A flow scheduling method and device, electronic equipment and storage medium

Abstract

The application discloses a traffic scheduling method and device, electronic equipment and storage medium, and relates to the technical field of communication. The method comprises the following steps: obtaining bandwidth information of a network link, and calculating a link potential energy value; determining a target scheduling path from a candidate scheduling path based on the link potential energy value, and scheduling traffic according to the target scheduling path; recalculating the potential energy value according to actual traffic borne by the link after scheduling, and determining a scheduling warning level; and performing a corresponding scheduling control operation according to the warning level. The application solves the problems of uneven load caused by a traditional static hash mechanism, single-point failure and expansion bottleneck of a centralized control scheme, and the technical problem that a traditional congestion control can only be passively reacted and cannot be actively prevented, and achieves the technical effects of realizing global load balancing under a distributed architecture, preventing traffic before congestion occurs, and guaranteeing differentiated quality of service of multi-priority services through hierarchical control.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a traffic scheduling method, apparatus, electronic device and storage medium. Background Technology

[0002] With the widespread adoption of distributed applications, especially large-scale artificial intelligence model training, traffic storms and long-distance flows pose unprecedented challenges to the operational quality of network links. For example, traditional equal-cost multi-path routing is prone to hash collisions in complex traffic scenarios, causing large data flows and small data flows to compete for the same link, resulting in small flow congestion and link utilization imbalance. Excessive link congestion directly leads to increased packet loss and queue latency, while other links in the network may be idle, resulting in low overall resource utilization efficiency. At the same time, modern data centers simultaneously handle throughput-intensive and latency-sensitive tasks, especially artificial intelligence workloads, whose traffic growth is rapid and unpredictable, posing differentiated service quality requirements to the network. To address the aforementioned issues, the industry has proposed several solutions: traditional static scheduling techniques, such as equal-cost multi-path routing, are simple to implement and have low overhead, but cannot detect dynamic congestion; improved dynamic scheduling techniques, such as the protective load balancing proposed by Microsoft, use explicit congestion notifications to switch paths at the host side, but the response speed is limited by feedback latency and the control granularity is limited; centralized scheduling techniques based on software-defined networks achieve optimal path decisions through global state collection, but face scalability challenges such as controller performance bottlenecks, single points of failure, and additional control overhead; end-to-end network collaboration and receiver-driven technologies, such as high-precision congestion control, use in-band telemetry to achieve microsecond-level precise rate adjustment, but have high deployment thresholds, requiring upgrades to switches and smart network cards that support in-band telemetry and deep modifications to the transmission protocol stack. Summary of the Invention

[0003] This application provides a traffic scheduling method, apparatus, electronic device, and storage medium to at least address the problems of related technologies, which struggle to achieve an effective balance between maximizing link utilization and maintaining low latency stability. These technologies also suffer from issues such as coarse perception granularity, slow response speed, high system complexity, and high deployment costs, and are unable to simultaneously address multiple challenges, including uneven traffic load, sudden congestion, and diverse workload requirements.

[0004] This application provides a traffic scheduling method, which includes: acquiring bandwidth information of network links; calculating a first link potential value of the network links based on the bandwidth information; responding to receiving a traffic scheduling instruction, determining candidate scheduling paths for the traffic to be scheduled based on the start node and end node in the traffic scheduling instruction; using the network links that make up the candidate scheduling paths as candidate network links, and determining a target scheduling path from the candidate scheduling paths based on the first link potential value of the candidate network links; using the network links that make up the target scheduling path as target network links, and scheduling the traffic to be scheduled according to the target network links; using the scheduled target network links as network links to be updated, and calculating a second link potential value of the network links to be updated based on the first link potential value of the network links to be updated; determining a scheduling warning level for the network links to be updated based on the second link potential value of the network links to be updated, and performing corresponding scheduling control operations on the network links to be updated based on the scheduling warning level.

[0005] This application also provides a traffic scheduling device, comprising: a potential energy calculation module, configured to: acquire bandwidth information of network links; calculate a first link potential energy value of the network links based on the bandwidth information; a traffic navigation module, configured to: in response to receiving a traffic scheduling instruction, determine candidate scheduling paths for the traffic to be scheduled based on the start node and end node in the traffic scheduling instruction; use the network links constituting the candidate scheduling paths as candidate network links, and determine a target scheduling path from the candidate scheduling paths based on the first link potential energy value of the candidate network links; and an execution control module, configured to: use the network links constituting the target scheduling path as target network links, and schedule the traffic to be scheduled according to the target network links; use the scheduled target network links as network links to be updated, and calculate a second link potential energy value of the network links to be updated based on the first link potential energy value of the network links to be updated; determine the scheduling warning level of the network links to be updated based on the second link potential energy value of the network links to be updated, and perform corresponding scheduling control operations on the network links to be updated based on the scheduling warning level.

[0006] This application also provides an electronic device, including: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the above-described traffic scheduling methods.

[0007] This application also provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of any of the above-described traffic scheduling methods.

[0008] This application obtains network link bandwidth information and calculates the first link potential value for each link. This first link potential value, by integrating three dimensions—current utilization, trend component, and prediction component—can comprehensively characterize the real-time load status, changing trend, and future congestion risk of the link. When a traffic scheduling instruction is received, candidate scheduling paths are determined based on the starting and ending nodes, and the target scheduling path is determined from the candidate paths based on the link potential values ​​that make up each candidate path. After scheduling traffic according to the target scheduling path, the second link potential value is recalculated based on the actual traffic carried by each link, realizing dynamic updating of the potential state. The second link potential value is compared with a preset threshold range to determine the scheduling warning level. According to the scheduling warning level, corresponding hierarchical control operations are executed: Level 1 maintains the status quo; Level 2 sends path switching suggestions; Level 3 reallocates traffic; Level 4 reduces the bandwidth allocation ratio of low-priority traffic; Level 5 discards low-priority traffic data packets. Through the above scheme, this application constructs a closed-loop autonomous system from potential energy perception, path decision-making, traffic scheduling, state update, hierarchical early warning to differentiated control, achieving the technical effects of global load balancing under a distributed architecture, preventive diversion before congestion occurs, and ensuring differentiated service quality for multi-priority businesses through hierarchical control. Attached Figure Description

[0009] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 A flowchart illustrating a traffic scheduling method provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a traffic scheduling device provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device for implementing traffic scheduling method steps, provided in an embodiment of this application. Detailed Implementation

[0011] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0012] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0013] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0014] Embodiments of this application provide a traffic scheduling method, such as... Figure 1 As shown, the traffic scheduling method includes: acquiring bandwidth information of network links; calculating a first link potential value of the network links based on the bandwidth information; responding to a received traffic scheduling instruction, determining candidate scheduling paths for the traffic to be scheduled based on the start and end nodes in the traffic scheduling instruction; using the network links that make up the candidate scheduling paths as candidate network links, and determining a target scheduling path from the candidate scheduling paths based on the first link potential value of the candidate network links; using the network links that make up the target scheduling path as target network links, and scheduling the traffic to be scheduled according to the target network links; using the scheduled target network links as network links to be updated, and calculating a second link potential value of the network links to be updated based on the first link potential value of the network links to be updated; determining the scheduling warning level of the network links to be updated based on the second link potential value of the network links to be updated, and performing corresponding scheduling control operations on the network links to be updated based on the scheduling warning level.

[0015] Specifically, the above method constructs a closed-loop traffic scheduling framework encompassing perception, decision-making, execution, and feedback. First, quantitative perception of network status is achieved through the calculation of link potential values. Second, path selection based on potential values ​​provides a global perspective for traffic scheduling. Third, recalculating potential values ​​and determining warning levels after scheduling enables real-time evaluation of scheduling effectiveness. Finally, tiered control is implemented based on warning levels, allowing the system to take differentiated intervention measures according to the degree of congestion. This closed-loop mechanism ensures the transformation of traffic scheduling from passive response to proactive prevention.

[0016] Furthermore, the bandwidth information includes the current utilization rate, the historical average utilization rate within the historical scheduling period, and the historical maximum utilization rate within the current scheduling period. Based on the bandwidth information, the first link potential value of the network link is calculated, including: using the current utilization rate as the first component of the first link potential value of the network link; calculating the utilization rate offset based on the current utilization rate and the historical average utilization rate, and calculating the second component of the first link potential value of the network link based on the utilization rate offset and the historical maximum utilization rate; determining the current prediction factor of the network link, and calculating the third component of the first link potential value of the network link based on the first component, the second component, the current prediction factor, and the preset step size coefficient; and determining the first link potential value of the network link based on the weighted sum of the first component, the second component, and the third component.

[0017] Specifically, the above method decomposes the link potential energy value into three dimensions for calculation. The current utilization component reflects the instantaneous load state of the link, providing a real-time benchmark for potential energy calculation; the trend component, calculated using utilization offset and historical maximum utilization, characterizes the direction and speed of link load changes, enabling the system to perceive load change trends; the prediction component predicts future load based on historical data, giving the potential energy value a forward-looking nature. The weighted fusion of the three components allows the link potential energy value to reflect the current state, characterize the changing trend, and predict future trends, achieving a leap from passive response to proactive prevention.

[0018] In a specific implementation, the first link potential energy value E link The calculation formula is: ; Among them, U current U represents the instantaneous component of bandwidth utilization at the current time t, i.e., the first component of the first link potential energy value; trend This represents the trend component of bandwidth utilization at the current time t, which is the second component of the first link potential value; U predicted U represents the predicted component of bandwidth utilization at the current time t, which is the third component of the first link potential value; β represents the weighting coefficient of the trend component, and γ represents the weighting coefficient of the predicted component, and both weighting coefficients are greater than 0 and less than 1. current The value is the bandwidth utilization rate at the current time t; the second component U of the first link potential energy value. trend The calculation formula is: ; U predicted The calculation formula is: ; Where h represents the prediction step size; for example, when h=1, this calculation formula represents the bandwidth utilization for predicting the next cycle; L tT represents the baseline value at the current time t. t This represents the adjustment value at the current time t. The formulas for calculating both are as follows: ; ; in, This represents the instantaneous component of bandwidth utilization at the current time t. This represents the trend component of bandwidth utilization at the current time t.

[0019] Furthermore, after calculating the first link potential value of the network link based on bandwidth information, the traffic scheduling method further includes: determining the directly connected neighbor nodes of the network node, and taking the network link between the network node and the directly connected neighbor node as the directly connected neighbor link, wherein the directly connected neighbor node is adjacent to the network node and directly connected to the network node; obtaining the first link potential value of the directly connected neighbor link, and determining the maximum value of the first link potential value among the directly connected neighbor links as the maximum potential value of the directly connected link of the network node in this round; obtaining the node potential value of the directly connected neighbor node in this round, and calculating the average potential value of the directly connected node of the network node in this round based on the node potential value of the directly connected neighbor node in this round; and calculating the average potential value of the directly connected node in this round and a preset... The attenuation factor is used to calculate the propagation potential energy value of the network node in the current round of direct links. The maximum value between the maximum potential energy value of the direct links in the current round and the propagation potential energy value of the network node in the current round is determined as the node potential energy value of the network node in the current round. In response to the absolute value of the difference between the node potential energy value of the network node in the current round and the node potential energy value of the network node in the previous round being less than or equal to a preset convergence threshold, the iteration stops, and the node potential energy value of the network node in the current round is determined as the current node potential energy value of the network node. In response to the absolute value of the difference being greater than the preset convergence threshold, the node potential energy value of the network node in the current round is sent to the directly connected neighbor nodes to update the node potential energy values ​​of the directly connected neighbor nodes, and the next round of iteration is performed until the absolute value of the difference is less than or equal to the preset convergence threshold.

[0020] Specifically, the above method uses an iterative calculation mechanism that takes the maximum value of the direct link's maximum potential energy and the decaying propagation potential energy to enable each network node to perceive the global congestion situation in a purely distributed manner. The maximum potential energy of the direct link ensures that the node can respond immediately to the most severe congestion on its local direct link, while the decaying propagation potential energy allows the congestion information from distant locations to spread hop by hop along the network, but always maintains its influence on the node's state. Taking the maximum value of the two ensures that the node's potential energy is always dominated by the highest-risk link in the local worst link and the overall load of the surrounding area, avoiding the problem of congestion information being diluted by weighted summation. At the same time, through iterative determination of the convergence threshold, the node automatically stops propagation after the potential energy stabilizes, effectively controlling communication overhead. This mechanism has good scalability in ultra-large-scale data center networks and provides the traffic navigation module with clear high-potential barrier identification and low-potential guiding gradient.

[0021] In a specific implementation, for network node i, its current node potential value E [i] The calculation formula is: ; Among them, E i This represents the maximum potential energy value of the direct link of the network node. This represents the average potential energy of directly connected nodes in the network, where α represents a preset attenuation factor, which takes a value greater than 0 and less than 1. The propagation potential energy of directly connected nodes is expressed as... .

[0022] In a specific implementation, the preset convergence threshold is greater than or equal to 0.01 and less than or equal to 0.05. This threshold, serving as the boundary for determining potential energy changes, should be on the same order of magnitude as the potential energy value. The range of 0.01 to 0.05 represents a variation of 1% to 5% relative to the 0 to 1 range, effectively determining whether the node's potential energy is stable. The node's potential energy value is updated asynchronously. A new round of potential energy propagation is triggered only when the network node's own potential energy value changes beyond the preset threshold or when it receives a potential energy value from a neighboring node that exceeds the preset threshold, effectively controlling communication overhead. Furthermore, the number of iterations can be set. Through the iterative convergence process, each node obtains a stable overall potential energy view within a finite number of iterations, forming a potential energy gradient field that gradually decreases from high-load areas to low-load areas.

[0023] Further, the current prediction factor of the network link is determined, and the third component of the first link potential value of the network link is calculated based on the first component, the second component, the current prediction factor, and the preset step size coefficient. This includes: obtaining the previous reference value and the previous prediction factor of the network link; calculating the current reference value based on the first and second components, and calculating the reference value change based on the current reference value and the previous reference value; calculating the current prediction factor based on the reference value change and the previous prediction factor, and calculating the prediction adjustment based on the current prediction factor and the preset step size coefficient; and calculating the third component based on the prediction adjustment and the current reference value.

[0024] Specifically, the above method further refines the calculation mechanism of the predictive component. By introducing two state variables—a baseline value and a predictor factor—and employing an iterative update method to calculate the third component, the third component can continuously track the dynamic changes in network load. That is, the baseline value reflects the baseline level of link load, and the predictor factor characterizes the trend of load change; the combination of the two enables adaptive tracking of network fluctuations at different time scales. This recursive calculation method avoids the overhead of storing large amounts of historical data, and has the advantages of low computational complexity and low storage overhead, making it easy to implement line-rate processing on hardware switches.

[0025] In a specific implementation, the preset step size coefficient h is a positive integer.

[0026] Further, based on the start node and end node in the traffic scheduling instruction, candidate scheduling paths for the traffic to be scheduled are determined, including: obtaining the network topology and querying the reachable scheduling paths from the start node to the end node from the network topology; identifying the network links that make up the reachable scheduling paths as reachable network links and obtaining the total number of reachable network links; comparing the total number of reachable network links with a preset link number threshold; determining the reachable scheduling path as a candidate scheduling path in response to the total number of reachable network links being less than or equal to the link number threshold; and determining the reachable scheduling path as a non-candidate scheduling path in response to the total number of reachable network links being greater than the link number threshold.

[0027] Specifically, the above method defines the screening mechanism for candidate scheduling paths. By screening reachable scheduling paths using a preset link count threshold, paths with excessive hop counts (i.e., a large number of links) are excluded from the candidate set, avoiding exhaustive search in the subsequent path selection process and achieving a balance between path quality and computational overhead. Paths with fewer hops typically have lower end-to-end latency and higher transmission reliability, while the screening operation significantly reduces the computational complexity of path selection, ensuring that the decision-making process meets the microsecond-level scheduling latency requirements of the data center.

[0028] In a specific implementation, the preset link number threshold can be set to 5. When the total number of reachable network links in all reachable scheduling paths is greater than 5, the link number threshold can be increased to 6, 7, or 8, etc., so that the total number of reachable network links in at least one reachable scheduling path is less than or equal to the link number threshold.

[0029] In specific implementations, the direction with the fastest decrease in potential energy is selected with a high probability, while retaining a small probability of exploring other paths, effectively balancing the contradiction between local optima and global exploration. Candidate scheduling paths can also be considered from the perspective of link potential energy values. For example, when a link's potential energy value is detected to exceed the fifth threshold, the network link is marked as an obstacle to ensure that traffic scheduling does not pass through this network link, and a warning is issued when it is unavoidable.

[0030] Further, determining the target scheduling path from the candidate scheduling paths based on the first link potential value of the candidate network links includes: calculating the total path potential energy of the candidate scheduling paths according to the first link potential energy value of the candidate network links; matching the candidate scheduling paths with other candidate scheduling paths to obtain the number of links overlapping in the candidate scheduling paths; obtaining the total number of links in the candidate scheduling paths, and determining the target scheduling path from the candidate scheduling paths based on at least one of the following data: the total path potential energy of the candidate scheduling paths, the number of links overlapping, and the total number of links.

[0031] Specifically, the above method constructs a multi-dimensional path evaluation index system. The total path potential energy reflects the congestion level of the entire path; the number of overlapping links characterizes the independence between paths, used to avoid simultaneous congestion during multi-path traffic splitting; and the total number of links reflects the transmission efficiency of the path. By introducing multi-dimensional evaluation indexes, path selection can flexibly adjust the evaluation dimensions according to the needs of different scenarios, providing a decision-making basis for subsequent differentiated path selection strategies.

[0032] Further, based on at least one of the following data: the total path potential energy of the candidate scheduling paths, the number of overlapping links, and the total number of links, the target scheduling path is determined from the candidate scheduling paths, including: in response to the low priority of the traffic to be scheduled, the candidate scheduling path with the smallest total number of links is determined as the intermediate scheduling path; in response to the uniqueness of the intermediate scheduling path, the intermediate scheduling path is determined as the target scheduling path and the target scheduling path is used as the primary scheduling path; in response to the non-uniqueness of the intermediate scheduling path, the intermediate scheduling path with the lowest total path potential energy is determined as the target scheduling path and the target scheduling path is used as the primary scheduling path.

[0033] Specifically, considering that low-priority traffic has lower requirements for latency and bandwidth guarantees, it is guided to the path with the fewest hops. This satisfies its basic transmission needs while preventing it from consuming excessive network resources. When multiple paths with the same number of hops exist, the path with the lowest total potential energy is further selected, ensuring that the lightest-loaded path is chosen among the shortest paths. This dual-selection mechanism ensures the basic quality of service for low-priority traffic while minimizing its negative impact on the overall network load balancing.

[0034] In a specific embodiment, the traffic to be scheduled is scheduled in descending order of priority.

[0035] Further, determining the target scheduling path from the candidate scheduling paths based on at least one of the following data: the total path potential energy, the number of overlapping links, and the total number of links, further includes: in response to the traffic to be scheduled belonging to medium priority, determining the candidate scheduling path with the smallest total number of links as the intermediate scheduling path; in response to the intermediate scheduling path being unique, determining the intermediate scheduling path as the primary scheduling path in the target scheduling path, and determining the candidate scheduling path with the lowest total path potential energy as the backup scheduling path in the target scheduling path, wherein the primary scheduling path and the backup scheduling path are different scheduling paths; in response to the intermediate scheduling path not being unique, determining the intermediate scheduling path with the lowest total path potential energy as the primary scheduling path, and determining the candidate scheduling path with the lowest total path potential energy as the backup scheduling path among the other candidate scheduling paths besides the primary scheduling path.

[0036] Specifically, the above method introduces a primary and backup dual-path protection mechanism for medium-priority traffic. The primary path is determined based on the criterion of minimizing the total number of links, followed by minimizing the total path potential energy, ensuring optimal transmission efficiency for the traffic. The backup path is determined solely based on minimizing the total path potential energy, and unlike the primary path, it ensures that in the event of congestion or failure on the primary path, traffic can quickly switch to the least loaded backup path. This differentiated selection strategy for primary and backup paths provides reliability redundancy while ensuring transmission efficiency, enabling medium-priority traffic to achieve a balance between efficiency and reliability.

[0037] Further, determining the target scheduling path from the candidate scheduling paths based on at least one of the following data: the total path potential energy, the number of overlapping links, and the total number of links, further includes: in response to the high priority of the traffic to be scheduled, determining the candidate scheduling path with the smallest total number of links as the intermediate scheduling path; in response to the uniqueness of the intermediate scheduling path, determining the intermediate scheduling path as the primary scheduling path in the target scheduling path, and determining the candidate scheduling path with the fewest overlapping links with the primary scheduling path as the backup scheduling path in the target scheduling path; in response to the non-uniqueness of the intermediate scheduling path, determining the intermediate scheduling path with the lowest total path potential energy as the primary scheduling path, and determining the candidate scheduling path with the fewest overlapping links with the primary scheduling path as the backup scheduling path.

[0038] Specifically, the above method employs a dual-path protection strategy with maximum path independence for high-priority traffic. Unlike the selection of the lightest-loaded backup path for medium-priority traffic, the backup path for high-priority traffic is determined based on minimizing the number of links overlapping with the primary path, rather than minimizing the total potential energy of non-paths. Less link overlap reduces the probability of both paths experiencing congestion or failure simultaneously, thus providing the strongest transmission reliability guarantee for high-priority traffic. This design reflects a refined approach to ensuring differentiated service quality for services of different priorities.

[0039] Further, scheduling the traffic to be scheduled according to the target network link includes: in response to the target scheduling path including a primary scheduling path and a backup scheduling path, calculating the total path potential energy of the target scheduling path based on the total path potential energy of the primary scheduling path and the total path potential energy of the backup scheduling path; calculating the first allocation ratio of the primary scheduling path based on the total path potential energy of the target scheduling path and the total path potential energy of the primary scheduling path, and calculating the first allocation amount of the primary scheduling path based on the bandwidth occupancy of the traffic to be scheduled and the first allocation ratio; calculating the second allocation ratio of the backup scheduling path based on the total path potential energy of the target scheduling path and the total path potential energy of the backup scheduling path, and calculating the second allocation amount of the backup scheduling path based on the bandwidth occupancy and the second allocation ratio; in response to the target scheduling path including the primary scheduling path but not the backup scheduling path, determining the bandwidth occupancy as the allocation amount of the primary scheduling path.

[0040] Specifically, the above method defines the traffic allocation mechanism between primary and backup paths. When both primary and backup paths exist, the allocation ratio is inversely proportional to the sum of the path potential energy; that is, the path with the lower total potential energy is allocated more traffic, achieving load-aware traffic allocation and dynamically guiding traffic to the less loaded path, preventing multiple paths from becoming congested simultaneously. When only the primary path exists, all traffic is carried by the primary path, simplifying the scheduling logic. This mechanism ensures that traffic allocation fully utilizes idle path resources without creating new congestion points.

[0041] In a specific implementation, the allocation amount = bandwidth usage × allocation ratio. For example, the first allocation ratio = (total path potential energy of the target scheduling path - total path potential energy of the main scheduling path) / total path potential energy of the target scheduling path.

[0042] Further, based on the first link potential value of the network link to be updated, the second link potential value of the network link to be updated is calculated, including: in response to the target scheduling path including a primary scheduling path and a backup scheduling path, the network links constituting the primary scheduling path are designated as primary scheduling network links, and the second link potential value of the primary scheduling network links is calculated based on the first allocation and the first link potential value of the primary scheduling network links; the network links constituting the backup scheduling path are designated as backup scheduling network links, and the second link potential value of the backup scheduling network links is calculated based on the second allocation and the first link potential value of the backup scheduling network links; in response to the target scheduling path including the primary scheduling path but not the backup scheduling path, the second link potential value of the primary scheduling network links is calculated based on the bandwidth occupancy and the first link potential value of the primary scheduling network links.

[0043] Specifically, the above method implements a dynamic update mechanism for potential energy values. After traffic scheduling is completed, the state of network links changes. This method recalculates the potential energy values ​​of the links constituting each path based on the actual traffic allocation carried by each path. That is, the primary scheduling network link and the backup scheduling network link calculate new potential energy values ​​based on their respective traffic loads. This update mechanism ensures that the potential energy field can reflect the network state changes after traffic scheduling in real time, providing accurate state input for the next round of scheduling decisions. It is a key implementation of the feedback loop in the closed-loop control system.

[0044] In a specific implementation, the second link potential value = the first link potential value + (first allocation amount / total bandwidth of network links).

[0045] Further, based on the second link potential energy value of the network link to be updated, the scheduling warning level of the network link to be updated is determined, including: comparing the second link potential energy value of the primary scheduling network link or the backup scheduling network link with a preset potential energy interval threshold to determine the potential energy interval to which the second link potential energy value of the primary scheduling network link or the backup scheduling network link belongs; in response to the second link potential energy value of the primary scheduling network link or the backup scheduling network link being greater than or equal to the first threshold and less than the second threshold, determining that the second link potential energy value of the primary scheduling network link or the backup scheduling network link belongs to the first potential energy interval, and determining the scheduling warning level of the primary scheduling network link or the backup scheduling network link as the first warning level; in response to the second link potential energy value of the primary scheduling network link or the backup scheduling network link being greater than or equal to the second threshold and less than the third threshold, determining that the second link potential energy value of the primary scheduling network link or the backup scheduling network link belongs to the second potential energy interval, and determining the scheduling warning level of the primary scheduling network link or the backup scheduling network link as the second ... If the potential energy value of the primary or backup scheduling network link is greater than or equal to the third threshold and less than the fourth threshold, the second link potential energy value of the primary or backup scheduling network link is determined to belong to the third potential energy range, and the scheduling warning level of the primary or backup scheduling network link is determined to be the third warning level. If the potential energy value of the primary or backup scheduling network link is greater than or equal to the fourth threshold and less than the fifth threshold, the second link potential energy value of the primary or backup scheduling network link is determined to belong to the fourth potential energy range, and the scheduling warning level of the primary or backup scheduling network link is determined to be the fourth warning level. If the potential energy value of the primary or backup scheduling network link is greater than or equal to the fifth threshold and less than the sixth threshold, the second link potential energy value of the primary or backup scheduling network link is determined to belong to the fifth potential energy range, and the scheduling warning level of the primary or backup scheduling network link is determined to be the fifth warning level. Among these, the warning levels are ranked from lowest to highest according to their urgency: first warning level, second warning level, third warning level, fourth warning level, and fifth warning level.

[0046] Specifically, the above method constructs a five-level progressive early warning classification system. By setting five increasing threshold intervals—the first threshold, the second threshold, the third threshold, the fourth threshold, the fifth threshold, and the sixth threshold—the link potential value is mapped to five early warning levels with increasing urgency. The advantages of this classification system are: first, high granularity, with five levels precisely characterizing different risk levels of the link from completely normal to severely congested; second, clear threshold boundaries, providing clear operational triggering conditions for subsequent graded control; and third, standardization, allowing links of different capacities to adopt a unified early warning standard.

[0047] In a specific implementation, the first threshold is 0, the second threshold is 0.6, the third threshold is 0.7, the fourth threshold is 0.8, the fifth threshold is 0.9, and the sixth threshold is 1.

[0048] Further, based on the scheduling warning level, corresponding scheduling control operations are performed on the network link to be updated, including: responding to a first warning level for the primary or backup scheduling network link, performing a first control operation on the primary or backup scheduling network link, wherein the first control operation includes maintaining the current link state of the primary or backup scheduling network link; responding to a second warning level for the primary or backup scheduling network link, performing a second control operation on the primary or backup scheduling network link, wherein the second control operation includes sending a path switching suggestion to the network node corresponding to the primary or backup scheduling network link; responding to a third warning level for the primary or backup scheduling network link, performing a path switching suggestion to the network node corresponding to the primary or backup scheduling network link. A third control operation is performed on the scheduling network link, wherein the third control operation includes reallocating the traffic to be scheduled through the primary scheduling network link or the backup scheduling network link; in response to the scheduling warning level of the primary scheduling network link or the backup scheduling network link being a fourth warning level, a fourth control operation is performed on the primary scheduling network link or the backup scheduling network link, wherein the fourth control operation includes reducing the bandwidth allocation ratio of low-priority traffic in the traffic to be scheduled through the primary scheduling network link or the backup scheduling network link; in response to the scheduling warning level of the primary scheduling network link or the backup scheduling network link being a fifth warning level, a fifth control operation is performed on the primary scheduling network link or the backup scheduling network link, wherein the fifth control operation includes dropping data packets of low-priority traffic in the traffic to be scheduled through the primary scheduling network link or the backup scheduling network link.

[0049] Specifically, the above method maps five warning levels to five control operations of increasing intensity, constructing a complete hierarchical warning and progressive control execution system. The first warning level, L0, takes no action, achieving normal monitoring with zero overhead; the second warning level, L1, only sends suggestions, alerting to risks at minimal cost; the third warning level, L2, implements traffic reallocation, proactively intervening while preserving link capacity; the fourth warning level, L3, reduces the bandwidth allocation ratio of low-priority traffic, prioritizing high-priority services; and the fifth warning level, L4, discards low-priority data packets in extreme cases, ensuring the link does not collapse due to overload. The control intensity of the five levels progresses progressively from zero intervention to hard discarding, achieving a refined balance between system stability and service quality.

[0050] In specific implementations, L0 is normal operation with no action; L1 is preventative guidance with mild rate limiting and path suggestions; L2 is destined load balancing with dynamic weight adjustment and multi-path enhancement; L3 avoids congestion by forcing traffic migration and priority scheduling; and L4 handles congestion by selectively dropping traffic and performing emergency rerouting.

[0051] In a specific embodiment, it is assumed that the network includes the following network nodes A, B, C, D, E, F, G, H, and I, totaling 9 nodes, and the total bandwidth of the network links is uniformly set to 1000 Mbps. The used bandwidth of each network link is also given. For example, link AB has occupied 400 Mbps, so its current utilization rate is 40%; BC has occupied 800 Mbps, so its current utilization rate is 80%; CF has occupied 900 Mbps, so its current utilization rate is 90%; BE has occupied 600 Mbps, so its current utilization rate is 60%; DE has occupied 200 Mbps, so its current utilization rate is 20%; EF has occupied 700 Mbps, so its current utilization rate is 70%; EH has occupied 500 Mbps, so its current utilization rate is 50%; GH has occupied 150 Mbps, so its current utilization rate is 15%; HI has occupied 450 Mbps, so its current utilization rate is 45%; DG has occupied 100 Mbps, so its current utilization rate is 10%; and AD has occupied 300 Mbps, so its current utilization rate is 30%.

[0052] Four users have submitted four new traffic requests, each with a different start and end point. Furthermore, the priorities of these four traffic requests are also different. Specifically: User 1: Traffic from G to C, bandwidth usage 80Mbps, high priority; User 2: Traffic from A to F, bandwidth usage 120Mbps, medium priority; User 3: Traffic from D to I, bandwidth usage 60Mbps, low priority; User 4: Traffic from B to H, bandwidth usage 100Mbps, medium priority. Therefore, the total new traffic is 80 + 120 + 60 + 100 = 360Mbps.

[0053] The following potential energy values ​​for the first link were calculated using the formulas: Link AB, 0.40 + 0.3 × 0.05 + 0.2 × 0.42 ≈ 0.50; BC, 0.80 + 0.3 × 0.08 + 0.2 × 0.85 ≈ 0.99; CF, 0.90 + 0.3 × 0.10 + 0.2 × 0.92 ≈ 1.11; BE, 0.60 + 0.3 × 0.06 + 0.2 × 0.62 ≈ 0.74; DE, 0.20 + 0.3 × 0.02 + 0.2 × 0.21 ≈ 0.25; EF, 0. 70+0.3×0.07+0.2×0.72≈0.87; EH, 0.50+0.3×0.05+0.2×0.52≈0.62; GH, 0.15+0.3×0.01+0.2×0.16≈0.19; HI , 0.45+0.3×0.04+0.2×0.47≈0.56; DG, 0.10+0.3×0.01+0.2×0.11≈0.13; AD, 0.30+0.3×0.03+0.2×0.32≈0.37.

[0054] User 1, traffic is sent from G to C, bandwidth usage is 80Mbps, high-priority traffic processing: the following 4 candidate scheduling paths are identified: Path G→H→E→B→C, potential energy sequence [0.19, 0.62, 0.74, 0.99], maximum potential energy is 0.99 (link BC), total potential energy is 2.54. The maximum potential energy exceeds the fifth threshold of 0.9, link bottleneck assessment: extremely severe; path G→D→A→B→C, potential energy sequence [0.13, 0.37, 0.50, 0.99], maximum potential energy is 0.99 (link BC), total potential energy is 1.99. The maximum potential energy exceeds the fifth threshold of 0.9, link bottleneck assessment: extremely severe; path G→D→E→B→C, potential energy sequence [0.13, 0.25, 0.74, 0.99], maximum potential energy is 0.99 (link BC), total potential energy is 2.11. The maximum potential energy is greater than the fifth threshold of 0.9, link bottleneck assessment: extremely severe. Path The potential energy sequence for the path G→D→E→F→C is [0.13, 0.25, 0.87, 1.11]. The maximum potential energy is 1.11 (link FC), and the total potential energy is 2.36. The maximum potential energy exceeds the fifth threshold of 0.9, indicating a severe link bottleneck. In summary, the primary path is G→D→A→B→C, distributing 40Mbps. Although it passes through BC, the potential energy of other links is relatively low. The backup path is G→H→E→B→C, also distributing 40Mbps, thus reducing the pressure on the GH and DG links.

[0055] User 2, traffic is sent from A to F, bandwidth usage is 120Mbps, medium priority traffic processing: the following 3 candidate scheduling paths are identified: Path The potential energy sequence for the path A→B→C→F is [0.50, 0.99, 1.11]. The maximum potential energy is 1.11, and the total potential energy is 2.6. The maximum potential energy exceeds the fifth threshold of 0.9, indicating a severe bottleneck in the link. (Path) The potential energy sequence for the path A→B→E→F is [0.50, 0.74, 0.87]. The maximum potential energy is 0.87, and the total potential energy is 2.11. The maximum potential energy exceeds the fourth threshold of 0.8. Link bottleneck assessment: Severe. Path The potential energy sequence for path A→D→E→F is [0.37, 0.25, 0.87]. The maximum potential energy is 0.87, and the total potential energy is 1.49. The maximum potential energy exceeds the fourth threshold of 0.8, indicating a severe link bottleneck. Therefore, the primary path (A→D→E→F) will allocate 70Mbps because it has the lowest total potential energy. The backup path (A→B→E→F) will allocate 50Mbps.

[0056] User 3, traffic is sent from D to I, bandwidth usage is 60Mbps, low priority traffic processing: the following two candidate scheduling paths are identified: Path D→E→H→I, potential energy sequence [0.25, 0.62, 0.56], maximum potential energy is 0.62, total potential energy is 1.43. The maximum potential energy is less than the third threshold of 0.7, link bottleneck assessment: moderate. Path The potential energy sequence for the path D→G→H→I is [0.13, 0.19, 0.56]. The maximum potential energy is 0.56, and the total potential energy is 0.88. The maximum potential energy is less than the second threshold of 0.6, indicating a good link bottleneck assessment. Therefore, the optimal path is D→G→H→I, distributing the entire 60Mbps.

[0057] User 4, traffic is sent from B to H, bandwidth usage is 100Mbps, medium priority traffic processing: the following two candidate scheduling paths are identified: Path B→E→H, potential energy sequence [0.74, 0.62], maximum potential energy is 0.74, total potential energy is 1.36, maximum potential energy is less than the fourth threshold of 0.8, link bottleneck assessment: moderate. Path The potential energy sequence for the path B→A→D→G→H is [0.50, 0.37, 0.13, 0.19]. The maximum potential energy is 0.5, and the total potential energy is 1.19. The maximum potential energy is less than the second threshold of 0.6, indicating a good link bottleneck assessment. In summary, the main path, B→E→H, distributes 60Mbps due to its short hop count and high efficiency, allowing for greater traffic distribution. The offloading path, B→A→D→G→H, distributes 40Mbps.

[0058] Analysis of load changes on critical links: Link BC, original load 800Mbps, added 80Mbps, predicted load 880Mbps, potential energy 0.88 (L3 warning); CF, original load 900Mbps, no new traffic, predicted load 900Mbps, potential energy 0.9 (L4 warning); EF, original load 700Mbps, added 120Mbps, predicted load 820Mbps, potential energy 0.82 (L3 warning); BE, original load 600Mbps, added 150Mbps, predicted load 750Mbps, potential energy 0.75 (L2 warning); DG, original load 100Mbps, added 140Mbps, predicted load 240Mbps, potential energy 0.24 (L0 normal); GH, original load 150Mbps, added 140Mbps, predicted load 290Mbps, potential energy 0.29 (L0 normal).

[0059] Using the above traffic scheduling method Load balancing: Successfully distributed the additional 360Mbps of traffic across 10 links; Congestion avoidance: All link utilization is kept below 90%, and no new congestion is generated. Business assurance: Different priority businesses receive differentiated services; Resource utilization: Idle links (DG, GH, etc.) are effectively utilized; Preventive control: An early warning is issued when the link reaches 70%; Global optimization: Although decisions are made locally, global load balancing is achieved through potential energy propagation.

[0060] In a specific embodiment, assuming α = 0.3, for network node B, according to the network topology, B's directly connected neighbor links are: BA (potential energy 0.5), BC (potential energy 0.99), and BE (potential energy 0.74). The maximum potential energy value of the directly connected links in this round is 0.99. B's directly connected neighbor nodes are A, C, and E. Therefore, it is necessary to calculate the node potential energy of these directly connected neighbor nodes in this round: A's potential energy calculation: where the directly connected links AB (potential energy 0.5) and AD (potential energy 0.37), the maximum potential energy value of the directly connected links is 0.5. Considering that A's directly connected neighbor nodes also include B, in the actual system, due to the interdependence of node potential energy calculations, the above calculation requires multiple rounds of iteration. Therefore, in the first round of calculation, all nodes use the maximum potential energy value of the directly connected links as the initial node potential energy value, that is, A's initial node potential energy value is 0.5. C's potential energy calculation: directly connected links CB (potential energy 0.99) and CF (potential energy 1.11), the maximum potential energy value of the directly connected links is 1.11. Potential energy calculation for node E: Direct links EB (potential energy 0.74), ED (potential energy 0.25), EF (potential energy 0.87), EH (potential energy 0.62), therefore the maximum potential energy of the direct links is 0.87. The average potential energy of B's ​​directly connected neighbor nodes (A, C, E) is (0.5 + 1.11 + 0.87) / 3 = 0.827, therefore B's current node potential energy is max(0.99, 0.3 × 0.827) = 0.99. The above method can be used to calculate the node potential energy of other network nodes.

[0061] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.

[0062] Embodiments of this application also provide a traffic scheduling device, such as... Figure 2As shown, the traffic scheduling device includes: a potential energy calculation module, which is used to: acquire bandwidth information of network links; and calculate a first link potential energy value of the network links based on the bandwidth information; a traffic navigation module, which is used to: respond to receiving a traffic scheduling instruction, determine candidate scheduling paths for the traffic to be scheduled based on the start node and end node in the traffic scheduling instruction; use the network links that make up the candidate scheduling paths as candidate network links, and determine a target scheduling path from the candidate scheduling paths based on the first link potential energy value of the candidate network links; and an execution control module, which is used to: use the network links that make up the target scheduling path as target network links, and schedule the traffic to be scheduled according to the target network links; use the scheduled target network links as network links to be updated, and calculate a second link potential energy value of the network links to be updated based on the first link potential energy value of the network links to be updated; determine the scheduling warning level of the network links to be updated based on the second link potential energy value of the network links to be updated, and perform corresponding scheduling control operations on the network links to be updated based on the scheduling warning level.

[0063] Furthermore, the bandwidth information includes the current utilization rate, the historical average utilization rate within the historical scheduling period, and the historical maximum utilization rate within the current scheduling period. The potential energy calculation module is also used to: use the current utilization rate as the first component of the first link potential energy value of the network link; calculate the utilization rate offset based on the current utilization rate and the historical average utilization rate, and calculate the second component of the first link potential energy value of the network link based on the utilization rate offset and the historical maximum utilization rate; determine the current prediction factor of the network link, and calculate the third component of the first link potential energy value of the network link based on the first component, the second component, the current prediction factor, and the preset step size coefficient; and determine the first link potential energy value of the network link based on the weighted sum of the first component, the second component, and the third component.

[0064] Furthermore, the potential energy calculation module is also used to: obtain the previous reference value and the previous prediction factor of the network link; calculate the current reference value based on the first and second components, and calculate the reference value change based on the current reference value and the previous reference value; calculate the current prediction factor based on the reference value change and the previous prediction factor, and calculate the prediction adjustment based on the current prediction factor and the preset step size coefficient; and calculate the third component based on the prediction adjustment and the current reference value.

[0065] In a specific implementation, the potential energy calculation module's workflow comprises three key stages: First, multi-source data is acquired to monitor link throughput, queue depth, and packet loss rate in real time, aiming to obtain the link's bandwidth information over a given period. Next, the calculations of the three components are performed in parallel. Finally, a weighted fusion is performed to output a standardized potential energy value. The entire process employs an incremental calculation method, updating only the changing quantities in each period to ensure minimal computational overhead.

[0066] Furthermore, the traffic scheduling device also includes a potential energy propagation and diffusion module. The potential energy calculation module and the potential energy propagation and diffusion module are used to: determine the directly connected neighbor nodes of the network node, and take the network link between the network node and the directly connected neighbor node as the directly connected neighbor link, wherein the directly connected neighbor node is adjacent to the network node and directly connected to the network node; obtain the first link potential energy value of the directly connected neighbor link, and determine the maximum value of the first link potential energy value in the directly connected neighbor link as the maximum potential energy value of the network node in this round of direct connection links; obtain the node potential energy value of the directly connected neighbor node in this round, and calculate the average potential energy value of the network node in this round of direct connection links based on the node potential energy value of the directly connected neighbor node in this round of direct connection links; and calculate the average potential energy value of the directly connected nodes in this round of direct connection links based on the average potential energy value of the directly connected nodes in this round and the pre- Define an attenuation factor and calculate the current-round propagation potential energy value of the network node's directly connected nodes. Determine the maximum value between the current-round maximum potential energy value of the direct links and the current-round propagation potential energy value of the directly connected nodes as the current-round node potential energy value of the network node. If the absolute value of the difference between the current-round node potential energy value and the previous-round node potential energy value of the network node is less than or equal to a preset convergence threshold, stop the iteration and determine the current-round node potential energy value of the network node as its current node potential energy value. If the absolute value of the difference is greater than the preset convergence threshold, send the current-round node potential energy value of the network node to its directly connected neighbor nodes to update their node potential energy values, and proceed to the next iteration until the absolute value of the difference is less than or equal to the preset convergence threshold.

[0067] In a specific implementation, the potential energy propagation module transforms the potential energy state of local network links into a consistent situational awareness across the entire network.

[0068] Furthermore, the traffic navigation module is also used to: obtain the network topology and query the reachable scheduling path from the starting node to the ending node from the network topology; identify the network links that make up the reachable scheduling path as reachable network links and obtain the total number of reachable network links; compare the total number of reachable network links with a preset link number threshold; determine the reachable scheduling path as a candidate scheduling path in response to the total number of reachable network links being less than or equal to the link number threshold; and determine the reachable scheduling path as a non-candidate scheduling path in response to the total number of reachable network links being greater than the link number threshold.

[0069] Furthermore, the traffic navigation module is also used to: calculate the total path potential energy of the candidate scheduling paths based on the first link potential energy value of the candidate network links; match the candidate scheduling paths with other candidate scheduling paths to obtain the number of links overlapping in the candidate scheduling paths; obtain the total number of links in the candidate scheduling paths, and determine the target scheduling path from the candidate scheduling paths based on at least one of the following data: the total path potential energy of the candidate scheduling paths, the number of links overlapping, and the total number of links.

[0070] Furthermore, the traffic navigation module is also used to: in response to the traffic to be scheduled being of low priority, determine the candidate scheduling path with the smallest total number of links as the intermediate scheduling path; in response to the intermediate scheduling path being unique, determine the intermediate scheduling path as the target scheduling path and use the target scheduling path as the main scheduling path; in response to the intermediate scheduling path being not unique, determine the intermediate scheduling path with the lowest total path potential as the target scheduling path and use the target scheduling path as the main scheduling path.

[0071] Furthermore, the traffic navigation module is also used to: in response to the traffic to be scheduled being of medium priority, determine the candidate scheduling path with the smallest total number of links as the intermediate scheduling path; in response to the intermediate scheduling path being unique, determine the intermediate scheduling path as the primary scheduling path in the target scheduling path, and determine the candidate scheduling path with the lowest total path potential energy as the backup scheduling path in the target scheduling path, wherein the primary scheduling path and the backup scheduling path are different scheduling paths; in response to the intermediate scheduling path being not unique, determine the intermediate scheduling path with the lowest total path potential energy as the primary scheduling path, and determine the candidate scheduling path with the lowest total path potential energy as the backup scheduling path among the other candidate scheduling paths besides the primary scheduling path.

[0072] Furthermore, the traffic navigation module is also used to: in response to the traffic to be scheduled being of high priority, determine the candidate scheduling path with the smallest total number of links as the intermediate scheduling path; in response to the intermediate scheduling path being unique, determine the intermediate scheduling path as the primary scheduling path in the target scheduling path, and determine the candidate scheduling path with the fewest links overlapping with the primary scheduling path as the backup scheduling path in the target scheduling path; in response to the intermediate scheduling path being not unique, determine the intermediate scheduling path with the lowest total path potential as the primary scheduling path, and determine the candidate scheduling path with the fewest links overlapping with the primary scheduling path as the backup scheduling path.

[0073] In a specific implementation, the traffic navigation module is responsible for converting potential energy field information into actual routing decisions. Based on the gradient descent principle and multi-path optimization theory, this module achieves intelligent guidance and load balancing of data flows. Its working principle is as follows: The traffic navigation module calculates the potential energy of directly connected neighbor links at each network node, selecting the direction with the fastest potential energy decrease with high probability, while retaining a small probability of exploring other paths, effectively balancing the contradiction between local optima and global exploration. Multi-path exploration is designed for large data flows, using the K-shortest path algorithm to find multiple edge-separated paths, and allocating traffic weights based on the sum of the potential energy of each path to maximize throughput. The obstacle avoidance router establishes a dynamic potential energy barrier mechanism. When a link's potential energy exceeds a preset danger threshold (the fifth threshold of 0.9), it is automatically marked as an obstacle and assigned an extremely high temporary potential energy, such as 0.99, ensuring that new traffic completely bypasses severely congested areas. Path evaluation continuously monitors the performance indicators of each active path, providing quantitative decision-making basis for the components.

[0074] Furthermore, the potential energy calculation module is also used to: in response to the target scheduling path including a primary scheduling path and a backup scheduling path, calculate the total path potential energy of the target scheduling path based on the total path potential energy of the primary scheduling path and the total path potential energy of the backup scheduling path; calculate the first allocation ratio of the primary scheduling path based on the total path potential energy of the target scheduling path and the total path potential energy of the primary scheduling path, and calculate the first allocation amount of the primary scheduling path based on the bandwidth occupancy of the traffic to be scheduled and the first allocation ratio; calculate the second allocation ratio of the backup scheduling path based on the total path potential energy of the target scheduling path and the total path potential energy of the backup scheduling path, and calculate the second allocation amount of the backup scheduling path based on the bandwidth occupancy and the second allocation ratio; in response to the target scheduling path including the primary scheduling path but not the backup scheduling path, determine the bandwidth occupancy as the allocation amount of the primary scheduling path.

[0075] Furthermore, the potential energy calculation module is also used to: in response to the target scheduling path including a primary scheduling path and a backup scheduling path, to use the network links constituting the primary scheduling path as primary scheduling network links, and to calculate the second link potential energy value of the primary scheduling network links based on the first allocation and the first link potential energy value of the primary scheduling network links; to use the network links constituting the backup scheduling path as backup scheduling network links, and to calculate the second link potential energy value of the backup scheduling network links based on the second allocation and the first link potential energy value of the backup scheduling network links; and in response to the target scheduling path including the primary scheduling path but not the backup scheduling path, to calculate the second link potential energy value of the primary scheduling network links based on the bandwidth occupancy and the first link potential energy value of the primary scheduling network links.

[0076] Further, the execution control module is also configured to: compare the second link potential energy value of the primary scheduling network link or the backup scheduling network link with a preset potential energy interval threshold to determine the potential energy interval to which the second link potential energy value of the primary scheduling network link or the backup scheduling network link belongs; in response to the second link potential energy value of the primary scheduling network link or the backup scheduling network link being greater than or equal to the first threshold and less than the second threshold, determine that the second link potential energy value of the primary scheduling network link or the backup scheduling network link belongs to the first potential energy interval, and determine that the scheduling warning level of the primary scheduling network link or the backup scheduling network link is the first warning level; in response to the second link potential energy value of the primary scheduling network link or the backup scheduling network link being greater than or equal to the second threshold and less than the third threshold, determine that the second link potential energy value of the primary scheduling network link or the backup scheduling network link belongs to the second potential energy interval, and determine that the scheduling warning level of the primary scheduling network link or the backup scheduling network link is the second warning level; in response to the second link potential energy value of the primary scheduling network link or the backup scheduling network link being greater than or equal to the third threshold and If the potential energy value of the second link of the primary or backup scheduling network link is less than the fourth threshold, it is determined that the second link potential energy value belongs to the third potential energy range, and the scheduling warning level of the primary or backup scheduling network link is determined to be the third warning level. If the potential energy value of the second link of the primary or backup scheduling network link is greater than or equal to the fourth threshold and less than the fifth threshold, it is determined that the second link potential energy value belongs to the fourth potential energy range, and the scheduling warning level of the primary or backup scheduling network link is determined to be the fourth warning level. If the potential energy value of the second link of the primary or backup scheduling network link is greater than or equal to the fifth threshold and less than the sixth threshold, it is determined that the second link potential energy value belongs to the fifth potential energy range, and the scheduling warning level of the primary or backup scheduling network link is determined to be the fifth warning level. Among them, the warning levels are ranked from low to high according to their urgency: first warning level, second warning level, third warning level, fourth warning level, and fifth warning level.

[0077] Furthermore, the execution control module is also configured to: in response to a scheduling warning level of the primary or backup scheduling network link being a first warning level, perform a first control operation on the primary or backup scheduling network link, wherein the first control operation includes maintaining the current link state of the primary or backup scheduling network link; in response to a scheduling warning level of the primary or backup scheduling network link being a second warning level, perform a second control operation on the primary or backup scheduling network link, wherein the second control operation includes sending a path switching suggestion to the network node corresponding to the primary or backup scheduling network link; and in response to a scheduling warning level of the primary or backup scheduling network link being a third warning level, perform a third control operation on the primary or backup scheduling network link. The control operations include: a third control operation, which involves reallocating the traffic to be scheduled through the primary or backup scheduling network link; a fourth control operation, which is performed on the primary or backup scheduling network link in response to a fourth warning level, wherein the fourth control operation includes reducing the bandwidth allocation ratio of low-priority traffic in the traffic to be scheduled through the primary or backup scheduling network link; and a fifth control operation, which is performed on the primary or backup scheduling network link in response to a fifth warning level, wherein the fifth control operation includes dropping data packets of low-priority traffic in the traffic to be scheduled through the primary or backup scheduling network link.

[0078] In a specific implementation, the execution control module is responsible for translating the decisions of the traffic navigation module into specific network control actions, ensuring accurate traffic scheduling and system stability. Furthermore, the execution control module constructs a hierarchical evaluation model for links and a congestion prevention control model. The working principle of the execution control module is analyzed as follows: its internal distributed scheduler performs traffic classification and queue management locally on each network node. Based on the priority labels of the traffic to be scheduled (i.e., data streams), it dynamically adjusts forwarding strategies and scheduling weights to ensure that high-priority traffic receives the necessary quality of service guarantees. The congestion prevention control model implements a five-level early warning response mechanism. When the link potential energy enters different threshold ranges, corresponding control actions are automatically triggered, ranging from L1-level mild rate limiting and path suggestions to L4-level emergency rerouting and selective packet loss, forming a progressive and differentiated congestion response system.

[0079] For a description of the features in the embodiment corresponding to the traffic scheduling device, please refer to the relevant description of the embodiment corresponding to the traffic scheduling method, which will not be repeated here.

[0080] Embodiments of this application also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above-described traffic scheduling method embodiments.

[0081] When the program instructions are read and executed by one or more processors, they can also perform operations corresponding to the steps in the above method embodiments, as described above, and will not be repeated here. Reference Figure 3 This example illustrates the architecture of an electronic device, which may include a processor 310, a video display adapter 311, a disk drive 312, an input / output interface 313, a network interface 314, and a memory 320. The processor 310, video display adapter 311, disk drive 312, input / output interface 313, network interface 314, and memory 320 can communicate with each other via a communication bus 330.

[0082] The processor 310 can be implemented using a general-purpose central processing unit (CPU), microprocessor, application specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to achieve the technical solution provided in this application.

[0083] The memory 320 can be implemented as a read-only memory (ROM), random access memory (RAM), static storage device, dynamic storage device, etc. The memory 320 can store an operating system 321 for controlling the operation of the electronic device 300, and a basic input / output system (BIOS) 322 for controlling the low-level operations of the electronic device 300. Additionally, it can store a web browser 323, data storage management 324, and an icon / font processing system 325. The icon / font processing system 325 can be the application program that specifically implements the aforementioned steps in this embodiment. In summary, when implementing the technical solution provided in this application through software or firmware, the relevant program code is stored in the memory 320 and executed by the processor 310.

[0084] Input / output interface 313 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0085] Network interface 314 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0086] Bus 330 includes a pathway for transmitting information between various components of the device, such as processor 310, video display adapter 311, disk drive 312, input / output interface 313, network interface 314, and memory 320.

[0087] In addition, the electronic device 300 can also obtain information on specific acquisition conditions from the virtual resource object acquisition condition information database (not shown in the figure) for condition judgment.

[0088] It should be noted that although the above-described electronic device 300 only shows a processor 310, a video display adapter 311, a disk drive 312, an input / output interface 313, a network interface 314, a memory 320, and a bus 330, in specific implementations, the electronic device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the solution of this application, and does not necessarily include all the components shown in the figures.

[0089] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause an electronic device (which may be a personal computer, cloud server, or network device, etc.) to execute the methods of various embodiments or some parts of the embodiments of this application.

[0090] Embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above-described traffic scheduling method embodiments at runtime.

[0091] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0092] Embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above-described traffic scheduling method embodiments.

[0093] Embodiments of this application also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in any of the above-described traffic scheduling method embodiments.

[0094] Any of the components, modules, units, parts, methods, and operations described herein can be implemented using software, firmware, hardware (e.g., fixed logic circuitry), manual processing, or any combination thereof. Alternatively or additionally, any functionality described herein can be executed at least in part by one or more hardware logic components, such as, but not limited to, a central processing unit (CPU), a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), an application-specific standard product (ASSP), a system-on-a-chip (SoC), a complex programmable logic device (CPLD), a microprocessor (MCU), etc. The terms "system," "computing device," or "apparatus" as used herein encompass various means, devices, and machines for processing data, including, for example, one or more programmable processors, computers, SoCs, or combinations thereof. The apparatus may also include code that creates an execution environment for the computer program in question, such as code constituting processor firmware, a protocol stack, a database management system, an operating system, a cross-platform runtime environment, a virtual machine, or one or more combinations thereof. The aforementioned computer program (also known as a program, software, software application, app, script, or code) can be written in any form of programming language, including compiled or interpreted languages, declarative or procedural languages, and can be deployed in any form, including as a standalone program or as a module, component, subroutine, object, or other unit suitable for a computing environment.

[0095] Those skilled in the art will further 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, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. 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.

[0096] The foregoing has provided a detailed description of a traffic scheduling method, apparatus, electronic device, and storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A traffic scheduling method, characterized in that, The traffic scheduling method includes: Obtain bandwidth information for network links; Based on the bandwidth information, calculate the first link potential value of the network link; In response to receiving a traffic scheduling instruction, a candidate scheduling path for the traffic to be scheduled is determined based on the start node and end node in the traffic scheduling instruction. The network links that make up the candidate scheduling paths are used as candidate network links, and the target scheduling path is determined from the candidate scheduling paths based on the first link potential value of the candidate network links. The network links that make up the target scheduling path are used as target network links, and the traffic to be scheduled is scheduled according to the target network links. The scheduled target network link is taken as the network link to be updated, and the second link potential value of the network link to be updated is calculated based on the first link potential value of the network link to be updated. Based on the second link potential energy value of the network link to be updated, the scheduling warning level of the network link to be updated is determined, and based on the scheduling warning level, the corresponding scheduling control operation is performed on the network link to be updated. The bandwidth information includes the current utilization rate, the historical average utilization rate within the historical scheduling period, and the historical maximum utilization rate within the current scheduling period. The step of calculating the first link potential value of the network link based on the bandwidth information includes: The current utilization rate is used as the first component of the first link potential value of the network link; Based on the current utilization rate and the historical average utilization rate, calculate the utilization rate offset, and based on the utilization rate offset and the historical maximum utilization rate, calculate the second component of the first link potential value of the network link; Determine the current prediction factor of the network link, and calculate the third component of the first link potential value of the network link based on the first component, the second component, the current prediction factor, and the preset step size coefficient; The first link potential value of the network link is determined based on the weighted sum of the first component, the second component, and the third component.

2. The traffic scheduling method according to claim 1, characterized in that, The step of determining the current prediction factor of the network link and calculating the third component of the first link potential value of the network link based on the first component, the second component, the current prediction factor, and a preset step size coefficient includes: Obtain the previous reference value and the previous prediction factor of the network link; Calculate the current reference value based on the first component and the second component, and calculate the change in the reference value based on the current reference value and the reference value at the previous moment; The current prediction factor is calculated based on the change in the baseline value and the prediction factor at the previous time step, and the prediction adjustment is calculated based on the current prediction factor and the preset step size coefficient. The third component is calculated based on the predicted adjustment amount and the current baseline value.

3. The traffic scheduling method according to claim 2, characterized in that, The step of determining the candidate scheduling path for the traffic to be scheduled based on the start node and end node in the traffic scheduling instruction includes: Obtain the network topology and query the network topology for a reachable scheduling path from the starting node to the ending node; The network links that make up the reachable scheduling path are defined as reachable network links, and the total number of reachable network links is obtained. The total number of reachable network links is compared with a preset link number threshold. In response to the total number of reachable network links being less than or equal to the link number threshold, the reachable scheduling path is determined as the candidate scheduling path; In response to the total number of reachable network links being greater than the link number threshold, the reachable scheduling path is determined as a non-candidate scheduling path.

4. The traffic scheduling method according to claim 3, characterized in that, The step of determining the target scheduling path from the candidate scheduling paths based on the first link potential value of the candidate network links includes: Based on the first link potential value of the candidate network link, calculate the total path potential of the candidate scheduling path; The candidate scheduling path is matched with other candidate scheduling paths besides the candidate scheduling path to obtain the number of links overlapping for the candidate scheduling paths; Obtain the total number of links in the candidate scheduling paths, and determine the target scheduling path from the candidate scheduling paths based on at least one of the following data: the total path potential energy of the candidate scheduling paths, the number of overlapping links, and the total number of links.

5. The traffic scheduling method according to claim 4, characterized in that, The step of determining the target scheduling path from the candidate scheduling paths based on at least one of the following data: the total path potential energy, the number of overlapping links, and the total number of links, includes: In response to the fact that the traffic to be scheduled is of low priority, the candidate scheduling path with the smallest total number of links is determined as the intermediate scheduling path; Since the intermediate scheduling path is unique, the intermediate scheduling path is determined as the target scheduling path, and the target scheduling path is used as the primary scheduling path. In response to the fact that the intermediate scheduling path is not unique, the intermediate scheduling path with the lowest total potential energy is determined as the target scheduling path, and the target scheduling path is used as the main scheduling path.

6. The traffic scheduling method according to claim 4, characterized in that, The step of determining the target scheduling path from the candidate scheduling paths based on at least one of the following data: the total path potential energy of the candidate scheduling paths, the number of overlapping links, and the total number of links, further includes: In response to the fact that the traffic to be scheduled belongs to medium priority, the candidate scheduling path with the smallest total number of links is determined as the intermediate scheduling path; In response to the uniqueness of the intermediate scheduling path, the intermediate scheduling path is determined as the primary scheduling path in the target scheduling path, and the candidate scheduling path with the lowest total path potential energy is determined as the backup scheduling path in the target scheduling path, wherein the primary scheduling path and the backup scheduling path belong to different scheduling paths; In response to the fact that the intermediate scheduling path is not unique, the intermediate scheduling path with the lowest total path potential energy is determined as the primary scheduling path, and among the other candidate scheduling paths besides the primary scheduling path, the candidate scheduling path with the lowest total path potential energy is determined as the backup scheduling path.

7. The traffic scheduling method according to claim 4, characterized in that, The step of determining the target scheduling path from the candidate scheduling paths based on at least one of the following data: the total path potential energy of the candidate scheduling paths, the number of overlapping links, and the total number of links, further includes: In response to the fact that the traffic to be scheduled belongs to high priority, the candidate scheduling path with the smallest total number of links is determined as the intermediate scheduling path; In response to the uniqueness of the intermediate scheduling path, the intermediate scheduling path is determined as the primary scheduling path in the target scheduling path, and the candidate scheduling path with the fewest links overlapping with the primary scheduling path is determined as the backup scheduling path in the target scheduling path. In response to the fact that the intermediate scheduling path is not unique, the intermediate scheduling path with the lowest total potential energy is determined as the primary scheduling path, and the candidate scheduling path with the fewest links overlapping with the primary scheduling path is determined as the backup scheduling path.

8. The traffic scheduling method according to any one of claims 5 to 7, characterized in that, The step of scheduling the traffic to be scheduled according to the target network link includes: In response to the target scheduling path including the primary scheduling path and the backup scheduling path, the total path potential energy of the target scheduling path is calculated based on the total path potential energy of the primary scheduling path and the total path potential energy of the backup scheduling path. Based on the sum of the path potential energy of the target scheduling path and the sum of the path potential energy of the main scheduling path, the first allocation ratio of the main scheduling path is calculated, and based on the bandwidth occupancy of the traffic to be scheduled and the first allocation ratio, the first allocation amount of the main scheduling path is calculated. Based on the sum of the path potential energy of the target scheduling path and the sum of the path potential energy of the backup scheduling path, calculate the second allocation ratio of the backup scheduling path, and based on the bandwidth occupancy and the second allocation ratio, calculate the second allocation amount of the backup scheduling path. In response to the target scheduling path including the primary scheduling path but excluding the backup scheduling path, the bandwidth usage is determined as the allocation amount for the primary scheduling path.

9. The traffic scheduling method according to claim 8, characterized in that, The step of calculating the second link potential value of the network link to be updated based on the first link potential value of the network link to be updated includes: In response to the target scheduling path including the primary scheduling path and the backup scheduling path, the network links constituting the primary scheduling path are designated as primary scheduling network links, and the second link potential value of the primary scheduling network link is calculated based on the first allocation amount and the first link potential value of the primary scheduling network link. The network links that make up the backup scheduling path are used as backup scheduling network links, and the second link potential value of the backup scheduling network link is calculated based on the second allocation amount and the first link potential value of the backup scheduling network link. In response to the target scheduling path including the primary scheduling path but excluding the backup scheduling path, the second link potential value of the primary scheduling network link is calculated based on the bandwidth occupancy and the first link potential value of the primary scheduling network link.

10. The traffic scheduling method according to claim 9, characterized in that, The step of determining the scheduling warning level of the network link to be updated based on the second link potential value of the network link to be updated includes: The potential energy value of the second link of the primary scheduling network link or the backup scheduling network link is compared with a preset potential energy range threshold to determine the potential energy range to which the potential energy value of the second link of the primary scheduling network link or the backup scheduling network link belongs. In response to the second link potential energy value of the primary scheduling network link or the backup scheduling network link being greater than or equal to a first threshold and less than a second threshold, it is determined that the second link potential energy value of the primary scheduling network link or the backup scheduling network link belongs to a first potential energy range, and the scheduling warning level of the primary scheduling network link or the backup scheduling network link is determined to be a first warning level. In response to the second link potential energy value of the primary scheduling network link or the backup scheduling network link being greater than or equal to the second threshold and less than the third threshold, it is determined that the second link potential energy value of the primary scheduling network link or the backup scheduling network link belongs to the second potential energy range, and the scheduling warning level of the primary scheduling network link or the backup scheduling network link is determined to be the second warning level. In response to the second link potential energy value of the primary scheduling network link or the backup scheduling network link being greater than or equal to the third threshold and less than the fourth threshold, it is determined that the second link potential energy value of the primary scheduling network link or the backup scheduling network link belongs to the third potential energy range, and the scheduling warning level of the primary scheduling network link or the backup scheduling network link is determined to be the third warning level. In response to the second link potential energy value of the primary scheduling network link or the backup scheduling network link being greater than or equal to the fourth threshold and less than the fifth threshold, it is determined that the second link potential energy value of the primary scheduling network link or the backup scheduling network link belongs to the fourth potential energy range, and the scheduling warning level of the primary scheduling network link or the backup scheduling network link is determined to be the fourth warning level. In response to the second link potential energy value of the primary scheduling network link or the backup scheduling network link being greater than or equal to the fifth threshold and less than the sixth threshold, it is determined that the second link potential energy value of the primary scheduling network link or the backup scheduling network link belongs to the fifth potential energy range, and the scheduling warning level of the primary scheduling network link or the backup scheduling network link is determined to be the fifth warning level. The warning levels are arranged in order of increasing urgency: the first warning level, the second warning level, the third warning level, the fourth warning level, and the fifth warning level.

11. The traffic scheduling method according to claim 10, characterized in that, The step of performing corresponding scheduling control operations on the network link to be updated according to the scheduling warning level includes: In response to the scheduling warning level of the primary scheduling network link or the backup scheduling network link being the first warning level, a first control operation is performed on the primary scheduling network link or the backup scheduling network link, wherein the first control operation includes maintaining the current link state of the primary scheduling network link or the backup scheduling network link; In response to the scheduling warning level of the primary scheduling network link or the backup scheduling network link being the second warning level, a second control operation is performed on the primary scheduling network link or the backup scheduling network link, wherein the second control operation includes sending a path switching suggestion to the network node corresponding to the primary scheduling network link or the backup scheduling network link; In response to the scheduling warning level of the primary scheduling network link or the backup scheduling network link being the third warning level, a third control operation is performed on the primary scheduling network link or the backup scheduling network link, wherein the third control operation includes reallocating the traffic to be scheduled that passes through the primary scheduling network link or the backup scheduling network link. In response to the scheduling warning level of the primary scheduling network link or the backup scheduling network link being the fourth warning level, a fourth control operation is performed on the primary scheduling network link or the backup scheduling network link, wherein the fourth control operation includes reducing the bandwidth allocation ratio of low-priority traffic in the traffic to be scheduled passing through the primary scheduling network link or the backup scheduling network link. In response to the scheduling warning level of the primary scheduling network link or the backup scheduling network link being the fifth warning level, a fifth control operation is performed on the primary scheduling network link or the backup scheduling network link, wherein the fifth control operation includes discarding data packets of low-priority traffic in the traffic to be scheduled passing through the primary scheduling network link or the backup scheduling network link.

12. A traffic scheduling apparatus for implementing the traffic scheduling method of claim 1, characterized in that, The traffic scheduling device includes: Potential energy calculation module, the potential energy calculation module is used for: Obtain bandwidth information for network links; Based on the bandwidth information, calculate the first link potential value of the network link; Traffic navigation module, the traffic navigation module is used for: In response to receiving a traffic scheduling instruction, a candidate scheduling path for the traffic to be scheduled is determined based on the start node and end node in the traffic scheduling instruction. The network links that make up the candidate scheduling paths are used as candidate network links, and the target scheduling path is determined from the candidate scheduling paths based on the first link potential value of the candidate network links. The execution control module is used for: The network links that make up the target scheduling path are used as target network links, and the traffic to be scheduled is scheduled according to the target network links. The scheduled target network link is taken as the network link to be updated, and the second link potential value of the network link to be updated is calculated based on the first link potential value of the network link to be updated. Based on the second link potential value of the network link to be updated, the scheduling warning level of the network link to be updated is determined, and based on the scheduling warning level, the corresponding scheduling control operation is performed on the network link to be updated.

13. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the traffic scheduling method as described in any one of claims 1 to 11.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the steps of the traffic scheduling method as described in any one of claims 1 to 11.

Images