, and determining a propagation path and a terminal signal source of the logical particle according to the principle of minimum action; and generating a final price according to the path attribute. Through the coupling mechanism of the time recovery entropy and the opportunity cost field, the application enables the future flexibility value to be considered in the resource configuration decision, realizes adaptive decoupling of a price signal and signal source selection, and thus improves the ability of the system to cope with sudden high-value tasks.

Description

Technical Field

[0001] This invention relates to a method and system for decoupling information source configuration and price based on a dynamic pricing mechanism, belonging to the field of electronic digital data processing technology. Background Technology

[0002] In the field of digital data processing, dynamic pricing and information source allocation mainly rely on real-time load indicators (such as CPU utilization and network traffic) to construct a direct price-resource mapping mechanism. This approach meets immediate needs through reactive adjustment, but generally ignores the opportunity cost dimension implicit in resource allocation—that is, the strategic flexibility lost in responding to high-value tasks after information sources are occupied by low-value tasks. Taking cloud platforms simultaneously processing latency-sensitive AI training and batch data processing as an example, existing technologies prioritize expanding low-priced resources under high load, resulting in high-flexibility nodes being deeply occupied by low-value tasks. When a sudden high-value task request occurs, the system is forced to delay response or migrate at high cost due to a lack of available resources, exposing the rigid coupling defect between resource allocation and price signals.

[0003] The industry has attempted to introduce predictive models to optimize scheduling, but these efforts are still limited by three fundamental contradictions: 1. Price signals only reflect historical / current supply and demand and cannot quantify the time cost of restoring the information source to a flexible state; 2. Multi-source monitoring data is simplified into a single-dimensional price instruction, severing the connection between resource selection and opportunity cost; 3. The prediction algorithms and feedback loops added to compensate for the above defects exacerbate system latency and computing power consumption.

[0004] While existing methods have improved resource utilization efficiency to some extent, they still cannot effectively measure the responsiveness of information sources in future tasks through price mechanisms. This results in a lack of strategic consideration for potentially high-value tasks during resource scheduling, and the over-reliance on real-time monitoring and complex calculations makes it difficult to adapt to the large-scale deployment needs of edge computing nodes. Therefore, how to construct an information source allocation mechanism based on endogenous assessment of opportunity cost, and achieve adaptive decoupling of price signals and resource allocation while avoiding high-frequency monitoring and complex calculations, is the technical problem that this invention aims to solve. Summary of the Invention

[0005] This invention provides a method and system for decoupling information source allocation and price based on a dynamic pricing mechanism. Its main purpose is to solve the problem that the lack of opportunity cost assessment in existing technologies leads to rigid coupling between resource allocation and price signals, which in turn restricts the strategic response capability of the system.

[0006] To achieve the above objectives, this invention provides a method for decoupling information source configuration and price based on a dynamic pricing mechanism, the method comprising the following steps:

[0007] Step a: For each source in the source pool, determine a Time Recovery Entropy (TRE). The Time Recovery Entropy is a scalar value used to characterize the time required for a source to be released from its current service state and recover to an idle state capable of accepting any type of new task. The determination of the Time Recovery Entropy is based on the estimated state recovery time parameter of the task currently carried by the source, and the value of the Time Recovery Entropy is only updated infrequently when the type of task carried by the source changes.

[0008] Step b: Based on the time recovery entropy value of the information source, a dynamic virtual opportunity cost field is constructed in the cloud controller, where the time recovery entropy value determines the potential energy corresponding to the information source in the opportunity cost field;

[0009] Step c: When a service request and its initial quote are received, the service request is abstracted into a logical particle. The initial quote carried by the logical particle serves as its initial kinetic energy and propagates in the opportunity cost field.

[0010] Step d: Determine the final propagation path of the logical particle in the opportunity cost field according to the principle of least action. The principle of least action minimizes the total action of the final propagation path. The total action includes the physical distance action and the potential energy change action.

[0011] Step e involves determining the information source corresponding to the endpoint of the final propagation path as the configuration target of the service request, and determining the final service price based on the kinetic energy consumed by the logical particle along the final propagation path, thereby decoupling the information source configuration from the price.

[0012] Preferably, the determination of time recovery entropy in step a is based on the estimated recovery time contained in the metadata of the task currently carried by the information source; the estimated recovery time is a predefined value that can indicate the reload time or context switching time after cache failure, and the update frequency of the time recovery entropy value is less than once per minute.

[0013] Preferably, the construction of the dynamic virtual machine opportunity cost field in step b includes: the cloud controller periodically collecting the time recovery entropy values ​​of all information sources in the information source pool, and using an inverse distance weighted interpolation algorithm to generate discretized potential field data on the logical grid based on the time recovery entropy values ​​of the information sources and their logical positions to form the opportunity cost field.

[0014] Preferably, the propagation rule of logical particles in the opportunity cost field in step c is as follows: logical particles tend to move towards regions with lower potential energy in the opportunity cost field; when logical particles pass through regions with higher potential energy during propagation, their kinetic energy is consumed.

[0015] Preferably, the final propagation path following the principle of least action in step d is found using a variant of the A* algorithm (A* algorithm), where the cost function of the A* algorithm is... Integrating physical distance cost Cost of potential energy change Their relationship can be expressed as:

[0016] ,

[0017] in, To adjust the preset coefficients for the relative weights of physical distance cost and potential energy change cost, This represents the total change in potential energy experienced by the logical particle along the path.

[0018] Preferred, This is a preset coefficient, with a value range of 0.1 to 100, and the value can be configured and adjusted according to the system administrator's preference for the system's future responsiveness and short-term benefits.

[0019] Preferably, the determination of the final service price includes combining the kinetic energy consumed by the logical particle along the final propagation path with the initial quote of the service request to obtain the final settlement price.

[0020] Preferably, the information sources in the information source pool include virtual machines, containers, physical servers, and network slice resource units; the types of service requests include data cleaning tasks, AI model training tasks, data analysis tasks, and high real-time business tasks.

[0021] Preferably, the source node maintains only one scalar value, the time recovery entropy, and the update of the scalar value is only triggered when the task type carried by the source changes, thereby reducing the computational and communication overhead of the source node.

[0022] A system for decoupling information source configuration and price based on a dynamic pricing mechanism, the system comprising:

[0023] The source node is configured to determine and maintain the time recovery entropy (TRE). The time recovery entropy is a scalar value used to characterize the time required for the source to be released from its current service state and recover to an idle state that can accept any type of new task. The determination of the time recovery entropy is based on the estimated state recovery time parameter of the task currently carried by the source, and the value of the time recovery entropy is only updated infrequently when the type of task carried by the source changes.

[0024] The cloud controller, communicating with the source node, is configured to: construct a dynamic virtual opportunity cost field based on the time recovery entropy value of the source node, where the time recovery entropy value determines the potential energy corresponding to the source in the opportunity cost field; when a service request and its initial quote are received, the service request is abstracted into a logical particle, and the initial quote carried by the logical particle serves as its initial kinetic energy, which propagates in the opportunity cost field; the final propagation path of the logical particle in the opportunity cost field is determined according to the principle of least action, which minimizes the total action of the final propagation path, including the physical distance action and the potential energy change action; the source corresponding to the endpoint of the final propagation path is determined as the configuration target of the service request, and the final service price is determined based on the kinetic energy consumed by the logical particle along the final propagation path, thereby decoupling the source configuration from the price.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] 1. The source node only needs to update a single scalar value (TRE) when switching tasks. This value comes from the preset recovery time parameter in the task metadata, avoiding the computational power consumption of continuously monitoring multiple dimensions of indicators in traditional solutions. This allows edge nodes to run with near-zero computational overhead. When the cloud controller constructs an opportunity cost field based on discrete TRE values, the system naturally forms a quantitative capability for resource strategic elasticity. High TRE sources form high potential energy zones in the virtual potential field, while low TRE sources form low potential energy channels. This design makes resource allocation decisions no longer dependent on complex prediction models, but spontaneously guides service requests to bypass highly sticky resources through the principle of physical fields, reserving agile response space for sudden high-value tasks.

[0027] 2. The initial quote carried in the service request is transformed into the initial kinetic energy of the logical particle, which propagates in the potential field constructed by the TRE value. The process of the particle consuming kinetic energy as it traverses the high potential energy region is essentially a process of transforming the opportunity cost of resources into a price adjustment factor. When the particle reaches the destination according to the principle of minimum action, its path endpoint automatically points to the source of information with the optimal overall cost, and the amount of kinetic energy consumed directly corresponds to the final service price. This mechanism integrates the two originally separate decision-making processes of resource selection and price generation into a natural motion process in a single physical field, avoiding the rigid coupling between price signals and resource allocation in traditional schemes.

[0028] 3. The cloud controller uses inverse distance weighted interpolation, which only requires low-frequency TRE broadcasts from the source node to generate a global opportunity cost field. When a service request arrives, the A* algorithm variant uses a composite cost function that integrates physical distance cost and potential energy change cost. The system quickly outputs the optimal path on the logical grid. This process transforms the opportunity cost assessment in economics into a standard path planning problem, enabling the system to simultaneously complete resource matching and dynamic price adjustment without deploying a dedicated pricing model, thereby reducing the decision-making complexity of the central node.

[0029] 4. The value of Time Recovery Entropy (TRE) depends only on the task's own state recovery characteristics, such as cache loading time and context switching time, and is unrelated to the specific business type. When the system connects to new information sources, such as network slicing or carrying unknown tasks, it can be integrated into the existing scheduling framework simply by presetting TRE parameters during the task deployment phase. This design, which abstracts business characteristics into standardized metrics, enables the solution to maintain the consistency of resource allocation strategies through a unified potential field mechanism when facing heterogeneous loads such as AI training and real-time services. Furthermore, logical particles continuously consume kinetic energy when passing through high potential energy zones during propagation. This means that when a service request attempts to occupy high TRE resources, its effective bid will be automatically decayed. This mechanism forms a built-in price damping effect: competitive demand for scarce and flexible resources will be self-suppressed due to kinetic energy consumption, avoiding the pulse-like price fluctuations caused by traditional bidding models. The system achieves market equilibrium through physical principles rather than external regulation, thereby reducing the risk of frequent resource allocation switching caused by sudden price changes. Attached Figure Description

[0030] Figure 1 This is a flowchart illustrating the source configuration and dynamic pricing process of this invention.

[0031] Figure 2 A comparison chart of pricing responses for service requests related to this invention;

[0032] Figure 3 This is a timing diagram of the high-value request response of the present invention.

[0033] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0034] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0035] This application provides a method and system for decoupling information source configuration and price based on a dynamic pricing mechanism. The method includes the following steps:

[0036] Step a: For each source in the source pool, determine a Time Recovery Entropy (TRE). The Time Recovery Entropy is a scalar value used to characterize the time required for a source to be released from its current service state and recover to an idle state capable of accepting any type of new task. The determination of the Time Recovery Entropy is based on the estimated state recovery time parameter of the task currently carried by the source, and the value of the Time Recovery Entropy is only updated infrequently when the type of task carried by the source changes.

[0037] Step b: Based on the time recovery entropy value of the information source, a dynamic virtual opportunity cost field is constructed in the cloud controller, where the time recovery entropy value determines the potential energy corresponding to the information source in the opportunity cost field;

[0038] Step c: When a service request and its initial quote are received, the service request is abstracted into a logical particle. The initial quote carried by the logical particle serves as its initial kinetic energy and propagates in the opportunity cost field.

[0039] Step d: Determine the final propagation path of the logical particle in the opportunity cost field according to the principle of least action. The principle of least action minimizes the total action of the final propagation path. The total action includes the physical distance action and the potential energy change action.

[0040] Step e involves determining the information source corresponding to the endpoint of the final propagation path as the configuration target of the service request, and determining the final service price based on the kinetic energy consumed by the logical particle along the final propagation path, thereby decoupling the information source configuration from the price.

[0041] Preferably, the determination of time recovery entropy in step a is based on the estimated recovery time contained in the metadata of the task currently carried by the information source; the estimated recovery time is a predefined value that can indicate the reload time or context switching time after cache failure, and the update frequency of the time recovery entropy value is less than once per minute.

[0042] Preferably, the construction of the dynamic virtual machine opportunity cost field in step b includes: the cloud controller periodically collecting the time recovery entropy values ​​of all information sources in the information source pool, and using an inverse distance weighted interpolation algorithm to generate discretized potential field data on the logical grid based on the time recovery entropy values ​​of the information sources and their logical positions to form the opportunity cost field.

[0043] Preferably, the propagation rule of logical particles in the opportunity cost field in step c is as follows: logical particles tend to move towards regions with lower potential energy in the opportunity cost field; when logical particles pass through regions with higher potential energy during propagation, their kinetic energy is consumed.

[0044] Preferably, the final propagation path following the principle of least action in step d is found using a variant of the A* algorithm, where the cost function of the A* algorithm is... Integrating physical distance cost Cost of potential energy change Their relationship can be expressed as:

[0045] ,

[0046] in, To adjust the preset coefficients for the relative weights of physical distance cost and potential energy change cost, This represents the total change in potential energy experienced by the logical particle along the path.

[0047] Preferred, This is a preset coefficient, with a value range of 0.1 to 100, and the value can be configured and adjusted according to the system administrator's preference for the system's future responsiveness and short-term benefits.

[0048] Preferably, the determination of the final service price includes combining the kinetic energy consumed by the logical particle along the final propagation path with the initial quote of the service request to obtain the final settlement price.

[0049] Preferably, the information sources in the information source pool include virtual machines, containers, physical servers, and network slice resource units; the types of service requests include data cleaning tasks, AI model training tasks, data analysis tasks, and high real-time business tasks.

[0050] Preferably, the source node maintains only one scalar value, the time recovery entropy, and the update of the scalar value is only triggered when the task type carried by the source changes, thereby reducing the computational and communication overhead of the source node.

[0051] A system for decoupling information source configuration and price based on a dynamic pricing mechanism, the system comprising:

[0052] The source node is configured to determine and maintain the time recovery entropy (TRE). The time recovery entropy is a scalar value used to characterize the time required for the source to be released from its current service state and recover to an idle state that can accept any type of new task. The determination of the time recovery entropy is based on the estimated state recovery time parameter of the task currently carried by the source, and the value of the time recovery entropy is only updated infrequently when the type of task carried by the source changes.

[0053] The cloud controller, communicating with the source node, is configured to: construct a dynamic virtual opportunity cost field based on the time recovery entropy value of the source node, where the time recovery entropy value determines the potential energy corresponding to the source in the opportunity cost field; when a service request and its initial quote are received, the service request is abstracted into a logical particle, and the initial quote carried by the logical particle serves as its initial kinetic energy, which propagates in the opportunity cost field; the final propagation path of the logical particle in the opportunity cost field is determined according to the principle of least action, which minimizes the total action of the final propagation path, including the physical distance action and the potential energy change action; the source corresponding to the endpoint of the final propagation path is determined as the configuration target of the service request, and the final service price is determined based on the kinetic energy consumed by the logical particle along the final propagation path, thereby decoupling the source configuration from the price.

[0054] In a specific implementation of this invention, the propagation process of logical particles refers to the existence of explicit discrete scheduling behaviors and operation paths within the system. Once a service request is generated, it is assigned a resource allocation intention with an initial budget weight. This intention is transformed into a series of scheduling attempts in the controller logic. The controller calculates the total cost of the service request moving from its current location to each candidate information source based on the current opportunity cost field state. The cost mainly includes resource state update delay, scheduling channel queuing length estimation, and system load trend feedback. Each scheduling attempt is equivalent to a logical particle jumping once in a certain direction. The path planning process is actually a finite-step search performed by the controller within the logical coordinate grid, and the path selection is based on the target cost. To achieve optimal control of path selection, this invention introduces the principle of minimum action path into the scheduling engine. The calculation of action mainly refers to the reachability cost of the information source on the request propagation path and the resource scarcity level difference of the logical distance traversed. This information comes from the statistical average of the historical response delay of each information source and the resource recovery capability prediction table. Without relying on continuous mathematical formulas, the system maintains a discrete resource response cost table and a set of movement offset rules based on resource level gradients. This ensures that each path jump has clear selection logic and execution conditions, and completes the propagation decision of shortest path and optimal resources without relying on an external physical simulation system.

[0055] Example 1: This invention provides a method for decoupling information source configuration and price based on a dynamic pricing mechanism. It optimizes resource allocation decisions and decouples information source configuration from service prices by coupling the Time Recovery Entropy (TRE) mechanism with an opportunity cost field. The method mainly includes the following steps: determining the Time Recovery Entropy (TRE) for each information source in the information source pool, which represents the time required for the information source to recover from its current task state to an idle state; constructing a dynamic virtual opportunity cost field in the cloud controller based on the TRE; abstracting service requests into logical particles and giving them initial kinetic energy according to the initial price, propagating them in the opportunity cost field; determining the propagation path according to the principle of least action, minimizing the action required during propagation; determining the final information source configuration and determining the final service price through the kinetic energy consumption of the logical particles. Time recovery entropy (TRE) is a scalar value characterizing the time required for a source to recover from its current service state to an idle state capable of undertaking new tasks. This value is calculated based on the recovery time parameter of the source's current task and is updated when the source's task type changes. The main function of this feature is to provide a method for quantifying the flexibility of a source. This metric describes the recovery time required for a source to go from its current occupied state to an idle state, thereby determining the potential energy of the source in the dynamic virtual machine meeting cost field. Specifically, the time recovery entropy is determined based on the estimated recovery time of the source's current task, such as cache reloading and context switching time. The accuracy of this value is maintained by periodically updating the source's state through the system. The update frequency of this value is usually set to less than once per minute to reduce computational and communication overhead.

[0056] In the cloud controller, a dynamic virtual opportunity cost field is constructed based on the time recovery entropy values ​​of information sources. This field guides the selection of information sources by reflecting the corresponding potential energy of the recovery entropy of each information source in the virtual potential field. Specifically, the system periodically collects the time recovery entropy values ​​of all information sources in the information source pool and uses an inverse distance weighted interpolation algorithm to generate discretized potential field data on the logical grid, thereby forming an opportunity cost field. The design of this potential field reflects the opportunity cost of resources, so that information sources with high recovery entropy form high potential energy regions, and information sources with low recovery entropy form low potential energy regions, thus providing flexibility in resource selection. Service requests are abstracted as [data points] within the opportunity cost field. The propagating logical particle uses the initial price as its initial kinetic energy. During particle propagation, it follows the principle of least action to select the path that minimizes the propagation path. The total action of this path includes the physical distance cost and the potential energy change cost. Specifically, logical particles tend to move towards regions with lower potential energy and consume their kinetic energy in regions with higher potential energy. Through a variant of the A* algorithm, the system combines the cost functions of physical distance cost and potential energy change cost to quickly determine the optimal path. Finally, when the particle reaches its destination, the information source corresponding to that destination is determined as the configuration target of the service request, and the kinetic energy consumption directly corresponds to the final service price. The principle of least action is applied to particle path propagation, causing particles to move from the initial point to the destination along the path of least action. This process combines the opportunity cost of resources with the price adjustment factor, avoiding the rigid coupling between traditional pricing and resource allocation. Specifically, a variant of the A* algorithm is used to determine the path of least action. The cost function of this path integrates the physical distance cost and the potential energy change cost, and the preset coefficients in the cost function adjust the relative weights of the physical distance cost and the potential energy change cost. This method determines the final service price by calculating the kinetic energy consumed by the particle along the path, thus naturally integrating the decision-making process of resource allocation and price generation.

[0057] Example 2: In this example, to improve the dynamic response capability of resources and reduce the risk of high-flexibility information sources being occupied by low-value tasks, the system adopts the following five stages to decouple information source configuration and price decision-making: First, the system sets and maintains a scalar value called time recovery entropy on the information source node side, which is used to quantify the time required for the information source node to recover from the current task state to the idle state. The setting of this value is based on the recovery time parameters preset in the metadata attached to the task, mainly involving indicators such as cache reload time and context switching time. After normalization, this value is represented as a unified time scale and is updated infrequently when the task type changes, thereby ensuring its stability and practical maintainability at the system level and avoiding communication and computing overhead caused by frequent state fluctuations; Second, the cloud controller periodically collects the time recovery entropy values ​​of each node in the information source pool and constructs a dynamic virtual machine meeting cost field on the logical topology mapping. This process is based on a regular logical grid, mapping the logical position of the information source and its recovery entropy value as potential energy weights, and through The inverse distance weighted interpolation algorithm generates a discretized potential energy field data structure across the entire grid. The distribution characteristics of this potential energy field reflect the opportunity cost levels corresponding to different information sources. The information source regions corresponding to high time recovery entropy values ​​form high potential energy regions, while the regions corresponding to low time recovery entropy values ​​form low potential energy channels, providing a quantitative basis for subsequent resource scheduling. Then, upon receiving a service request, the system abstracts the request into a logical particle and uses the initial quote carried by the request as its initial kinetic energy. This logical particle propagates discretely in the potential energy field, and its propagation path is dynamically determined according to the principle of minimum action. Specifically, each time the particle propagates and jumps, the system comprehensively considers the path cost to each neighboring grid point. This cost consists of the basic cost caused by physical distance and the potential energy change experienced during the jump. To control the relative influence of these two types of costs in path decision-making, the system introduces an adjustable coefficient as a weighting factor, the value of which can be configured within a preset range to reflect the platform's preference between resource response flexibility and short-term communication costs.

[0058] During particle propagation, kinetic energy is continuously consumed. Each jump reduces the corresponding path cost until the kinetic energy is exhausted or the maximum allowed number of jumps is reached. To ensure path calculation efficiency, the system uses a variant of the A* algorithm based on a heuristic function to search for paths. The heuristic function is designed with the logical distance between the particle's current position and the low-potential-energy region as an estimation index, thereby seeking the globally optimal solution for the path between resource cost and opportunity value. After the particle propagation is completed, the system uses the source node corresponding to its final landing point as the resource allocation target for the service request, and determines the final settlement price based on the difference between the particle's kinetic energy consumption in the path and its initial bid. Since the high-potential-energy region causes the logical particle's kinetic energy to decay rapidly, the source with a high recovery entropy value forms a natural damping effect in the price mechanism, making it difficult to achieve path arrival when the demand for scarce resources is not supported by high bids. This effectively avoids the irrational occupation of highly flexible resources in traditional solutions. This invention maps task scheduling decisions to particle propagation path calculations and embeds the pricing mechanism into the construction of the physical potential field, achieving adaptive decoupling between information source selection and service pricing. This helps improve the stability and response efficiency of resource scheduling and has the ability to expand to various application scenarios such as edge computing and multi-tenant scenarios.

[0059] Example 3: This example aims to verify the source configuration and price decoupling method based on dynamic pricing mechanism disclosed in this invention, and its effectiveness in improving the system's ability to respond to high-value tasks and optimizing resource allocation efficiency. The experiment was conducted on a cloud-native computing platform, which includes a cloud controller and a group of heterogeneous source nodes. The types of source nodes include: computing source nodes (based on container technology, suitable for short-term high-concurrency tasks), storage source nodes (simulating distributed storage services, focusing on data persistence and access performance), and hybrid source nodes (integrating computing and storage capabilities, with a certain degree of flexibility).

[0060] The parameters are set as follows: Each source in the source pool is configured with Time Recovery Entropy (TRE). TRE is a scalar value representing the time required for a source to recover from its current service state to an idle state capable of handling any type of new task. This value is determined based on the estimated recovery time parameters of the tasks currently being carried by the source, such as cache reload time or context switching time. In this experiment, the recovery time parameter can be reasonably set based on statistical analysis of historical task data and engineering experience for different task types. For example, high real-time business tasks can be set to a lower recovery time parameter (e.g., approximately 50 milliseconds), while AI model training tasks can be set to a lower parameter. The recovery time parameter is set to a relatively high value (e.g., approximately 500 milliseconds). The time recovery entropy value is updated only infrequently when the task type carried by the information source changes, typically less than once per minute, to effectively control computational and communication overhead. The cloud controller periodically collects the time recovery entropy values ​​of all information sources in the information source pool. Through an inverse distance weighted interpolation algorithm, combined with the logical location of the information source nodes, discretized potential field data is generated on the logical grid, forming a dynamic virtual opportunity cost field. The time recovery entropy value of the information source determines its corresponding potential energy in this opportunity cost field. The relative weights of physical distance cost and potential energy change cost are determined by preset coefficients. Adjustments should be made. The value can be configured and adjusted according to the system's weight preference for future responsiveness and short-term benefits. It can be set in the range of 0.1 to 100 according to actual needs to balance the consideration of physical distance and future flexibility. Service requests are abstracted as logical particles carrying initial quotes. The initial quotes serve as the initial kinetic energy for the logical particles to propagate in the opportunity cost field. During propagation, logical particles tend to move towards areas with lower potential energy in the opportunity cost field. When crossing areas with higher potential energy, their kinetic energy is consumed. Logical particles follow the principle of minimum action in the opportunity cost field to determine their final propagation path. The principle of minimum action minimizes the total action of the final propagation path. The total action includes the physical distance action and the potential energy change action. Path search is performed through a variant of the A* algorithm, whose cost function integrates the physical distance cost and the potential energy change cost. Finally, the information source corresponding to the endpoint of the final propagation path is determined as the configuration target of the service request, and the final service price is determined according to the kinetic energy consumed by the logical particles along the final propagation path, thereby achieving decoupling of information source configuration and price.

[0061] The experiment was conducted in two phases: benchmark testing and verification of the method proposed in this invention. In the first phase, a traditional direct mapping mechanism based on real-time load indicators (such as CPU utilization and network traffic) was used to construct price signals. The system selects the information source and generates prices according to the current load. The test scenario simulated an environment with gradually increasing load. Initially, low-value, long-running data cleaning tasks were deployed, and then high-priority, high-real-time-requirement high-value tasks, such as high-real-time business tasks, were gradually introduced. This phase focused on the response latency and resource allocation efficiency of the system in processing high-value tasks under high load. The dynamic pricing mechanism based on time recovery entropy and opportunity cost field proposed in this invention was used in this phase. The test scenario was consistent with the benchmark test to ensure the fairness of the comparison. In this phase, the system's response performance in processing high-value tasks under high load was observed.

[0062] By comparing benchmark test data with data from the verification phase of this invention, the following phenomena and trends were observed: In simulated high-concurrency, high-real-time business scenarios, the system using this invention's method exhibits a noticeable reduction in the average allocation latency of high-value tasks compared to the traditional coupled pricing mechanism under benchmark testing. This improvement is mainly attributed to this invention's quantification of the future flexibility value of information sources through time recovery entropy, and its representation of high-flexibility information sources as high-potential-energy zones in the opportunity cost field. Logical particles (service requests) tend to avoid high-potential-energy zones during propagation, thereby guiding high-value tasks to more suitable resources, reserving critical resources to cope with sudden high-value tasks. Observation results... The results show that after applying the method of this invention, the situation where high-time-recovery-entropy information sources are deeply occupied by low-value tasks during high-load periods is reduced. Analysis of the service request propagation path shows that logical particles carrying higher initial kinetic energy (high priority) can complete the crossing by consuming kinetic energy even when facing some high-potential-energy regions, but their final service price will reflect the cost of occupying high-flexibility resources. On the other hand, logical particles carrying lower initial kinetic energy (low priority) tend to choose low-potential-energy regions, that is, information sources that are currently idle and have low recovery entropy, thereby reducing resource competition with high-value tasks. This dynamic guidance mechanism helps to improve the strategic flexibility of resource scheduling. Experimental results show that this invention, by abstracting service requests into logical particles and propagating them in an opportunity cost field according to the principle of least action, integrates the source selection and price generation processes into a unified physical field motion. The kinetic energy consumed by the logical particles along the path, combined with the initial quote, naturally generates the final service price. This automatic price adjustment creates a price damping effect, which helps to suppress blind competition for scarce and flexible resources, thereby enabling the system to achieve market equilibrium in a dynamic environment. Compared with traditional dynamic pricing methods that require continuous monitoring and frequent updates of multi-dimensional indicators, this invention only requires the source node to update a single scalar value time recovery entropy at low frequency, and the opportunity cost field is periodically constructed by the cloud controller. This helps to reduce the computational overhead and communication burden on the source node, and is especially suitable for the needs of large-scale, edge deployment.

[0063] Example 4: This example combines Figures 1 to 3 This paper describes the implementation of a method and system for decoupling information source configuration and price based on a dynamic pricing mechanism. Figure 1As shown, the system first determines the time required for a source to recover to an idle state using the Time Recovery Entropy (TRE) module, and employs a low-frequency update mechanism to reduce system overhead. Based on this, the system constructs a virtual machine cost field, where TRE determines potential energy, causing sources with high TRE to form high-potential-energy regions and sources with low TRE to form low-potential-energy channels. Simultaneously, inverse distance weighted interpolation is used to achieve spatial continuity of the potential field. Subsequently, service requests are abstracted as logical particles, each carrying an initial quote and serving as initial kinetic energy to enter the virtual machine cost field. These particles then undergo simulated physical motion in the propagation environment. The propagation of these particles follows the principle of minimum action, which is optimized through a variant of the A* algorithm and a composite cost function to minimize the total action of the path. The entire process is handled by a cloud controller that tracks the logical particles, determines the path, and calculates the cost. Based on the movement of the logical particles, the system ultimately completes two tasks: source configuration and dynamic pricing. Source configuration achieves endpoint source selection and adaptive decoupling, while dynamic pricing calculates the price through kinetic energy consumption and reflects a price damping effect, effectively preventing high-value resources from being occupied by low-value tasks.

[0064] like Figure 2 As shown in the figure, the vertical axis represents service requests, including high-value requests A, B, and C, and low-value requests D, E, and F. The horizontal axis represents price (unit: cost), with a scale from 0 to 180. The figure also distinguishes three elements: the type of information source allocated, the price type, and the price. High-TRE information sources are represented by solid lines, low-TRE information sources by dashed lines, initial quotes by solid dots, and final prices by diamond symbols. High-value requests A, B, and C tend to be allocated to high-TRE information sources, and their price curves are solid lines, indicating higher energy consumption based on the initial quote, resulting in a final price much higher than the initial value. Low-value requests D, E, and F are more often allocated to low-TRE information sources, and their price curves are dashed lines, indicating relatively lower energy consumption, with a limited difference in final price compared to the initial quote.

[0065] like Figure 3As shown, firstly, the cloud controller activates logical particles (injecting logical particles carrying initial kinetic energy) and calls the path planning engine (Variant A) to initiate path cost evaluation, i.e., requesting path costs from neighboring nodes. During subsequent path traversal, the logical particles sequentially determine whether they are traversing a high-potential-energy zone or a low-potential-energy zone. The former will return a high path cost and trigger a significant consumption of particle kinetic energy, while the latter will return a low path cost and trigger a slight consumption of particle kinetic energy. After the path traversal is completed, the path planning engine (Variant A) returns the optimal path and total kinetic energy consumption to the cloud controller. Based on this, the cloud controller executes the path scheduling and price generation process according to the path division results. Subsequently, it enters the parallel processing (execution processing) stage. In the price generation module, a request is initiated to calculate the final service price, and based on the calculation of kinetic energy consumption and initial quote, the final service price is output and returned. At the same time, the cloud controller issues a target resource allocation instruction according to the path endpoint, and the target information source confirms that the configuration is successful.

[0066] Example 5: In the edge intelligence platform, the system needs to dynamically respond to a large number of real-time data processing requests from various emergency subsystems, including tasks such as streaming analysis of public safety monitoring videos, road traffic congestion prediction, and emergency event image recognition. The platform is deployed on multiple physical area nodes, each node has a source pool composed of several virtualized computing resources. The source types include container resource nodes, edge microserver nodes, and lightweight computing units. Each source has heterogeneous performance and task adaptability. This example focuses on describing the scheduling process of the system in the scenario of handling an emergency traffic accident image recognition request, demonstrating how the dynamic pricing mechanism based on time recovery entropy driven by the present invention coordinates the efficient operation of the source configuration logic and ensures the natural decoupling of price signals and resource selection.

[0067] When the platform detects a sudden emergency, it automatically generates a high-priority image recognition service request with initial budget information. This request is uploaded to the cloud controller via edge nodes. The controller then abstracts this request into logical particles in real time to initiate the scheduling process. The initial kinetic energy is the kinetic energy scalar mapped by the system to the budget value using a standard conversion factor. The selection of this mapping factor combines the platform's historical scheduling cost range with estimates of the average path consumption for emergency task processing, ensuring that the logical particles have sufficient kinetic energy to traverse some high-potential energy zones without wasting resources. At this time, the status of the information source pool in the platform is reported to the controller at low frequency from the information source node side. The key indicator is the time recovery entropy. This entropy value is obtained based on the type of the current task on the node, combined with a typical estimation table established using historical task state transition latency samples. Specifically, it is calculated as follows: The expected recovery latency value is defined for each task, and the cache release latency and context cleanup time are integrated and normalized according to their weights to form a unified metric. The weighting is 7:3, which is an empirical setting based on the analysis of the system's internal task scheduling backtracking, which found that cache release has a greater impact on the scheduling latency of new tasks in typical container nodes. The controller constructs a virtual machine capital field in the logical space based on the time recovery entropy values ​​held by all source nodes. This field is expressed as a logical grid, with each grid point corresponding to one or more source nodes. Its potential energy value is the average or weighted value of the time recovery entropy of the source nodes within that grid point. To achieve spatial continuity, the controller uses an inverse distance weighted interpolation algorithm to fill areas without direct data points, so that the entire potential field has computable gradient characteristics, thereby guiding the propagation path of logical particles with directional selectivity. During propagation, each time a logical particle jumps one grid, it incurs a propagation path cost, which consists of the physical jump distance and the change in potential energy. The physical jump distance is a constant per unit distance, while the change in potential energy is determined by the potential energy difference between the current and target grid points. The system has an adjustment coefficient to balance the influence of both on the propagation direction decision. This coefficient is preset to 20 in this scenario. The selection of this value was verified by the platform through simulation of a large number of emergency request scheduling historical behaviors. The aim is to suppress the risk of particles rapidly consuming kinetic energy when crossing high potential energy regions in the early stages, thereby retaining greater path flexibility to cope with diverse target selections.

[0068] The path planning of logical particles is performed by a path search module in the controller based on a heuristic function. The heuristic function estimates the logical grid distance between the particle's current position and the lowest potential energy point. The path search algorithm is a simplified variant of the A* algorithm, removing the high dependence on path pruning strategies to accelerate the response of high-priority tasks. This design takes into account that in urgent task scenarios, the priority of search response latency is higher than path optimality guarantee. Platform tests show that after removing the pruning mechanism, the system search latency decreases by about 30%, and the performance loss is within an acceptable range. Finally, the particle path endpoint lands on a lightweight computing node with low time recovery entropy in a certain area. The current task of this node is low-priority log analysis, and its expected cleanup time is extremely short. Therefore, it appears as a low potential energy point when constructing the potential field, and the particle landing point is the service point. Based on the configuration target requested by the task, the controller issues scheduling instructions to the target information source. At the same time, it calculates the total kinetic energy loss of the particles along the path based on the total potential energy increase and physical distance. The system subtracts the loss value from the remaining kinetic energy under the original price mapping relationship to obtain the actual price output, which is the service fee. This price is recorded by the system and used for perceptual modeling of price fluctuation trends in subsequent scheduling behavior. It should be noted that, in order to enhance the adaptability of this mechanism in different scenarios, the system also monitors the particle kinetic energy decay rate during the path planning process. When the system recognizes that the kinetic energy consumption rate is too fast and the particle has not yet entered the low potential energy zone, it automatically triggers a secondary path re-exploration mechanism. This mechanism expands the reachable area by reducing the adjustment coefficient and allowing particle back jumps, so as to avoid scheduling failure of high-priority tasks due to resource stickiness.

[0069] Example 6: In this example, when constructing the logical particle path propagation model, the path cost is composed of physical distance cost and potential energy change cost. The physical distance cost measures the basic communication cost required for each hop of a service request within the logical mesh. This cost can be set based on engineering experience such as the average hop count delay of the network topology or the fixed transit time between nodes. It does not rely on real-time monitoring values ​​and only needs to be combined with the static configuration of the logical structure during deployment, thus possessing engineering simplicity and scenario adaptability. The potential energy change cost, on the other hand, is directly related to the time recovery entropy value held by the source node. It represents the indirect cost of a service request traversing different resource flexibility levels, reflecting the internal resource protection mechanism built by the system to maintain the availability of critical resources. The path cost... In the cost function, to achieve a dynamic balance between the two types of costs mentioned above, an adjustable proportional coefficient is introduced to reflect the platform's preferred configuration of future responsiveness and current communication efficiency under different operating strategies. In actual use, the value of this coefficient is not uniquely optimal, but is set by the administrator within a preset range in the early stage of system launch, combined with the platform's design goals and load characteristics. It can also be periodically adjusted based on platform operation feedback. In actual engineering implementation, the setting of this coefficient may be based on the path cost distribution of historical tasks, node resource scheduling offset statistics, or path fluctuation stability of typical service types, etc., but it does not need to rely on real-time training or external algorithm support. It can be statically loaded by the system management logic to ensure low complexity and high determinism in deployment and operation.

[0070] To avoid parameter mismatch during actual deployment, the system maintains a uniform measurement scale for all parameters and employs a normalization strategy in path cost calculation, unifying all cost factors into a dimensionless ratio system. This ensures that the propagation process of logical particles maintains consistent control logic across any topological scale. The path search module does not rely on precise comparison of absolute cost values ​​but instead uses a heuristic estimation function to evaluate the remaining propagation distance and potential energy gradient. This ensures that even with short-term fluctuations in the states of some nodes, the system can still output resource allocation paths with a stable strategy, avoiding problems such as severe path oscillations or premature particle energy depletion. Furthermore, the implementation process also includes measures to address particle propagation termination... The method for determining the points is designed with a mechanism. After path planning is completed, the system not only configures services based on the source node corresponding to the final landing position of the logical particle, but also records the total logical kinetic energy consumed by the particle during propagation. This total amount is obtained by deducting the path propagation cost from the initial budget value. Based on this, the system generates the final service settlement price, so that the price generation and path propagation have consistent internal logic. This logic is reflected at the mechanism level as the pricing feedback of service requests on resource flexibility consumption, so that the use of high-flexibility resources is superimposed with a natural inhibition mechanism. Thus, the allocation rhythm of high-value resources is controlled through path planning itself, and a price damping effect is constructed within the system.

[0071] Example 7: This example further illustrates how the present invention ensures that it finds the globally optimal source configuration target during the logical particle path planning process. For example, after a service request is abstracted as a logical particle carrying initial kinetic energy and enters the dynamic virtual machine cost field, determining its propagation path is the key to decoupling source configuration and dynamic pricing. Specific optimization strategies include:

[0072] First, a multi-path parallel evaluation mechanism is introduced. When a logical particle faces multiple possible path branches at a certain node, the system does not immediately choose the single path that seems optimal at the moment. Instead, within a certain range, for example, based on a preset path exploration breadth coefficient, it simultaneously evaluates and tracks several suboptimal paths. This parallel tracking allows the particle to follow the general trend and move towards the low-potential region during the initial exploration, while also retaining the possibility of exploring nearby high-potential regions that may lead to a better global low-potential channel. Ultimately, the path with the smallest total action among all tracked paths is determined as the final globally optimal propagation path. Secondly, the heuristic function for the A* algorithm has been optimized by guiding the global potential energy gradient. Traditional heuristic functions may focus more on estimating the physical distance from the current position to the target point. In this invention, the potential energy distribution trend of the global opportunity cost field is incorporated into the heuristic function, making it not only an estimate of physical distance but also a prediction of the direction of potential energy depressions. For example, the heuristic function combines the average potential energy of the current region, the average potential energy of the target source, and the logical topology to guide particles to tilt in directions more likely to lead to large low-potential areas, even if this may mean traversing small areas of high ground in the short term. Furthermore, to cope with extreme situations, such as dynamic adjustment and backtracking mechanisms, if the kinetic energy consumption rate of the logical particle is abnormally fast during path propagation, indicating that it may be stuck in a high-potential quagmire, the system will trigger a dynamic adjustment logic. At this time, the potential energy can be temporarily reduced. The weighting, or allowing particles to make limited backtracking attempts to re-explore different path branches, this dynamic adaptability ensures that even if the initial path selection is not ideal, the system still has the opportunity to correct errors and find the globally optimal path again.

[0073] Regarding the cost function in the principle of least action In its composition, This represents the cost of physical distance, specifically the basic resource overhead required for a logical particle to jump one step in the virtual field. It can be regarded as communication latency, scheduling link occupation, or consumption of basic computing resources. This cost is usually relatively stable and can be empirically set according to the logical topology and network performance during system deployment. This ensures that when resource availability is similar, the system will prioritize the source with closer distance and higher transmission efficiency.

[0074] This quantifies the cost of potential energy change. It is the total change in potential energy experienced by a logical particle along its propagation path. It is directly derived from the potential energy level determined by the time recovery entropy of the information source. When a logical particle moves from a low potential energy region to a high potential energy region, it occupies an information source that is currently busy and has a long recovery time. A positive and large value will increase the total action. The significant increase directly reflects the additional burden of occupying a high opportunity cost information source; conversely, if particles primarily move in low potential energy regions, If the value is small or even negative (if the potential energy decreases), its contribution to the total action is also small.

[0075] It is a preset adjustment coefficient, whose function is to balance the relative weights of physical distance cost and potential energy change cost in path decision-making. When the value is high, the system will place greater emphasis on the future flexibility of the information source and tend to avoid high-potential regions, even if this means a slightly longer physical distance; conversely, when... At lower levels, physical distance cost becomes dominant, and the system may prefer to choose the nearest information source, even if its opportunity cost is slightly higher. Therefore, the cost function... It's not just a simple addition; by assigning weights to changes in potential energy, it enables the system to intrinsically consider and quantify the opportunity cost inherent in occupying a source when selecting a source for a service request. The kinetic energy consumed by the logical particle along the path is the total action of that path. The manifestation of this will ultimately affect service prices. This mechanism ensures the occupation of information sources with high opportunity costs (high time recovery entropy), and will provide feedback through higher kinetic energy consumption and final price, thereby forming an adaptive price damping effect. This effectively guides resource demand and avoids the unreasonable occupation of high-value resources by low-value tasks. These are all extended implementation methods known to those skilled in the art.

[0076] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for decoupling information source allocation and price based on a dynamic pricing mechanism, characterized in that, The method includes the following steps: Step a: For each information source in the information source pool, determine a time recovery entropy. The time recovery entropy is a scalar value used to characterize the time required for the information source to be released from its current service state and recover to an idle state that can accept any type of new task. The determination of the time recovery entropy is based on the estimated state recovery time parameter of the task currently carried by the information source, and the value of the time recovery entropy is only updated infrequently when the type of task carried by the information source changes. Step b: Based on the time recovery entropy value of the information source, a dynamic virtual opportunity cost field is constructed in the cloud controller, where the time recovery entropy value determines the potential energy corresponding to the information source in the opportunity cost field; Step c: When a service request and its initial quote are received, the service request is abstracted into a logical particle. The initial quote carried by the logical particle serves as its initial kinetic energy and propagates in the opportunity cost field. Step d: Determine the final propagation path of the logical particle in the opportunity cost field according to the principle of least action. The principle of least action minimizes the total action of the final propagation path. The total action includes the physical distance action and the potential energy change action. Step e: Determine the information source corresponding to the end point of the final propagation path as the configuration target of the service request, and determine the final service price based on the kinetic energy consumed by the logical particle along the final propagation path. Furthermore, the construction of the dynamic virtual machine opportunity cost field in step b includes: the cloud controller periodically collects the time recovery entropy values ​​of all information sources in the information source pool, and uses the inverse distance weight interpolation algorithm to generate discretized potential field data on the logical grid based on the time recovery entropy values ​​of the information sources and their logical positions to form the opportunity cost field; The final propagation path following the principle of least action in step d is found using a variant of the A* algorithm, where the cost function of the A* algorithm is... Integrating physical distance cost Cost of potential energy change Their relationship can be expressed as: ， in, To adjust the preset coefficients for the relative weights of physical distance cost and potential energy change cost, This represents the total change in potential energy experienced by the logical particle along the path.

2. The method for decoupling information source configuration and price based on a dynamic pricing mechanism as described in claim 1, characterized in that, The determination of time recovery entropy in step a is based on the estimated recovery time contained in the metadata of the task currently being carried by the information source; the estimated recovery time is a predefined value that can indicate the reload time or context switching time after cache failure, and the update frequency of the time recovery entropy value is less than once per minute.

3. The method for decoupling information source configuration and price based on a dynamic pricing mechanism as described in claim 1, characterized in that, In step c, the propagation rule of logical particles in the opportunity cost field is as follows: logical particles tend to move towards regions with lower potential energy in the opportunity cost field; when logical particles pass through regions with higher potential energy during propagation, their kinetic energy is consumed.

4. The method for decoupling information source configuration and price based on a dynamic pricing mechanism as described in claim 1, characterized in that, This is a preset coefficient, and its value ranges from 0.1 to 100.

5. The method for decoupling information source configuration and price based on a dynamic pricing mechanism as described in claim 1, characterized in that, The determination of the final service price involves combining the kinetic energy consumed by the logical particle along the final propagation path with the initial quote for the service request to arrive at the final settlement price.

6. The method for decoupling information source configuration and price based on a dynamic pricing mechanism as described in claim 1, characterized in that, The information sources in the information source pool include virtual machines, containers, physical servers, and network slice resource units; the types of service requests include data cleaning tasks, AI model training tasks, data analysis tasks, and high real-time business tasks.

7. The method for decoupling information source configuration and price based on a dynamic pricing mechanism as described in claim 1, characterized in that, The source node maintains only one scalar value, the time recovery entropy, and updates to this scalar value are only triggered when the task type carried by the source changes.

8. A system for decoupling information source allocation and price based on a dynamic pricing mechanism, used to implement the method for decoupling information source allocation and price based on a dynamic pricing mechanism as described in claim 1, characterized in that, The system includes: The source node is configured to determine and maintain the time recovery entropy. The time recovery entropy is a scalar value used to characterize the time required for the source to be released from its current service state and recover to an idle state that can accept any type of new task. The determination of the time recovery entropy is based on the estimated state recovery time parameter of the task currently carried by the source, and the value of the time recovery entropy is only updated infrequently when the type of task carried by the source changes. The cloud controller, communicating with the source node, is configured to: construct a dynamic virtual opportunity cost field based on the time recovery entropy value of the source node, where the time recovery entropy value determines the potential energy corresponding to the source in the opportunity cost field; when a service request and its initial quote are received, the service request is abstracted into a logical particle, with the initial quote carried by the logical particle serving as its initial kinetic energy, and it propagates in the opportunity cost field; the final propagation path of the logical particle in the opportunity cost field is determined according to the principle of least action, which minimizes the total action of the final propagation path, including the physical distance action and the potential energy change action; the source corresponding to the endpoint of the final propagation path is determined as the configuration target of the service request, and the final service price is determined based on the kinetic energy consumed by the logical particle along the final propagation path.

Images