Bandwidth management method, system, electronic device and storage medium for multi-service modal isolation scenarios

Abstract

This application relates to the field of communication network resource management technology, and discloses a bandwidth management method, system, electronic device, and storage medium for multi-service modal isolation scenarios. The method includes: dividing service modes according to the service level agreement requirements of the services and determining the target mode and interfering modes; obtaining the actual performance data of the target mode under the condition of coexistence of interfering modes and its baseline performance data under the condition of no interfering mode disturbance; determining the real-time isolation degree based on the deviation between the actual performance data and the baseline performance data; and adjusting the hard-pipe bandwidth allocation corresponding to the target service when the real-time isolation degree meets a preset threshold condition. This application can objectively quantify the actual mutual interference degree between multiple services and adaptively adjust the bandwidth resource allocation accordingly, ensuring the service quality of critical services while improving the utilization rate of hard-pipe resources.

Description

Technical Field

[0001] This application relates to the field of communication network resource management technology, specifically to bandwidth management methods, systems, electronic devices, and storage media in multi-service modal isolation scenarios. Background Technology

[0002] In multi-service network scenarios such as power communication networks, services like relay protection, video inspection, and remote control have significantly different requirements for performance indicators such as bandwidth, latency, jitter, and packet loss rate. Related technologies primarily employ methods such as Quality of Service (QoS) scheduling, traffic engineering, network slicing, and Flexible Ethernet (FlexE) hard pipe isolation for resource management. However, QoS scheduling and network slicing are soft isolation mechanisms, meaning different services still share underlying ports and cache resources, making it difficult to guarantee deterministic transmission of critical services under sudden traffic surges or congestion. While FlexE hard pipes provide strong physical layer isolation, their bandwidth configuration typically relies on static planning or manual adjustments, lacking the ability to dynamically adapt to the operational status of services.

[0003] Furthermore, isolation assessment methods in related technologies typically rely on manually setting the weights of performance indicators or fixed thresholds, making it difficult to objectively reflect the true sensitivity of different services to each indicator under actual disturbances. During bandwidth adjustment, the hop-by-hop negotiation method between multi-hop nodes is prone to causing adjustment failures or inconsistent effective times for some nodes, and lacks an effective failure rollback mechanism. Therefore, how to objectively quantify the degree of mutual interference between services under multi-service shared network conditions, automatically generate reasonable isolation assessment parameters and adjustment thresholds, and achieve synchronous and reliable adjustment of end-to-end hard-pipe bandwidth remains an urgent problem to be solved. Summary of the Invention

[0004] This application provides a bandwidth management method, system, electronic device, and storage medium for multi-service modal isolation scenarios, in order to solve the problem in related technologies that it is difficult to objectively quantify the actual degree of mutual interference between multiple services and that bandwidth resource allocation cannot be adaptively adjusted based on the quantification results.

[0005] Firstly, this application provides a bandwidth management method, comprising: dividing the services in the network into multiple service modes according to the service level agreement requirements of each service carried in the network; determining the service mode to which a target service belongs from the multiple service modes as the target mode; and designating at least one service mode other than the target mode from the multiple service modes as an interfering mode; acquiring actual performance data of the target mode under the coexistence condition of the interfering mode, and baseline performance data of the target mode under the condition of no disturbance from the interfering mode; determining the real-time isolation degree of the interfering mode to the target mode based on the deviation between the actual performance data and the baseline performance data; and adjusting the hard pipe bandwidth allocation corresponding to the target service when the real-time isolation degree meets a preset isolation threshold condition.

[0006] Beneficial Effects: The bandwidth management method provided in this application divides services into multiple service modes according to the service level agreement requirements of each service carried in the network, and determines the target mode and interfering modes. It uses the deviation between the actual performance data of the target mode under the condition of coexistence of interfering modes and its baseline performance data under the condition of no interfering mode disturbance to determine the real-time isolation degree of the interfering mode from the target mode. When the real-time isolation degree meets the preset isolation degree threshold, the hard pipe bandwidth allocation corresponding to the target service is adjusted. This objectively quantifies the actual degree of mutual interference between multiple services and adaptively adjusts bandwidth resource allocation accordingly. Compared with related technologies that rely on manually setting performance index weights and statically configuring hard pipe bandwidth, this application avoids the subjectivity of manual settings, enabling bandwidth adjustment to be dynamically triggered according to the service operation status. This ensures the service quality of critical services when interfered with, and avoids the continuous idleness of hard pipe resources.

[0007] In one optional implementation, acquiring the actual performance data of the target mode under the coexistence condition of the interfering modes includes: determining the length of a preset sampling period based on the priority of the target mode, wherein the higher the priority, the shorter the corresponding sampling period; collecting the performance data of the target mode at the preset sampling period, including at least one of the following methods: acquiring the end-to-end latency, jitter, and packet loss rate of the target mode based on a bidirectional active measurement protocol; extracting the time slot occupancy rate from the flexible Ethernet overhead based on the operation, management, and maintenance overhead mechanism of the flexible Ethernet protocol, and determining the bandwidth occupancy data based on the time slot occupancy rate; and determining the actual service utilization rate of the target mode based on port traffic statistics or service traffic statistics.

[0008] Beneficial Effects: The bandwidth management method provided in this application determines the preset sampling period based on the priority of the target modality. Higher-priority services have shorter sampling periods, enabling more timely monitoring and response to high-priority services. Simultaneously, it acquires end-to-end latency, jitter, and packet loss rate through a bidirectional active measurement protocol, and extracts time slot occupancy from the flexible Ethernet overhead through the operation, management, and maintenance overhead mechanism of the flexible Ethernet protocol to determine bandwidth usage data. These two acquisition methods complement each other and can be used individually or in combination. The latter method acquires bandwidth usage data without requiring additional probe traffic. Furthermore, determining the actual service utilization rate through port traffic statistics or service traffic statistics allows for a more accurate understanding of the true bandwidth usage of the target service. This solution ensures both the flexibility and comprehensiveness of performance data acquisition while avoiding additional interference from measurement traffic to the service itself.

[0009] In one optional implementation, determining the real-time isolation degree of the interference mode from the target mode based on the deviation between the actual performance data and the benchmark performance data includes: determining the deviation values ​​corresponding to each performance index of the target mode based on the actual performance data and the benchmark performance data, wherein the performance indexes include at least two of bandwidth, latency, jitter, and packet loss rate; determining the isolation coefficient corresponding to each performance index; performing dimensionless normalization on each deviation value and then weighting and fusing the deviation values ​​based on the isolation coefficients; and taking the maximum value in the fusion result as the real-time isolation degree.

[0010] Beneficial Effects: The bandwidth management method provided in this application determines the deviation values ​​corresponding to each performance index of the target mode based on actual performance data and benchmark performance data. Then, it performs dimensionless normalization on each deviation value to eliminate scale differences between different dimensions, determines the isolation coefficient corresponding to each performance index, and performs weighted fusion of each deviation value based on the isolation coefficient. The maximum value in the fusion result is used as the real-time isolation degree. This method can uniformly quantify multi-dimensional performance deviations such as bandwidth, latency, jitter, and packet loss rate into a comprehensive isolation degree index. The use of the maximum value corresponds to the Chebyshev norm, ensuring that the real-time isolation degree can reflect the most severe degradation of any performance dimension, effectively avoiding the risk of a performance index being severely degraded while the overall value remains within the normal range.

[0011] In one optional implementation, determining the isolation coefficients corresponding to each performance indicator includes: acquiring multiple historical samples of the target modality, each historical sample including a historical deviation vector of the target modality at a historical time and a corresponding service degradation label; determining the sample weights corresponding to each historical sample based on the service level agreement (SLA) constraint values ​​of each performance indicator; wherein historical samples close to the SLA critical state are assigned higher sample weights; training a sparse regression model with non-negative constraints using the historical deviation vector as input and the service degradation label as the training target, the sparse regression model having a contribution coefficient, a sparse regularization term, and a smoothing regularization term; wherein the magnitude of the contribution coefficient is used to characterize the degree of contribution of each performance indicator to the service degradation label, the sparse regularization term is used to highlight the performance indicator with the greatest impact on the service degradation label, and the smoothing regularization term is used to suppress the oscillations of the contribution coefficient; and normalizing the contribution coefficients to obtain the isolation coefficients corresponding to each performance indicator.

[0012] Beneficial Effects: The bandwidth management method provided in this application acquires multiple historical samples of the target modality and determines the sample weights corresponding to each historical sample based on the service level agreement (SLA) constraint values. This gives higher weights to historical samples approaching the SLA critical state, thereby guiding the model to focus more on the critical state where the service is approaching the degradation boundary. Then, a sparse regression model with non-negative constraints is trained using the historical deviation vector as input and the service degradation label as the training target. The contribution coefficient represents the degree of contribution of each performance indicator to degradation. Sparse regularization highlights the key performance indicators with the greatest impact on degradation, while smoothing regularization suppresses the oscillation of the contribution coefficient. Finally, the contribution coefficient is normalized to obtain the isolation coefficient. This scheme allows the isolation coefficient to be automatically generated entirely from actual service disturbance samples and performance contribution attribution, avoiding the experience and subjectivity of manually setting weights. This makes the isolation evaluation parameters more closely match the real sensitivity characteristics of the target service in different network environments. Simultaneously, the non-negative constraint ensures that the contribution direction of each performance indicator degradation to service degradation is consistent, sparse regularization enables automatic selection of key performance indicators, and normalization makes the isolation coefficients between different performance indicators comparable.

[0013] In one optional implementation, determining the isolation coefficient corresponding to each of the performance indicators includes: acquiring multiple historical samples of the target modality, each historical sample including a historical deviation vector of the target modality at a historical time and a corresponding service degradation label; training a nonlinear prediction model using the historical deviation vector as input and the service degradation label as the training target; calculating the average contribution of each performance indicator to the service degradation label using a feature contribution method; and normalizing each average contribution to obtain the isolation coefficient corresponding to each performance indicator.

[0014] Beneficial effects: The bandwidth management method provided in this application obtains multiple historical samples of the target mode, trains a nonlinear prediction model with the historical deviation vector as input and the service degradation label as the training target, and calculates the average contribution of each performance index to the degradation label using the feature contribution method. Finally, the isolation coefficient is obtained by normalization. It can effectively characterize the complex mapping relationship between each performance index and service degradation when there is a nonlinear relationship between service performance degradation and performance deviation. It makes up for the lack of expressive power of the linear model, so that the generation of the isolation coefficient can accurately reflect the actual contribution of each performance index under different network conditions.

[0015] In one optional implementation, obtaining multiple historical samples of the target modality includes: determining the prior importance coefficients corresponding to each performance indicator based on the service level agreement constraint value of the target service; and calculating the degree of degradation corresponding to each historical sample based on the historical deviation vector in each historical sample and the corresponding prior importance coefficient, as the service degradation label.

[0016] Beneficial effects: The bandwidth management method provided in this application determines the prior importance coefficients corresponding to each performance index based on the service level agreement constraint value of the target service, and then calculates the degree of degradation corresponding to each historical sample as a service degradation label based on the historical deviation vector in each historical sample and the corresponding prior importance coefficient. This makes the generated degradation label reflect the actual deviation of each performance index relative to the benchmark value, and also incorporates the prior information of the sensitivity of different services to different performance indexes. For example, the contribution of latency deviation to degradation of latency-sensitive services is amplified, thereby providing a high-quality training target with business semantics for the subsequent supervised learning of isolation coefficients, and further improving the accuracy of isolation coefficient generation.

[0017] In one optional implementation, the preset isolation threshold condition includes an expansion threshold determined based on service level agreement (SLP) critical samples in the historical samples whose degradation level reaches the allowable upper limit, and a reduction threshold determined based on samples in the historical samples whose degradation level meets preset excellent conditions. When the real-time isolation meets the preset isolation threshold condition, adjusting the hard pipe bandwidth allocation corresponding to the target service includes: increasing the hard pipe bandwidth corresponding to the target service when the real-time isolation is greater than the expansion threshold; decreasing the hard pipe bandwidth corresponding to the target service when the real-time isolation is less than the reduction threshold; and maintaining the hard pipe bandwidth corresponding to the target service unchanged when the real-time isolation is greater than or equal to the reduction threshold and less than or equal to the expansion threshold.

[0018] Beneficial Effects: The bandwidth management method provided in this application determines the expansion threshold based on service level agreement (SLP) critical samples whose degradation level reaches the allowable upper limit in historical samples, and determines the reduction threshold based on samples whose degradation level meets preset excellent conditions. This ensures that both thresholds are derived from actual business operation samples rather than manually fixed values. Based on this, the real-time isolation degree is compared with both the expansion and reduction thresholds. When the isolation degree is greater than the expansion threshold, the bandwidth is increased; when it is less than the reduction threshold, the bandwidth is decreased; and when it is between the two, the bandwidth remains unchanged. This achieves adaptive triggering and precise control of bandwidth adjustment. The setting of the reduction threshold below the expansion threshold naturally creates a hysteresis interval, avoiding frequent adjustments triggered when the real-time isolation degree fluctuates slightly around the threshold, thus improving system stability.

[0019] In one optional implementation, adjusting the hard-pipe bandwidth allocation corresponding to the target service further includes: generating a version number and a unified effective time for this adjustment; sending a reservation request message in parallel to each forwarding node on the end-to-end path corresponding to the target service, the reservation request message carrying the version number, the number of time slots to be adjusted, and the unified effective time; each forwarding node pre-locking a candidate time slot in its local inactive candidate calendar table, and returning a confirmation message after successful locking; when all forwarding nodes on the end-to-end path return the confirmation message, sending a commit message to all forwarding nodes; under the control of the commit message, each forwarding node switching the inactive candidate calendar table to the effective calendar table within the unified effective time to complete the hard-pipe bandwidth allocation adjustment for the target service.

[0020] Beneficial Effects: The bandwidth management method provided in this application generates a version number and a unified effective time, and sends a reservation request message carrying the version number, the number of time slots to be adjusted, and the unified effective time to each forwarding node on the end-to-end path in parallel. Each node pre-locks candidate time slots in its local inactive candidate calendar table and returns a confirmation message. After all nodes confirm, a unified commit message is sent. Each node synchronously switches its candidate calendar table to the effective calendar table at the unified effective time, realizing parallel negotiation and synchronous switching of end-to-end hard pipe bandwidth. Compared with the hop-by-hop serial negotiation method in related technologies, this application significantly shortens the negotiation latency of multi-hop link adjustment, and each node synchronously switches the calendar table based on a unified clock, effectively avoiding instantaneous packet loss, time slot misalignment, or service interruption caused by inconsistent effective times of different nodes, ensuring the consistency of end-to-end hard pipe configuration.

[0021] In one optional implementation, adjusting the hard pipe bandwidth allocation corresponding to the target service further includes: if no confirmation message is received from all forwarding nodes on the end-to-end path, sending a stop message to the forwarding nodes that have returned confirmation messages, so as to instruct the forwarding nodes to release the locked candidate time slots in the inactive candidate calendar table and keep the current effective calendar table unchanged.

[0022] Beneficial effects: The bandwidth management method provided in this application ensures the atomicity of bandwidth adjustment by sending a stop message to the confirmed node when no confirmation message is received from all forwarding nodes on the end-to-end path. This instructs the node to release the locked candidate time slots in the inactive candidate calendar table and keep the current effective calendar table unchanged. Either all nodes successfully reserve the slots and the adjustment takes effect uniformly, or all nodes abandon the reservation and maintain the original configuration. This avoids the inconsistency of the end-to-end hard pipe caused by some nodes successfully reserving slots and some nodes failing to reserve slots, and improves the reliability of the bandwidth adjustment process.

[0023] In one optional implementation, adjusting the hard pipe bandwidth allocation corresponding to the target service further includes: after sending the commit message, if there is a failed handover forwarding node on the end-to-end path, sending a rollback message to each forwarding node on the end-to-end path according to the version number to instruct each forwarding node to restore to the calendar table configuration before the handover.

[0024] Beneficial Effects: The bandwidth management method provided in this application, after sending the commit message, if there are forwarding nodes that have failed to switch over on the end-to-end path, sends a rollback message to each forwarding node according to the version number to instruct each node to restore to the calendar table configuration before the switchover. This ensures that even in extreme cases where some nodes fail to switch over during the unified activation phase, the system can still achieve precise rollback based on the adjusted version number, consistently restoring all nodes on the end-to-end path to their pre-adjustment state. This avoids the hard pipe configuration splitting and service damage caused by some nodes successfully adjusting while others fail, further improving the fault tolerance and recoverability of bandwidth adjustments. This rollback mechanism works in conjunction with the abort mechanism, covering two types of failure scenarios: the reservation phase and the activation phase, respectively, together forming a complete failure handling system.

[0025] In an optional implementation, adjusting the hard pipe bandwidth allocation corresponding to the target service further includes: after increasing the hard pipe bandwidth corresponding to the target service, if the real-time isolation does not drop below the sum of the scaling-down threshold and the preset backlash, then the step of reducing the hard pipe bandwidth corresponding to the target service is prohibited.

[0026] Beneficial effects: The bandwidth management method provided in this application prohibits the reduction of bandwidth after increasing the hard pipe bandwidth corresponding to the target service, when the real-time isolation degree has not yet dropped below the sum of the reduction threshold and the preset hysteresis. Based on the threshold hysteresis interval, a clear time dimension protection is further introduced - even if the isolation degree is lower than the expansion threshold after expansion but has not fully fallen back to near the reduction threshold, reduction is still prohibited. This effectively prevents the ping-pong effect caused by the frequent oscillation of isolation degree between expansion and reduction thresholds due to the instantaneous fluctuation of service traffic, and ensures the long-term stability of bandwidth adjustment.

[0027] Secondly, this application provides a bandwidth management system, comprising: an application layer device for dividing services in the network into multiple service modes according to the service level agreement requirements of each service carried in the network; a control layer device, communicatively connected to the application layer device, for determining the service mode to which a target service belongs from the multiple service modes as the target mode, and designating at least one service mode other than the target mode from the multiple service modes as an interference mode; and a forwarding layer device, communicatively connected to the control layer device, including a performance data acquisition unit and a flexible Ethernet hard isolation execution unit; the performance data acquisition unit is used to acquire actual performance data of the target mode under the condition of coexistence of the interference modes; the control layer device is further used to acquire baseline performance data of the target mode under the condition of no interference mode disturbance, determine the real-time isolation degree of the interference mode to the target mode based on the deviation between the actual performance data and the baseline performance data, and send a bandwidth adjustment command to the flexible Ethernet hard isolation execution unit when the real-time isolation degree meets a preset isolation threshold condition; the flexible Ethernet hard isolation execution unit is used to adjust the hard pipe bandwidth allocation corresponding to the target service according to the bandwidth adjustment command.

[0028] Thirdly, this application provides an electronic device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the bandwidth management method of the first aspect or any corresponding embodiment described above.

[0029] Fourthly, this application provides a computer-readable storage medium storing computer instructions for causing a computer to perform the bandwidth management method described in the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0031] Figure 1 This is a schematic flowchart of a first method for bandwidth management according to an embodiment of this application; Figure 2 This is a second flowchart illustrating the bandwidth management method according to an embodiment of this application; Figure 3 This is a schematic diagram of a third process of the bandwidth management method according to an embodiment of this application; Figure 4 This is a schematic diagram of the fourth process of the bandwidth management method according to the embodiments of this application; Figure 5 This is a structural block diagram of a bandwidth management system according to an embodiment of this application; Figure 6 This is a diagram illustrating the overall system deployment architecture of the bandwidth management system according to an embodiment of this application. Figure 7 This is a diagram illustrating the specific process of determining the baseline value, isolation coefficient, and tolerance threshold according to an embodiment of this application; Figure 8 This is a flowchart of hard pipe bandwidth adjustment according to an embodiment of this application; Figure 9 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, 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, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0033] It is understood that before using the technical solutions disclosed in the various embodiments of this application, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this application in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0034] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0035] As an optional application scenario of this application, the bandwidth management method provided by this application can be applied to multi-service co-network carrying scenarios such as power communication networks, data center interconnection networks or operator metropolitan area networks. The following uses power communication networks as an example for explanation, but the application scenarios of this application are not limited to this.

[0036] This application scenario includes application-layer devices deployed in local server rooms or the cloud, control-layer devices centrally deployed in SDN controllers or network management systems, and forwarding-layer devices deployed at various network nodes. The application-layer devices run a service modality management module and an operation and maintenance management interface. Network operation and maintenance personnel input the service level agreement requirements for each service through this management interface and send them to the control-layer devices. The control-layer devices and forwarding-layer devices communicate via a network (e.g., wired or wireless network). The control-layer devices send collection policies and bandwidth adjustment commands to the forwarding-layer devices via the NETCONF / YANG protocol, while the forwarding-layer devices report performance data to the control-layer devices via gRPC Telemetry or REST interfaces.

[0037] The forwarding layer equipment includes multiple forwarding devices deployed on network nodes (e.g., substation exit PE routers, core equipment room PE routers, core switches, FlexE interface cards, etc.). These forwarding devices are connected by physical optical fibers to form an end-to-end service transmission path. Each forwarding device's FlexE interface card divides the physical interface (e.g., a 100GE optical port) into multiple time slots (e.g., 5G per time slot). Different service modes achieve hard isolation on the same physical link by exclusively occupying different numbers of time slots.

[0038] The infrastructure layer is equipped with a time synchronization server (such as a PTP IEEE 1588v2 time server or an NTP time server). The time synchronization server is connected to the control layer devices and all forwarding layer devices through a wired network, providing a unified time reference for the control layer devices and all forwarding layer devices, ensuring that the unified effective time issued by the control layer can be accurately executed on all forwarding layer devices.

[0039] For example, relay protection services (Mode A), video inspection services (Mode B), and remote control services (Mode C) share the same physical fiber optic cable and transmission equipment. The three services are carried on different sets of time slots via flexible Ethernet hard pipes; for instance, relay protection services occupy 10 time slots (50G bandwidth), video inspection services occupy 6 time slots (30G bandwidth), and remote control services occupy 4 time slots (20G bandwidth).

[0040] In this application scenario, the control layer device, based on the service level agreement requirements of each service issued by the application layer, divides the network services into the three service modes mentioned above, and determines the service mode to which the target service (e.g., relay protection service) belongs as the target mode, while the other two service modes are considered interference modes. The control layer device acquires the actual performance data of the target mode on various performance indicators in real time through the performance data acquisition unit of the forwarding layer device, calculates the real-time isolation degree by combining it with pre-acquired baseline performance data, and generates a bandwidth adjustment command when the real-time isolation degree meets the preset isolation threshold condition. The flexible Ethernet hard isolation execution unit of the forwarding layer device, based on this command, pre-locks candidate time slots in the local inactive candidate calendar table, and completes the calendar table switch at a unified effective time after all nodes on the end-to-end path are ready, thereby achieving dynamic expansion and contraction of hard-pipe bandwidth without interrupting services.

[0041] The exemplary descriptions of the above application scenarios are for the purpose of facilitating understanding of this application and do not constitute any limitation on the scope of protection of this application.

[0042] In related technologies, bandwidth resource management in multi-service shared network scenarios mainly relies on manually setting performance indicator weights and statically configuring hard-pipe bandwidth. This makes it difficult to objectively quantify the actual degree of mutual interference between services and to adaptively adjust bandwidth allocation based on service operation status. Excessive reserved bandwidth can lead to resource idleness, while insufficient reserved bandwidth can cause critical services to be affected by sudden traffic surges from other services, resulting in degraded latency, jitter, or packet loss rates. Furthermore, existing multi-hop bandwidth adjustment processes use a hop-by-hop negotiation approach, which is prone to problems such as adjustment failures at some nodes or inconsistent effective times among nodes. This application provides a bandwidth management method, system, electronic device, and storage medium for multi-service modal isolation scenarios. By constructing an automatic service modal isolation coefficient generation mechanism and an end-to-end hard-pipe synchronous adjustment mechanism based on a shadow calendar, it can objectively quantify the actual degree of mutual interference between multiple services and adaptively adjust bandwidth resource allocation based on the quantification results, while ensuring the consistency and rollbackability of multi-hop node adjustments.

[0043] According to an embodiment of this application, a bandwidth management method embodiment for a multi-service modal isolation scenario is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0044] This embodiment provides a bandwidth management method for multi-service modal isolation scenarios, which can be used in control layer devices such as SDN controllers or network management systems in the above application scenarios. Figure 1 This is a flowchart of a bandwidth management method according to an embodiment of this application, referred to... Figure 1As shown, the process includes the following steps: Step S101: Based on the service level agreement requirements of each service carried in the network, the services in the network are divided into multiple service modes. The service mode to which the target service belongs is determined from the multiple service modes and is taken as the target mode. At least one service mode other than the target mode is taken as the interference mode.

[0045] In the above steps, "services" refers to the various application services carried in the network, such as relay protection services, video inspection services, and remote control services in power communication networks. Service Level Agreement (SLA) requirements refer to the specific network performance requirements of each service; for example, relay protection services require latency of less than 5 milliseconds, jitter of less than 50 microseconds, and packet loss rate of less than 10%. -6 Video inspection services require a bandwidth greater than 20 megabits per second and jitter less than 5 milliseconds, while remote control services require a packet loss rate of less than 10%. -3 The latency is less than 50 milliseconds. A service modality refers to a collection of services with the same or similar service level agreement requirements. Services within the same service modality are considered to have the same performance guarantee requirements in network resource management.

[0046] Segmentation refers to the process of classifying and aggregating services based on the similarity of their service level agreement requirements. Determination refers to the process of selecting one of the multiple segmented service modalities as the current evaluation object.

[0047] For example, in a power communication network, the control layer equipment, based on the service level agreement requirements of each service issued by the application layer equipment, classifies the relay protection service, video inspection service, and remote control service carried in the network into Mode A, Mode B, and Mode C, respectively. If it is necessary to assess the interference status of the relay protection service, the control layer equipment determines Mode A as the target mode and Modes B and C as interference modes. It is understandable that the target mode can be flexibly switched according to the assessment needs—when it is necessary to assess the video inspection service, Mode B becomes the target mode, and Modes A and C become interference modes.

[0048] Step S102: Obtain the actual performance data of the target mode under the condition of coexistence of disturbance modes, and the benchmark performance data of the target mode under the condition of no disturbance mode.

[0049] In the above steps, performance data refers to quantifiable indicators reflecting the quality of service transmission, including bandwidth, latency, jitter, and packet loss rate. Actual performance data refers to the measured performance index values ​​of the service corresponding to the target mode during transmission in a real-world operating environment with multiple services coexisting. Baseline performance data refers to the performance index values ​​corresponding to the target mode under stable operating conditions without interference—this value reflects the performance level of the target mode in an ideal environment and serves as a reference benchmark for subsequent judgments regarding whether interference has occurred.

[0050] Acquisition refers to the process of collecting performance data through performance data acquisition units deployed on forwarding layer devices. Acquisition methods may include acquiring end-to-end latency, jitter, and packet loss rate based on bidirectional active measurement protocols, and extracting time slot occupancy from flexible Ethernet overhead based on the operation, management, and maintenance overhead mechanisms of flexible Ethernet protocols to determine bandwidth occupancy data.

[0051] For example, in the evaluation scenario of relay protection service (Mode A), the control layer device first continuously collects performance data for 5 minutes under the condition that only Mode A is running in the network and there is no additional disturbance traffic, and calculates the average value of each index to obtain the baseline performance data of Mode A: bandwidth of 50G, latency of 2ms, jitter of 20μs, and packet loss rate of 10. -7 Subsequently, under the actual operating conditions of simultaneously running Mode B (video inspection service) and Mode C (remote control service) in the network, the control layer device continuously collected real-time performance data of Mode A with a sampling period of 10ms, obtaining the following actual performance data: bandwidth of 50G, latency of 4.5ms, jitter of 48μs, and packet loss rate of 5×10⁻⁶. -7 It is understandable that benchmark performance data can be obtained during the initial testing phase of network deployment, or it can be dynamically updated during periods of light workload and minimal interference during network operation.

[0052] Step S103: Determine the real-time isolation degree of the interference mode from the target mode based on the deviation between the actual performance data and the benchmark performance data.

[0053] In the above steps, deviation refers to the change in actual performance data relative to baseline performance data, reflecting the degree of influence of the interference mode on the performance indicators of the target mode. Isolation is a comprehensive quantitative indicator used to characterize the overall interference degree of the interference mode on the target mode; a higher isolation value indicates more severe interference.

[0054] Determination refers to the process of weighting and fusing the deviation values ​​of each performance indicator according to their respective isolation coefficients to generate a comprehensive isolation index.

[0055] For example, the control layer device calculates the deviations of the real-time performance data of mode A from the baseline performance data in various performance indicators: bandwidth deviation is 0 (50G-50G=0), latency deviation is 2.5ms (4.5ms-2ms=2.5ms), jitter deviation is 28μs (48μs-20μs=28μs), and packet loss rate deviation is 4×10 -7 (5×10) -7 -10 -7 =4×10 -7 The control layer device performs weighted fusion of each deviation value based on the isolation coefficient corresponding to each performance indicator—for example, in this network scenario, the isolation coefficients for latency, jitter, packet loss rate, and bandwidth are pre-determined to be 0.5, 0.2, 0.2, and 0.1, respectively—and takes the maximum value of the fusion result as the real-time isolation degree. Since the weighted deviation value for latency is the largest, this maximum value is the real-time isolation degree experienced by mode A at the current moment.

[0056] Step S104: If the real-time isolation meets the preset isolation threshold, adjust the hard pipe bandwidth allocation corresponding to the target service.

[0057] In the above steps, the preset isolation threshold condition is a pre-set trigger condition used to determine whether bandwidth adjustment needs to be performed. A hard pipe refers to a physically isolated data transmission channel provided by flexible Ethernet technology, where different services achieve non-interference hard isolation by exclusively occupying different sets of time slots. Bandwidth allocation refers to the number of time slot resources occupied by the target service within the hard pipe.

[0058] Adjustment refers to the process of changing the number of time slots occupied by the target service—increasing time slots (expanding capacity) when bandwidth is insufficient, and decreasing time slots (shrinking capacity) when bandwidth is redundant.

[0059] For example, the preset isolation threshold conditions include an expansion threshold and a reduction threshold, both determined based on historical samples. Assume the expansion threshold is 0.6 and the reduction threshold is 0.3. If the real-time isolation calculated in step S103 is 0.65, which is greater than the expansion threshold of 0.6, the control layer device determines that the interference experienced by the current mode A has exceeded the tolerance limit and triggers the bandwidth adjustment mechanism. The control layer device generates a bandwidth adjustment command, instructing the forwarding layer device to increase the hard-pipe bandwidth of mode A from the current 10 time slots to 12 time slots (expansion). After expansion, the real-time isolation of mode A drops to 0.25, which is lower than the reduction threshold of 0.3. The control layer device can then further trigger reduction adjustment to restore the bandwidth to a reasonable level and avoid resource waste. If the real-time isolation is 0.45, which is between the expansion threshold of 0.6 and the reduction threshold of 0.3, the control layer device maintains the current bandwidth configuration unchanged.

[0060] The bandwidth management method provided in this embodiment divides service modes according to service level agreement requirements and determines the target mode and interfering modes. It obtains the actual performance data of the target mode under conditions where interfering modes coexist, and its baseline performance data under interference-free conditions. The deviation between the actual performance data and the baseline performance data objectively quantifies the actual interference degree of the interfering mode on the target mode—a larger deviation indicates more severe interference. Then, it decides whether to adjust the hard-pipe bandwidth allocation based on whether the quantified real-time isolation degree meets a preset threshold condition. This method achieves objective quantification of the degree of mutual interference between multiple services, avoiding the subjectivity of relying on manual experience to set indicator weights in related technologies. It enables bandwidth adjustment decisions to be dynamically triggered based on the actual operating status of the services—timely expansion to ensure the service quality of critical services when isolation exceeds limits, and proactive reduction to release idle resources when isolation is low. This improves the overall utilization rate of hard-pipe resources while ensuring the service quality of critical services.

[0061] This embodiment provides a bandwidth management method that can be used in control layer devices in the above application scenarios, such as SDN controllers or network management systems. Figure 2 This is a flowchart of a bandwidth management method according to an embodiment of this application, referred to... Figure 2 As shown, the process includes the following steps: Step S201: Based on the service level agreement requirements of each service carried in the network, the services in the network are divided into multiple service modes. The service mode to which the target service belongs is determined from these multiple service modes and designated as the target mode. At least one service mode other than the target mode is designated as an interference mode. For details, please refer to [link to relevant documentation]. Figure 1 Step S101 of the illustrated embodiment will not be described again here.

[0062] Step S202: Obtain the actual performance data of the target mode under the condition of coexistence of interfering modes.

[0063] Specifically, step S202 includes: Step S2021: Determine the length of the preset sampling period according to the priority of the target mode, wherein the higher the priority, the shorter the corresponding sampling period.

[0064] In the above steps, service priority refers to the priority level determined based on the importance of the service and the requirements of the service level agreement. In power communication networks, relay protection services have extremely strict requirements for latency and reliability, and their priority is higher than that of video inspection services and remote control services. Therefore, they need to be monitored with a shorter sampling period in order to detect transient interferences in a timely manner and trigger adjustments.

[0065] For example, Mode A (relay protection service) has the highest priority, with a corresponding sampling period of 10ms; Mode B (video inspection service) has a medium priority, with a corresponding sampling period of 100ms; and Mode C (remote control service) has a relatively low priority, with a corresponding sampling period of 1s. The length of the sampling period determines the system's sensitivity to changes in service performance—the shorter the sampling period, the earlier the system can detect interference and trigger adjustments, thereby better ensuring the service quality of high-priority services.

[0066] Step S2022: Collect performance data of the target mode at a preset sampling period, including at least one of the following methods: obtain end-to-end latency, jitter and packet loss rate of the target mode based on a bidirectional active measurement protocol; extract time slot occupancy rate from the flexible Ethernet overhead based on the operation, management and maintenance overhead mechanism of the flexible Ethernet protocol, and determine bandwidth occupancy data based on the time slot occupancy rate; and determine the actual service utilization rate of the target mode based on port traffic statistics or service traffic statistics.

[0067] In the above steps, the performance data acquisition unit includes two types: active measurement probes based on bidirectional active measurement protocols and physical layer indicator extraction units based on flexible Ethernet protocol operation management and maintenance overhead mechanisms.

[0068] Specifically, in step S2022 above, obtaining the end-to-end latency, jitter, and packet loss rate of the target mode based on the bidirectional active measurement protocol includes: an active measurement probe deployed at the source PE node sends a measurement message to the destination PE node; the destination PE node reflects the message; and the source PE node calculates the round-trip latency based on the message's transmission and reception times, thereby obtaining the end-to-end latency, jitter, and packet loss rate. This method does not require the participation of intermediate nodes and can directly reflect the actual end-to-end transmission quality.

[0069] Based on the operation, management, and maintenance overhead mechanism of the Flexible Ethernet protocol, time slot occupancy is extracted from the Flexible Ethernet overhead, and bandwidth occupancy data is determined according to the time slot occupancy rate. This includes: the Flexible Ethernet protocol defines operation, management, and maintenance overhead channels to carry operation and management information of the Flexible Ethernet group. The forwarding layer device reads the current time slot occupancy status from the Flexible Ethernet overhead, counts the number of time slots occupied by the target service and its proportion of the total time slots, and calculates the bandwidth occupancy data by combining this with the total bandwidth of the physical interface. This method does not require additional probe traffic injection, avoiding interference from measurement traffic to the service itself. Furthermore, the actual service utilization rate is determined based on port traffic statistics or service traffic statistics, including: the forwarding layer device uses built-in counters to count the actual traffic of each port or the traffic of a specific service flow, and calculates the actual service utilization rate by combining this with the total bandwidth of the physical interface.

[0070] Step S203: Based on the deviation between the actual performance data and the baseline performance data, determine the real-time isolation degree of the interference mode from the target mode. For details, please refer to [link to relevant documentation]. Figure 1 Step S103 of the illustrated embodiment will not be described again here.

[0071] Step S204: If the real-time isolation meets the preset isolation threshold, adjust the hard pipe bandwidth allocation corresponding to the target service. For details, please refer to [link to relevant documentation]. Figure 1 Step S104 of the illustrated embodiment will not be described again here.

[0072] The bandwidth management method provided in this embodiment determines the sampling period based on the priority of the target modality—the higher the priority, the shorter the sampling period—achieving more timely monitoring and response to high-priority critical services. It also provides two acquisition methods: a bidirectional active measurement protocol and a flexible Ethernet protocol operation, management, and maintenance overhead mechanism. The former can directly measure end-to-end latency, jitter, and packet loss rate, while the latter can acquire bandwidth usage data without injecting additional probe traffic. These two methods complement each other and can be flexibly selected or combined according to the actual network environment and service requirements, ensuring both the comprehensiveness and flexibility of performance data acquisition while avoiding additional interference from measurement traffic to the services themselves.

[0073] This embodiment provides a bandwidth management method that can be used in control layer devices in the above application scenarios, such as SDN controllers or network management systems. Figure 3 This is a flowchart of a bandwidth management method according to an embodiment of this application, referred to... Figure 3 As shown, the process includes the following steps: Step S301: Based on the service level agreement requirements of each service carried in the network, the services in the network are divided into multiple service modes. The service mode to which the target service belongs is determined from these multiple service modes and designated as the target mode. At least one service mode other than the target mode is designated as an interference mode. For details, please refer to [link to relevant documentation]. Figure 1 Step S101 of the illustrated embodiment will not be described again here.

[0074] Step S302: Obtain the actual performance data of the target mode under the condition of coexistence of disturbing modes, and the benchmark performance data of the target mode under the condition of no disturbing mode perturbation. For details, please refer to [link to relevant documentation]. Figure 1 Step S102 of the illustrated embodiment will not be described again here.

[0075] Step S303: Determine the real-time isolation degree of the interference mode from the target mode based on the deviation between the actual performance data and the benchmark performance data.

[0076] Specifically, step S303 includes: Step S3031: Based on the actual performance data and the benchmark performance data, determine the deviation values ​​corresponding to each performance index of the target mode. The performance indexes include at least two of bandwidth, latency, jitter and packet loss rate.

[0077] Performance data of the target mode is collected under conditions of no additional disturbance traffic. After the network enters a stable state, performance data is continuously collected for a period of time (e.g., 5 minutes), and the average value of each index is calculated to obtain the baseline performance vector of the target mode.

[0078] in The reference bandwidth representing the service mode (m) Indicates the reference delay. Indicates reference jitter, This represents the baseline packet loss rate. The above baseline values ​​reflect the ideal performance level without additional interference. The system injects disturbance traffic of other service modes into the network according to a disturbance strategy. This strategy includes changing one or more of the following: the traffic rate, burst period, packet length, transmission interval, and service priority of the disturbance service. During the disturbance period, the system periodically collects real-time performance data of the target service mode at intervals Tr to obtain a real-time performance vector.

[0079] Calculate the normalized deviation vector based on the real-time performance vector and the baseline performance vector:

[0080] Deviation refers to the change in actual performance data relative to baseline performance data. Performance metrics include at least two of the following: bandwidth, latency, jitter, and packet loss rate. The specific performance metrics selected depend on the sensitivity of the target service. For example, relay protection services are sensitive to latency, jitter, and packet loss rate, so all four metrics can be selected; video inspection services are sensitive to bandwidth and jitter, so only bandwidth and jitter metrics can be selected.

[0081] Step S3032: Determine the isolation coefficient corresponding to each performance index, perform weighted fusion on each deviation value based on the isolation coefficient, and take the maximum value in the fusion result as the real-time isolation degree.

[0082] The isolation coefficient is a pre-determined weight value trained using historical samples, used to reflect the relative contribution of each performance indicator to the degradation of the target business service quality. Weighted fusion refers to the process of normalizing the deviation values ​​of each performance indicator without dimension and then weighting them according to their corresponding isolation coefficients. The maximum value in the fusion result is taken as the real-time isolation degree, corresponding to the Chebyshev norm, so that the real-time isolation degree can reflect the most severe degradation of any performance dimension.

[0083] For example, the deviation values ​​of the four performance indicators of mode A are bandwidth 0, latency 2.5ms, jitter 28μs, and packet loss rate 4×10⁻⁶. -7 The corresponding isolation coefficients are 0.1, 0.5, 0.2, and 0.2, respectively. Among the weighted deviation values, the weighted deviation value of time delay is the largest (2.5 × 0.5 = 1.25), therefore the real-time isolation degree of mode A at the current moment is 1.25.

[0084] Specifically, in step S3032 above, determining the isolation coefficient corresponding to each performance index includes: Step a1: Obtain multiple historical samples of the target modality. Each historical sample includes the historical deviation vector of the target modality at a historical moment and the corresponding business degradation label.

[0085] The SLA constraint vector for the service mode (m) is generated based on the SLA constraint values ​​of the target mode as follows:

[0086] Based on the degree of change in real-time performance relative to the baseline value, and combined with the prior importance coefficient αm of the business SLA, the business degradation label is calculated:

[0087] in, This represents the prior importance coefficient for the business SLA.

[0088] Historical samples refer to data records collected during historical operation and used to train the isolation coefficient generation model. Each historical sample contains a historical deviation vector (a vector consisting of the deviations of the actual values ​​of each performance indicator from the benchmark value) and a corresponding business degradation label (used to mark the degree of business degradation at that moment).

[0089] Step a2: Determine the sample weight corresponding to each historical sample based on the service level agreement constraint values ​​of each performance indicator; among them, historical samples that are close to the service level agreement critical state are assigned higher sample weights.

[0090] A service level agreement (SLA) critical state refers to a state where the degree of service degradation is close to but has not yet exceeded the allowable upper limit of the SLA. Samples close to the critical state are given higher weights, meaning that the model pays more attention to the critical state where the service is close to the default boundary during training.

[0091] To make the generated isolation coefficients more relevant to the critical degradation state of the business, the sample weights are expressed as follows:

[0092] in, The critical label value for business degradation (calculated based on the SLA constraint vector). For boundary sample enhancement coefficients, This is the attenuation parameter. This forms the SLA boundary-weighted training sample set:

[0093] Step a3: Using the historical deviation vector as input and the business degradation label as the training target, train a sparse regression model with non-negative constraints. The sparse regression model has contribution coefficients, sparse regularization terms, and smoothing regularization terms. The contribution coefficients are used to characterize the degree of contribution of each performance indicator to the business degradation label. The sparse regularization terms are used to highlight the performance indicators that have the greatest impact on the business degradation label. The smoothing regularization terms are used to suppress the oscillation of the contribution coefficients.

[0094] Based on the above training sample set, train a sparse regression model with non-negativity constraints:

[0095] in, The performance contribution coefficient to be solved. and These are the sparse regularization coefficient and the smooth regularization coefficient, respectively.

[0096] Non-negativity constraints restrict all contribution coefficients to be greater than or equal to zero, ensuring that the degradation of each performance indicator contributes in the same direction to the service degradation label. For example, after the model is trained, the output contribution coefficients are bandwidth 0.1, latency 0.5, jitter 0.2, and packet loss rate 0.2—latency has the largest contribution coefficient, indicating that latency deviation contributes the most to the degradation of relay protection services in the current network scenario.

[0097] Step a4: Normalize the contribution coefficients to obtain the isolation coefficients corresponding to each performance index.

[0098] The contribution coefficient is obtained by solving. Then, it is normalized to obtain the isolation coefficient vector corresponding to the target business mode:

[0099] in:

[0100] To prevent extremely small positive numbers with a denominator of zero.

[0101] For example, after normalization, the above four contribution coefficients of 0.1, 0.5, 0.2 and 0.2 are used to obtain isolation coefficients: latency 0.5, jitter 0.2, packet loss rate 0.2 and bandwidth 0.1.

[0102] In some optional implementations, when there is a significant nonlinear relationship between the performance degradation and performance deviation of the target service mode, the system can train a nonlinear prediction model fm(x) and use Shapley values ​​to calculate the average contribution of each performance index:

[0103] Then, normalize the average contribution to obtain the isolation coefficient:

[0104] The resulting isolation coefficient is not preset manually, but is generated by a combination of service disturbance samples, SLA constraint boundary samples, and performance contribution attribution. It can reflect the true sensitivity of different service modes to bandwidth, latency, jitter, and packet loss rate in different network environments.

[0105] After determining the isolation coefficient vector of the target service mode, critical SLA samples (degradation just reaching the allowable upper limit) and high-efficiency samples (excellent degradation) are selected from the dataset. Their perturbation levels are then calculated using the interference degree calculation formula and used as the tolerance upper limit. and tolerance limit The initial value.

[0106] In some optional implementations, when there is a significant non-linear relationship between service performance degradation and performance deviation, the determination of the isolation coefficients corresponding to each performance index in step S3032 above may include: Step b1 involves obtaining multiple historical samples of the target modality. Each historical sample includes the historical deviation vector of the target modality at a historical moment and the corresponding service degradation label. For the specific implementation of this step, please refer to the relevant description in step a1 above, which will not be repeated here.

[0107] Step b2: Train a nonlinear prediction model using the historical deviation vector as input and the business degradation label as the training target.

[0108] Nonlinear prediction models are machine learning models that can capture the nonlinear mapping relationship between input features and output targets, such as neural network models, random forest models, or support vector regression models. When there is a significant nonlinear relationship between business performance degradation and performance deviation—for example, the rate of degradation increases sharply after certain performance indicators reach a certain threshold—linear regression models struggle to accurately characterize this complex relationship, while nonlinear prediction models can effectively capture the complex mapping relationship between deviations in various performance indicators and business degradation.

[0109] Step b3: Calculate the average contribution of each performance indicator to the business degradation label using the feature contribution method.

[0110] Feature contribution methods are techniques used to interpret the decision-making processes of black-box models, quantifying the contribution of each input feature to the model's prediction results. Commonly used feature contribution methods include those based on cooperative game theory (such as the Shapley value method), which determines the importance of a feature by calculating the average marginal contribution of each feature across all possible feature combinations, thus fairly distributing prediction results among the input features.

[0111] Step b4 involves normalizing the average contribution values ​​to obtain the isolation coefficients corresponding to each performance index. The specific implementation of this step is described in step a4 above and will not be repeated here.

[0112] In some optional implementations, obtaining multiple historical samples of the target mode in step a1 or step b1 above includes: Step c1: Determine the prior importance coefficients corresponding to each performance indicator based on the service level agreement constraint values ​​of the target business.

[0113] The prior importance coefficient is a pre-determined importance weight based on the service level agreement (SLA) constraints of the service—the stricter the SLA constraints, the higher the corresponding prior importance coefficient. For example, relay protection services have extremely strict latency requirements (less than 5ms), so their latency prior importance coefficient should be set relatively high.

[0114] Step c2: Calculate the degree of degradation for each historical sample based on the historical deviation vector and the corresponding prior importance coefficient, and use it as a business degradation label.

[0115] For example, the prior importance coefficients of the performance indicators for mode A are latency 0.5, jitter 0.2, packet loss rate 0.2, and bandwidth 0.1. For a certain historical sample, its historical deviation vector is latency deviation 2.5ms, jitter deviation 28μs, and packet loss rate deviation 4×10. -7 The bandwidth deviation is 0. Since the dimensions of each performance indicator are different (latency is in milliseconds, jitter is in microseconds, and packet loss rate is a dimensionless ratio), they need to be normalized first. Taking latency as an example, if the latency baseline is 2ms, the normalized latency deviation is 2.5 / 2 = 1.25; the jitter baseline is 20μs, the normalized jitter deviation is 28 / 20 = 1.4; and the packet loss rate baseline is 10... -7 The normalized packet loss rate bias is 4×10 -7 / 10 -7=4; the bandwidth baseline is 50G, and the normalized bandwidth deviation is 0 / 50=0. Therefore, the normalized deviation vector is [0, 1.25, 1.4, 4]. The control layer device calculates the degradation degree based on this normalized deviation vector and the corresponding prior importance coefficient: 0×0.1+1.25×0.5+1.4×0.2+4×0.2=0+0.625+0.28+0.8=1.705, which is the corresponding service degradation label. The above normalization process eliminates the influence of different dimensions and numerical scales on various performance indicators, enabling the deviations of different performance indicators to be weighted and fused on the same scale.

[0116] Step S304: If the real-time isolation meets the preset isolation threshold, adjust the hard pipe bandwidth allocation corresponding to the target service. For details, please refer to [link to relevant documentation]. Figure 1 Step S104 of the illustrated embodiment will not be described again here.

[0117] The bandwidth management method provided in this embodiment achieves unified quantification of multi-dimensional performance indicators into a comprehensive isolation degree by weighted fusion of deviation values ​​and isolation coefficients to obtain the maximum value, enabling the isolation degree to reflect the most severe degradation of any performance dimension. The isolation coefficient is automatically generated by training a sparse regression model or nonlinear prediction model using historical samples, avoiding the experience and subjectivity of manually setting weights. Non-negativity constraints ensure that the contribution direction of each performance indicator degradation to business degradation is consistent. Sparse regularization and smoothing regularization respectively achieve automatic screening of key performance indicators and stability assurance of contribution coefficients. The prior importance coefficient incorporates the differentiated sensitivity information of different services to different performance indicators into the degradation label, providing a high-quality training target for supervised learning of the isolation coefficient.

[0118] This embodiment provides a bandwidth management method that can be used in control layer devices in the above application scenarios, such as SDN controllers or network management systems. Figure 4 This is a flowchart of a bandwidth management method according to an embodiment of this application, referred to... Figure 4 As shown, the process includes the following steps: Step S401: Based on the service level agreement requirements of each service carried in the network, the services in the network are divided into multiple service modes. The service mode to which the target service belongs is determined from these multiple service modes and designated as the target mode. At least one service mode other than the target mode is designated as an interference mode. For details, please refer to [link to relevant documentation]. Figure 1 Step S101 of the illustrated embodiment will not be described again here.

[0119] Step S402: Obtain the actual performance data of the target mode under the condition of coexistence of disturbing modes, and the benchmark performance data of the target mode under the condition of no disturbing mode perturbation. For details, please refer to [link to relevant documentation].Figure 1 Step S102 of the illustrated embodiment will not be described again here.

[0120] Step S403: Based on the deviation between the actual performance data and the baseline performance data, determine the real-time isolation degree of the interference mode from the target mode. For details, please refer to [link to relevant documentation]. Figure 1 Step S103 of the illustrated embodiment will not be described again here.

[0121] Step S404: If the real-time isolation meets the preset isolation threshold, adjust the hard pipe bandwidth allocation corresponding to the target service.

[0122] The system periodically collects network performance data at a preset sampling period (e.g., 10ms). The sampling period can be dynamically adjusted according to service priority, with higher priority services corresponding to shorter sampling periods to achieve more timely monitoring and response. The sampling strategy is uniformly distributed by the control layer to various sampling probes deployed at the source PE nodes, destination PE nodes, core nodes, and FlexE interfaces.

[0123] The performance data acquisition employs a composite acquisition mechanism: on the one hand, it utilizes the TWAMP Light protocol defined by the Internet Engineering Task Force (IETF) RFC5357 standard to achieve end-to-end latency. Shaking and packet loss rate Active measurement achieves timestamp accuracy down to the nanosecond level. Furthermore, utilizing the Operation, Management and Maintenance (OAM) overhead mechanism built into the FlexE protocol, physical layer status indicators such as slot occupancy rate and bit error rate are directly extracted from the FlexE Overhead. Bandwidth occupancy can be determined based on the slot occupancy rate. This method eliminates the need for additional probe traffic, thus avoiding interference from measurement traffic on the business itself.

[0124] The collected raw data is aggregated and then sent to the inter-modal influence detection module of the control layer. After preprocessing and time-aligning the raw performance data, the inter-modal influence detection module sends the preprocessed data to the interference level quantification module, which calculates the real-time interference level according to the p-norm weighted model formula. :

[0125] in, It is the obtained isolation coefficient vector. It is a normalized deviation vector obtained based on the collected data and the benchmark value, with the sign... Representing vectors p - Norm.

[0126] The preset isolation threshold conditions include an expansion threshold determined based on service level agreement (SLA) critical samples whose degradation level reaches the allowable upper limit from the aforementioned historical samples, and a reduction threshold determined based on samples from the aforementioned historical samples whose degradation level meets preset excellent conditions. Here, SLA critical samples refer to historical samples whose degradation level exactly reaches the allowable upper limit of the SLA; preset excellent conditions refer to degradation levels below a certain percentage of the allowable upper limit (e.g., below 30% of the allowable upper limit), indicating excellent service operation and resource redundancy. The control layer device selects samples that meet the above conditions from the historical samples, calculates their isolation levels, and uses these isolation levels as the initial values ​​for the expansion and reduction thresholds. For example, the expansion threshold is 0.6 and the reduction threshold is 0.3.

[0127] Calculated real-time interference level With tolerance threshold and tolerance lower limit and upper limit Comparison: If: This indicates that the interference level between the current modes is low, there is redundancy in system resources, and the hard pipe bandwidth adjustment mechanism is triggered; like This indicates that the current level of interference is within a reasonable range, the network status is stable, and the bandwidth remains unchanged. like This indicates that the interference from other modes to this mode exceeds the tolerance limit, and the current resources are insufficient, triggering the hard pipe bandwidth adjustment mechanism.

[0128] Specifically, in step S404 above, when the real-time isolation meets the preset isolation threshold, adjusting the hard bandwidth allocation corresponding to the target service includes: Step S4041: If the real-time isolation is greater than the expansion threshold, increase the hard pipe bandwidth corresponding to the target service.

[0129] For example, when the real-time isolation is 0.65, which is greater than the expansion threshold of 0.6, the control layer device triggers expansion adjustment.

[0130] Step S4042: If the real-time isolation is less than the scaling-down threshold, reduce the hard pipe bandwidth corresponding to the target service.

[0131] For example, when the real-time isolation is 0.25, which is less than the scaling-down threshold of 0.3, the control layer device triggers a scaling-down adjustment.

[0132] Step S4043: If the real-time isolation is greater than or equal to the shrinkage threshold and less than or equal to the expansion threshold, keep the hard pipe bandwidth corresponding to the target service unchanged.

[0133] For example, when the real-time isolation is 0.45, which is between the shrinking threshold of 0.3 and the expanding threshold of 0.6, the control layer device maintains the current bandwidth configuration unchanged.

[0134] Specifically, increasing or decreasing the hard pipe bandwidth corresponding to the target service in step S4041 or step S4042 above includes: Step d1: Generate the version number and unified effective time for this adjustment.

[0135] The version number is a unique identifier generated by the control layer device for each bandwidth adjustment operation, used to identify the corresponding adjustment operation in subsequent abort or rollback operations. The unified effective time refers to the time when all forwarding nodes synchronously perform calendar table switching, which is determined by the control layer device based on the current time plus a preset protection time. The protection time is preset based on the end-to-end path length, node processing latency, and clock synchronization accuracy.

[0136] Step d2: Send reservation request messages in parallel to each forwarding node on the end-to-end path corresponding to the target service. The reservation request message carries the version number, the number of time slots to be adjusted, and the unified effective time.

[0137] Parallel transmission refers to the control layer device simultaneously sending reservation request messages to all forwarding nodes along the end-to-end path, rather than sending them sequentially hop by hop. The end-to-end path refers to the sequence of all forwarding nodes that the target service traverses from the source to the destination, such as the source PE node, intermediate node P1, intermediate node P2, and the destination PE node.

[0138] Step d3: Each forwarding node pre-locks a candidate time slot in its local inactive candidate calendar table and returns a confirmation message after successful locking.

[0139] The Flexible Ethernet protocol provides two sets of client calendar configurations, A and B. One set is in use, and the other can be used as a configuration to be switched over. In this application, the currently inactive calendar configuration is referred to as the Shadow Calendar, which is used to temporarily store candidate time slot configurations without affecting current service transmission. After receiving the reservation request message, each forwarding node pre-locks the candidate time slot to be adjusted in its local inactive candidate calendar table. After successful locking, it returns a confirmation message to the control layer device.

[0140] Step d4: If all forwarding nodes on the end-to-end path return acknowledgment messages, send a commit message to all forwarding nodes.

[0141] The control layer device aggregates acknowledgment messages from all forwarding nodes along the end-to-end path. Only when all forwarding nodes return acknowledgment messages does the control layer device determine that the adjustment has met the conditions for consistent effectiveness and send a commit message to all forwarding nodes.

[0142] In step d5, under the control of the submission message, each forwarding node switches the inactive candidate calendar table to the effective calendar table within a unified effective time to complete the hard pipe bandwidth allocation adjustment for the target service.

[0143] Upon receiving the commit message, each forwarding node does not immediately switch its calendar. Instead, it waits locally until the unified effective time arrives before synchronously executing the calendar switch. Because the infrastructure layer has a time synchronization server deployed to provide a unified time base for all forwarding nodes, each forwarding node can perform the switch operation at the same time.

[0144] In some optional implementations, step d4 above further includes: if no acknowledgment messages are received from all forwarding nodes on the end-to-end path, sending a stop message to the forwarding nodes that have returned acknowledgment messages, instructing the forwarding nodes to release the locked candidate time slots in the inactive candidate calendar table and keep the current active calendar table unchanged. Not receiving all acknowledgment messages includes any forwarding node returning an unacknowledgment message or the control layer device not receiving all acknowledgment messages within a preset time.

[0145] In some optional implementations, step d5 above further includes: after sending the commit message, if there are any failed switching forwarding nodes on the end-to-end path, a rollback message is sent to each forwarding node on the end-to-end path according to its version number, instructing each forwarding node to restore its calendar table configuration before the switch. A failed switching forwarding node refers to a forwarding node that failed to switch successfully when the calendar table switch was performed at the unified effective time. Possible causes include node hardware failure, software anomalies, clock synchronization issues, etc.

[0146] In some optional implementations, after step S4041, the method further includes: after increasing the hard bandwidth corresponding to the target service, if the real-time isolation does not drop below the sum of the scaling-down threshold and the preset hysteresis, then the step of reducing the hard bandwidth corresponding to the target service is prohibited. The preset hysteresis is a pre-set protection value, and the sum of the scaling-down threshold and the preset hysteresis constitutes the minimum isolation threshold that allows scaling-down to be triggered. For example, if the sum of the scaling-down threshold 0.3 and the preset hysteresis 0.1 is 0.4, and the real-time isolation drops from 0.65 to 0.45—which is lower than the scaling-up threshold 0.6 but not yet below 0.4—the control layer device prohibits scaling-down.

[0147] The bandwidth management method provided in this embodiment determines expansion and contraction thresholds by screening critical and excellent service level agreement (SLP) samples from actual historical samples, ensuring that both thresholds originate from actual business operation data rather than manually fixed values. By comparing real-time isolation with the expansion and contraction thresholds respectively, expansion, contraction, or maintaining the current value is executed based on the comparison results, achieving adaptive triggering and precise control of bandwidth adjustments. Furthermore, by generating version numbers and unified effective times, reservation request messages are sent in parallel to each forwarding node on the end-to-end path, and each node pre-locks candidate time slots in its local inactive candidate calendar. After all confirmations are made, the calendar table is switched synchronously at a unified effective time. Compared with the hop-by-hop serial negotiation method, this significantly shortens the negotiation latency of multi-hop link adjustments and effectively avoids instantaneous packet loss or service interruption caused by inconsistent effective times of different nodes. The abort mechanism releases locked time slots and maintains the original configuration when the reservation phase fails, and the rollback mechanism restores the configuration before adjustment when the effective phase fails, ensuring the atomicity and fault tolerance of bandwidth adjustments. The hysteresis protection mechanism prohibits scaling down when the isolation has not fully fallen back after expansion, effectively preventing the ping-pong effect of frequent expansion and scaling down and ensuring the long-term stability of bandwidth adjustments.

[0148] This embodiment also provides a bandwidth management system, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0149] This embodiment provides a bandwidth management system, referring to... Figure 5 As shown, it includes: Application layer device 501 is used to divide the services in the network into multiple service modes according to the service level agreement requirements of each service carried in the network.

[0150] The control layer device 502, which is communicatively connected to the application layer device, is used to determine the service mode to which the target service belongs from multiple service modes, and to designate at least one service mode other than the target mode as an interference mode; to acquire the baseline performance data of the target mode under the condition of no interference mode disturbance; to receive the actual performance data of the target mode reported by the forwarding layer device under the condition of coexistence of interference modes; to determine the real-time isolation degree of the interference mode to the target mode based on the deviation between the actual performance data and the baseline performance data; and to send a bandwidth adjustment command to the forwarding layer device when the real-time isolation degree meets the preset isolation degree threshold condition.

[0151] The forwarding layer device 503 is communicatively connected to the control layer device and includes a performance data acquisition unit and a flexible Ethernet hard isolation execution unit. The performance data acquisition unit is used to acquire the actual performance data of the target mode under the condition of coexistence of interfering modes at a preset sampling period. The flexible Ethernet hard isolation execution unit is used to adjust the hard pipe bandwidth allocation corresponding to the target service according to the bandwidth adjustment command.

[0152] In some alternative implementations, the control layer device 502 includes: The intermodal influence detection module 5021 is used to send the collection strategy to the forwarding layer device, receive the raw performance data such as bandwidth, latency, jitter and packet loss rate reported by the forwarding layer device, and preprocess and time-align the raw performance data.

[0153] The disturbance quantification module 5022 is used to determine the deviation values ​​corresponding to each performance index of the target mode based on actual performance data and benchmark performance data, determine the isolation coefficient corresponding to each performance index, perform weighted fusion of each deviation value based on the isolation coefficient, and take the maximum value in the fusion result as the real-time isolation degree; it is also used to determine the expansion threshold based on the service level protocol critical samples in the historical samples whose degradation degree reaches the allowable upper limit, and to determine the reduction threshold based on the samples in the historical samples whose degradation degree meets the preset excellent conditions.

[0154] The dynamic hard pipe resource scheduling module 5023 is used to generate and send bandwidth adjustment instructions to the forwarding layer device when the real-time isolation meets the preset isolation threshold. Specifically, it is used to: generate a bandwidth adjustment instruction to increase the hard pipe bandwidth corresponding to the target service when the real-time isolation is greater than the expansion threshold; generate a bandwidth adjustment instruction to decrease the hard pipe bandwidth corresponding to the target service when the real-time isolation is less than the reduction threshold; and not generate a bandwidth adjustment instruction when the real-time isolation is greater than or equal to the reduction threshold and less than or equal to the expansion threshold.

[0155] In some optional implementations, the disturbance quantification module 5022 determines the isolation coefficient corresponding to each performance index, including: Multiple historical samples of the target modality are obtained. Each historical sample includes the historical deviation vector of the target modality at a historical moment and the corresponding service degradation label. The sample weights of each historical sample are determined according to the service level agreement (SLA) constraints of each performance indicator. Historical samples that are close to the critical state of the SLA are given higher sample weights. A sparse regression model with non-negative constraints is trained using the historical deviation vector as input and the service degradation label as the training target. The sparse regression model has contribution coefficients, sparse regularization terms, and smoothing regularization terms. The magnitude of the contribution coefficients is used to characterize the degree of contribution of each performance indicator to the service degradation label. The sparse regularization term is used to highlight the performance indicators that have the greatest impact on the service degradation label. The smoothing regularization term is used to suppress the oscillation of the contribution coefficients. The contribution coefficients are normalized to obtain the isolation coefficients corresponding to each performance indicator.

[0156] In some optional implementations, when there is a significant non-linear relationship between service performance degradation and performance deviation, the disturbance quantification module 5022 determines the isolation coefficient corresponding to each performance index, including: Multiple historical samples of the target modality are obtained. Each historical sample includes the historical deviation vector of the target modality at a historical moment and the corresponding business degradation label. A nonlinear prediction model is trained using the historical deviation vector as input and the business degradation label as the training target. The average contribution of each performance index to the business degradation label is calculated using the feature contribution method. The average contribution is normalized to obtain the isolation coefficient corresponding to each performance index.

[0157] In some optional implementations, the disturbance quantification module 5022 acquires historical samples, including: Based on the service level agreement constraints of the target business, determine the prior importance coefficients corresponding to each performance indicator; based on the historical deviation vectors in each historical sample and the corresponding prior importance coefficients, calculate the degree of degradation corresponding to each historical sample, which serves as the business degradation label.

[0158] In some optional implementations, the dynamic hard pipe resource scheduling module 5023 is specifically used for: Generate the version number and unified effective time for this adjustment; send reservation request messages in parallel to each forwarding layer device on the end-to-end path corresponding to the target service. The reservation request message carries the version number, the number of time slots to be adjusted, and the unified effective time; if all forwarding layer devices on the end-to-end path return confirmation messages, send a commit message to all forwarding layer devices so that each forwarding layer device can switch its local ineffective candidate calendar table to the effective calendar table at the unified effective time.

[0159] In some optional implementations, the dynamic hardpipe resource scheduling module 5023 is also used for: If no acknowledgment messages are received from all forwarding layer devices on the end-to-end path, an abort message is sent to the forwarding layer devices that have returned acknowledgment messages to instruct the forwarding layer devices to release the locked candidate time slots in the inactive candidate calendar table and keep the current active calendar table unchanged.

[0160] In some optional implementations, the dynamic hardpipe resource scheduling module 5023 is also used for: After sending the commit message, if there are forwarding layer devices on the end-to-end path that have failed to switch over, a rollback message is sent to each forwarding layer device on the end-to-end path according to the version number, instructing each forwarding layer device to restore to the calendar table configuration before the switchover.

[0161] In some optional implementations, the dynamic hardpipe resource scheduling module 5023 is also used for: After generating a bandwidth adjustment command to increase the hard pipe bandwidth corresponding to the target service, if the real-time isolation does not drop below the sum of the scaling-down threshold and the preset hysteresis, then generating a bandwidth adjustment command to decrease the hard pipe bandwidth corresponding to the target service is prohibited.

[0162] In some alternative implementations, the forwarding layer device 503 further includes: The performance data acquisition unit 5031 is used to acquire the actual performance data of the target mode under the condition of coexistence of interfering modes at a preset sampling period; the length of the sampling period is determined according to the priority of the target mode, and the higher the priority, the shorter the corresponding sampling period.

[0163] In some alternative implementations, the performance data acquisition unit 5031 includes at least one of the following: An active measurement probe that acquires end-to-end latency, jitter, and packet loss rate of the target mode based on a bidirectional active measurement protocol; and a physical layer indicator extraction unit that extracts time slot occupancy from the flexible Ethernet overhead based on the operation, management, and maintenance overhead mechanism of the flexible Ethernet protocol.

[0164] In some alternative implementations, the forwarding layer device 503 further includes: The Flexible Ethernet Hard Isolation Execution Unit 5032 is used to pre-lock candidate time slots in the local inactive candidate calendar table according to the bandwidth adjustment command. After successful locking, it returns a confirmation message to the control layer device. After receiving the submission message, it switches the inactive candidate calendar table to the active calendar table at the unified effective time to adjust the hard pipe bandwidth allocation corresponding to the target service.

[0165] In some alternative implementations, the system further includes: Infrastructure layer 504 is equipped with a time synchronization server to provide a unified time reference for control layer devices and various forwarding layer devices. The bandwidth management system provided in this application embodiment can execute the bandwidth management method provided in any embodiment of this application, possessing the corresponding functional modules and beneficial effects of the method execution. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0166] As a specific application embodiment of this application, the following describes the specific process of achieving dynamic isolation and resource optimization using the bandwidth management method provided in the above embodiments of this application when relay protection services (high priority, low latency), intelligent inspection video services (high bandwidth, low jitter), and remote control services (high reliability, low packet loss) share the same physical optical fiber and transmission equipment in a typical power communication access network. This specific application embodiment is only for the purpose of facilitating understanding of this application and does not constitute any limitation on the scope of protection of this application.

[0167] In practical applications, the bandwidth management method provided in this application can be applied to multi-service co-network carrying scenarios such as power communication networks, data center interconnection networks, or operator metropolitan area networks, and is not limited to the power communication access network described in this embodiment. The following will illustrate this with an example of a scenario where relay protection services, intelligent inspection video services, and remote control services in a power communication network share the same physical optical fiber.

[0168] (1) System deployment and initial configuration Reference Figure 6 The diagram shown is the overall system deployment architecture, which includes: Infrastructure layer: Deploy PTP (IEEE 1588v2) time synchronization across the entire network to ensure that the time deviation between nodes is less than 1 microsecond.

[0169] Forwarding Layer: Configure FlexE (Flexible Ethernet) interface cards on the substation exit PE router and the core equipment room PE router, with physical interfaces being 100GE optical ports. Divide this 100GE physical port into 20 5G time slots. The initial time slot allocation scheme is as follows: relay protection services (service mode A) occupy 10 time slots exclusively (guaranteed bandwidth 50G), video inspection services (service mode B) occupy 6 time slots (guaranteed bandwidth 30G), and remote control services (service mode C) occupy 4 time slots (guaranteed bandwidth 20G). All PE nodes enable the TWAMP Light reflection function and configure port traffic statistics and service flow statistics functions to collect the actual bandwidth usage of each service.

[0170] Control layer: The SDN controller manages all forwarding devices through NETCONF / YANG and deploys a disturbance quantification module, an inter-modal impact detection module, and a dynamic hard pipe resource scheduling module.

[0171] Application layer: Configure the business modality management module on the network management server, and operation and maintenance personnel can input the SLA requirements of each business through a graphical interface.

[0172] (2) Business mode segmentation and isolation degree model establishment Reference Figure 7 The process shown illustrates the specific configuration of the isolation metric model: Business Modal Division: Based on the SLA, the business is divided into three modalities: Mode A (Relay Protection): Requires a delay of <5ms, jitter of <50μs, and a packet loss rate of <10%. -6 ; Modal B (video inspection): Requires bandwidth > 20Mbps and jitter < 5ms; Modal C (Remote Control): Packet loss rate <10% required. - ³, latency <50ms.

[0173] Constructing an isolation model: Taking Mode A relay protection as an example, the system first collects 5 minutes of performance data under conditions of no additional disturbance flow to obtain the baseline performance vector of Mode A:

[0174] Subsequently, the system injects disturbance traffic from video inspection and remote control services, gradually changing the rate, burst period, and message length of the disturbance traffic. At each disturbance intensity, the system collects real-time performance data for Mode A with a sampling period of 10ms and calculates the normalized deviation vector. The components in this normalized bias vector have been dimensionlessly normalized by dividing by their respective reference values, eliminating the numerical scale differences in bandwidth, latency, jitter, and packet loss rate caused by different units (such as Gbps, ms, μs, percentage). Subsequently, service degradation labels are generated based on the SLA prior importance coefficients of mode A. .

[0175] Since relay protection services are more sensitive to time delay and packet loss rate, the SLA prior importance coefficient can be manually set as follows:

[0176] After obtaining the training sample set, the system assigns higher weights to samples close to the SLA critical state and trains a sparse regression model with non-negativity constraints. The model is solved and normalized to obtain the isolation coefficient vector:

[0177] The results indicate that, in the current network scenario, latency deviation contributes the most to the degradation of relay protection services, followed by jitter and packet loss rate, while bandwidth deviation contributes relatively little. Therefore, in subsequent real-time operation, the system uses the isolation coefficient vector to calculate the disturbance level of mode A and triggers the hard-pipe bandwidth adjustment process accordingly.

[0178] After obtaining the isolation coefficient vector, based on the selection of critical SLA samples (degradation just reaching the allowable upper limit) and high-efficiency samples (degradation excellent), the isolation coefficient values ​​are calculated for each, thereby determining the tolerance upper limit. and tolerance limit The initial value. For example, the lower threshold and the upper threshold are calculated. .

[0179] Choose norm Since mode A focuses on the degradation of the worst-case performance dimension (the weakest link effect), this embodiment chooses the Chebyshev norm, i.e. .

[0180] (3) Example of periodic performance acquisition and degree of interference The system collects data at a period of 10ms.

[0181] At a certain monitoring time Due to extreme weather conditions such as storms and heavy rain, the traffic of relay protection services (Mode A) surged, causing momentary congestion in the queues within the FlexE time slots. The real-time performance data reported by the probes is as follows (from the perspective of Mode A, calculating the impact of other modes):

[0182] Calculate the normalized bias vector :

[0183] Calculating the real-time interference level using the weighted Chebyshev norm : ; .

[0184] Judgment result: Calculated Far exceeding the upper limit threshold The surge in traffic caused severe latency and jitter interference to Mode A, necessitating increased bandwidth.

[0185] (4) Hard pipe bandwidth adjustment process because This triggers the hard-pipe bandwidth adjustment mechanism, adding isolation resources for mode A to reduce interference.

[0186] Reference Figure 8 The hard-pipe bandwidth adjustment process shown below includes the following specific steps: ①Initiation Phase: Based on the SLA requirements of Mode A, the current time slot utilization rate, and the remaining available time slot resources in the network, the SDN controller determines the target of this expansion as follows: increase the hard pipe bandwidth of Mode A from the original 10 FlexE time slots to 12 FlexE time slots, that is, add 2 5G time slots, so that the end-to-end hard pipe bandwidth of Mode A is increased from 50G to 60G.

[0187] In this embodiment, before officially switching to FlexE Calendar, the Shadow Calendar reservation process is first executed. The control layer generates a unique adjustment version number for this adjustment. And determine a unified effective date. .in, It is determined by adding the reserved protection time to the current time, for example:

[0188] in, The protection time is preset by the control layer based on the end-to-end path length, node processing delay, and clock synchronization accuracy.

[0189] Subsequently, the control layer sends reservation request messages in parallel to all relevant forwarding nodes on the Mode A service path:

[0190] Where A represents the target mode as relay protection service. This indicates that two new FlexE time slots need to be added. This indicates the version number of this bandwidth adjustment. This indicates the unified effective date.

[0191] Upon receiving the PRE-RESERVE message, the source PE node, intermediate nodes P1 and P2, and the destination PE node each check the available time slot resources in their local FlexE Calendar. If a node has an available time slot, it does not immediately modify the currently used Active Calendar, but instead pre-locks a candidate time slot in its local Shadow Calendar and returns the result to the control layer.

[0192] If a node does not have a suitable idle time slot, or if a candidate time slot conflicts with other high-priority services, then the node returns:

[0193] In this embodiment, the source PE node, intermediate nodes P1 and P2, and the destination PE node all successfully reserved two candidate time slots in the Shadow Calendar and returned RESERVE-ACK messages to the control layer. After receiving RESERVE-ACK messages from all nodes on the end-to-end path, the control layer determines that the expansion meets the conditions for consistent effectiveness and then sends commit messages to all relevant nodes in parallel.

[0194] After receiving the COMMIT message, each node does not immediately apply the new time slot allocation. Instead, it maintains the current ActiveCalendar to continue carrying the service flow and waits for the unified effective time. Because this system's infrastructure layer deploys a PTP / NTP time synchronization server, each forwarding node can synchronize based on a unified clock. Upon arrival, a Calendar switchover operation is performed synchronously. Specifically, each node... The local Shadow Calendar is switched to the new Active Calendar at any time, which increases the hard pipe bandwidth of Mode A on the end-to-end path from 10 time slots to 12 time slots.

[0195] In this way, the expansion and contraction of Mode A no longer rely on the traditional hop-by-hop serial handshake, but are completed through unified scheduling at the control layer, end-to-end parallel reservation, and unified timing activation. This method can reduce the negotiation waiting time in multi-hop links and avoid instantaneous packet loss, time slot misalignment, or service interruption caused by asynchronous Calendar switching between different nodes.

[0196] In the next sampling cycle after the capacity expansion is completed, the control layer continues to collect performance indicators such as bandwidth, latency, jitter, and packet loss rate of mode A, and recalculates the real-time isolation. If the isolation of mode A decreases to within the tolerance range after expansion, the control layer will maintain the current hard pipe configuration of 12 time slots. If the isolation is still higher than the tolerance limit, the control layer can continue to perform secondary expansion according to the shadow calendar reservation mechanism mentioned above until the SLA requirements are met or the maximum allocatable bandwidth limit for this service mode is reached.

[0197] If, during the reservation phase, any node on the end-to-end path returns RESERVE-NACK, or if the control layer does not receive RESERVE-ACKs from all nodes within a preset time, the control layer determines that the bandwidth adjustment does not meet the conditions for consistent effectiveness and sends the following to the nodes that have completed reservation:

[0198] Upon receiving the ABORT message, each node releases the locked candidate time slots in the Shadow Calendar and maintains the original Active Calendar. If a node anomaly occurs after the COMMIT phase but before the official implementation, the control layer will adjust the version number accordingly. Send to the relevant nodes:

[0199] The relevant nodes restore the original time slot allocation based on the locally saved Calendar configuration before the adjustment, thereby avoiding the end-to-end hard pipe inconsistency problem caused by some nodes successfully expanding while others fail to expand.

[0200] Through the bandwidth adjustment mechanism of this embodiment, when sudden service interference occurs, Mode A can trigger capacity expansion based on the isolation model, complete end-to-end time slot reservation through shadow calendar, reduce negotiation latency through parallel confirmation, ensure consistency of multi-hop calendar switching through unified timing, and ensure the reliability of the adjustment process through failure abort and rollback mechanisms. Compared with the traditional serial handshake method, this embodiment can achieve faster, more reliable, and rollback-capable dynamic adjustment of hard-pipe bandwidth without interrupting relay protection services.

[0201] Through the specific implementation of this embodiment, it is verified that this application can dynamically reclaim idle bandwidth resources while ensuring deterministic latency (jitter control at the microsecond level) of relay protection services, and realize refined isolation and efficient resource utilization under multi-modal power service co-network.

[0202] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0203] The following is a detailed reference. Figure 9 The diagram illustrates a structural schematic suitable for implementing the electronic device described in the embodiments of this application. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 901, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 902 or a program loaded from memory 908 into random access memory (RAM) 903. The RAM 903 also stores various programs and data required for the operation of the electronic device. The processor 901, ROM 902, and RAM 903 are interconnected via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0204] Typically, the following devices can be connected to I / O interface 905: input devices 906 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 907 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 908 including, for example, magnetic tapes, hard disks, etc.; and communication devices 909. Communication device 909 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 9 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0205] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 909, or installed from a memory 908, or installed from a ROM 902. When the computer program is executed by the processor 901, it performs the functions defined in the bandwidth management method of embodiments of this application.

[0206] Figure 9 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0207] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the bandwidth management method shown in the above embodiments is implemented.

[0208] A portion of this application can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to this application through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0209] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A bandwidth management method for multi-service modal isolation scenarios, characterized in that, include: Based on the service level agreement requirements of each service carried in the network, the services in the network are divided into multiple service modes. The service mode to which the target service belongs is determined from the multiple service modes and is taken as the target mode. At least one service mode other than the target mode is taken as an interference mode. Obtain the actual performance data of the target mode under the coexistence condition of the interference mode, and the baseline performance data of the target mode under the condition of no disturbance of the interference mode; The real-time isolation degree of the interference mode from the target mode is determined based on the deviation between the actual performance data and the benchmark performance data. If the real-time isolation meets the preset isolation threshold, the hard pipe bandwidth allocation corresponding to the target service is adjusted.

2. The bandwidth management method according to claim 1, characterized in that, Obtaining the actual performance data of the target mode under the coexistence condition of the interfering modes includes: The length of the preset sampling period is determined according to the priority of the target mode, wherein the higher the priority, the shorter the corresponding sampling period; The performance data of the target mode is collected at the preset sampling period, including at least one of the following methods: The end-to-end latency, jitter, and packet loss rate of the target mode are obtained based on a bidirectional active measurement protocol. Based on the operation, management, and maintenance overhead mechanism of the Flexible Ethernet protocol, time slot occupancy rate is extracted from the Flexible Ethernet overhead, and bandwidth occupancy data is determined based on the time slot occupancy rate; and... The actual service utilization rate of the target mode is determined based on port traffic statistics or service traffic statistics.

3. The bandwidth management method according to claim 1, characterized in that, Determining the real-time isolation degree of the interference mode from the target mode based on the deviation between the actual performance data and the benchmark performance data includes: Based on the actual performance data and the benchmark performance data, the deviation values ​​corresponding to each performance index of the target mode are determined respectively, and the performance index includes at least two of bandwidth, latency, jitter and packet loss rate; Determine the isolation coefficient corresponding to each of the aforementioned performance indicators, perform dimensionless normalization on each of the aforementioned deviation values, and then perform weighted fusion on each of the aforementioned deviation values ​​based on the isolation coefficients. The maximum value in the fusion result is taken as the real-time isolation degree.

4. The bandwidth management method according to claim 3, characterized in that, Determining the isolation coefficient corresponding to each of the aforementioned performance indicators includes: Obtain multiple historical samples of the target mode, each historical sample including the historical deviation vector of the target mode at a historical moment and the corresponding service degradation label; Based on the service level agreement constraint values ​​of each performance indicator, the sample weight corresponding to each historical sample is determined; wherein, historical samples that are close to the service level agreement critical state are assigned higher sample weights. Using the historical deviation vector as input and the business degradation label as the training target, a sparse regression model with non-negative constraints is trained. The sparse regression model has a contribution coefficient, a sparse regularization term, and a smoothing regularization term. The contribution coefficient is used to characterize the degree of contribution of each performance indicator to the business degradation label. The sparse regularization term is used to highlight the performance indicator with the greatest impact on the business degradation label. The smoothing regularization term is used to suppress the oscillation of the contribution coefficient. The contribution coefficients are normalized to obtain the isolation coefficients corresponding to each of the performance indicators.

5. The bandwidth management method according to claim 3, characterized in that, Determining the isolation coefficient corresponding to each of the aforementioned performance indicators includes: Obtain multiple historical samples of the target mode, each historical sample including the historical deviation vector of the target mode at a historical moment and the corresponding service degradation label; A nonlinear prediction model is trained using the historical deviation vector as input and the business degradation label as the training target. The average contribution of each performance indicator to the business degradation label is calculated using the feature contribution method; The average contribution of each of the above is normalized to obtain the isolation coefficient corresponding to each of the above performance indicators.

6. The bandwidth management method according to claim 4 or 5, characterized in that, The acquisition of multiple historical samples of the target modality includes: Based on the service level agreement constraint value of the target service, determine the prior importance coefficient corresponding to each of the performance indicators; Based on the historical deviation vector in each historical sample and the corresponding prior importance coefficient, the degree of degradation corresponding to each historical sample is calculated and used as the business degradation label.

7. The bandwidth management method according to claim 6, characterized in that, The preset isolation threshold conditions include an expansion threshold determined based on service level agreement critical samples in the historical samples whose degradation level reaches the allowable upper limit, and a reduction threshold determined based on samples in the historical samples whose degradation level meets preset excellent conditions. When the real-time isolation meets the preset isolation threshold, the hard pipe bandwidth allocation corresponding to the target service is adjusted, including: If the real-time isolation is greater than the expansion threshold, increase the hard pipe bandwidth corresponding to the target service; If the real-time isolation is less than the scaling-down threshold, reduce the hard pipe bandwidth corresponding to the target service; If the real-time isolation is greater than or equal to the shrinkage threshold and less than or equal to the expansion threshold, the hard pipe bandwidth corresponding to the target service remains unchanged.

8. The bandwidth management method according to claim 7, characterized in that, The adjustment of the hard bandwidth allocation corresponding to the target service also includes: Generate the version number and unified effective time for this adjustment; A reservation request message is sent in parallel to each forwarding node on the end-to-end path corresponding to the target service. The reservation request message carries the version number, the number of time slots to be adjusted, and the unified effective time. Each of the aforementioned forwarding nodes pre-locks a candidate time slot in its local inactive candidate calendar table and returns a confirmation message after successfully locking the slot; If all forwarding nodes on the end-to-end path return the confirmation message, a commit message is sent to all the forwarding nodes. Under the control of the submission message, each of the forwarding nodes switches the inactive candidate calendar table to the effective calendar table within the unified effective time period to complete the hard-pipe bandwidth allocation adjustment for the target service.

9. The bandwidth management method according to claim 8, characterized in that, The adjustment of the hard bandwidth allocation corresponding to the target service also includes: If no confirmation message is received from all forwarding nodes on the end-to-end path, a stop message is sent to the forwarding nodes that have returned confirmation messages to instruct the forwarding nodes to release the locked candidate time slots in the inactive candidate calendar table and keep the current active calendar table unchanged.

10. The bandwidth management method according to claim 8, characterized in that, The adjustment of the hard bandwidth allocation corresponding to the target service also includes: After sending the commit message, if there are forwarding nodes on the end-to-end path that have failed to switch over, a rollback message is sent to each forwarding node on the end-to-end path according to the version number to instruct each forwarding node to restore to the calendar table configuration before the switchover.

11. The bandwidth management method according to claim 8, characterized in that, The adjustment of the hard bandwidth allocation corresponding to the target service also includes: After increasing the hard bandwidth corresponding to the target service, if the real-time isolation does not drop below the sum of the scaling-down threshold and the preset backlash, the step of reducing the hard bandwidth corresponding to the target service is prohibited.

12. A bandwidth management system, characterized in that, include: Application layer devices are used to divide the services in the network into multiple service modes according to the service level agreement requirements of each service carried in the network. A control layer device, communicatively connected to the application layer device, is used to determine the service mode to which the target service belongs from the plurality of service modes, as the target mode, and to designate at least one service mode other than the target mode from the plurality of service modes as an interference mode; The forwarding layer device is communicatively connected to the control layer device and includes a performance data acquisition unit and a flexible Ethernet hard-isolated execution unit; The performance data acquisition unit is used to acquire the actual performance data of the target mode under the coexistence condition of the interference modes; The control layer device is also used to acquire the baseline performance data of the target mode under the condition of no disturbance of the interference mode, determine the real-time isolation degree of the interference mode to the target mode based on the deviation between the actual performance data and the baseline performance data, and send a bandwidth adjustment command to the flexible Ethernet hard isolation execution unit when the real-time isolation degree meets the preset isolation degree threshold condition. The flexible Ethernet hard isolation execution unit is used to adjust the hard pipe bandwidth allocation corresponding to the target service according to the bandwidth adjustment instruction.

13. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory stores computer instructions, and the processor executes the bandwidth management method according to any one of claims 1 to 11 by executing the computer instructions.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the bandwidth management method according to any one of claims 1 to 11.

Images