Optical communication network resource intelligent scheduling method and system

Abstract

The application relates to the technical field of intelligent scheduling of optical communication network resources, and discloses an intelligent scheduling method and system for optical communication network resources, which can comprehensively master the network operation state by acquiring equipment operation data, scheduling decision data and service performance data and performing time synchronization. On this basis, the method can identify weak fluctuations below a preset alarm threshold and perform risk marking, effectively solving the problem that weak fluctuations are difficult to be found in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent scheduling of optical communication network resources, and particularly relates to an intelligent scheduling method and system for optical communication network resources. BACKGROUND

[0002] Modern optical communication networks are the backbone of various digital services, and the core lies in an intelligent scheduling system that efficiently and automatically allocates network resources. The system continuously collects real-time running state information of various resources such as fiber links, wavelengths, optical power, and devices (such as optical transponders, optical amplifiers, and reconfigurable optical add-drop multiplexers (ROADM)) in the network, and constructs a dynamically updated network view. Based on this view, the system uses complex optimization methods to calculate a transmission path that meets service requirements (such as low latency and high bandwidth) and takes into account the overall network efficiency for new service requests, and automatically completes wavelength allocation and device configuration.

[0003] However, in actual operation, network devices may have characteristics that are difficult to detect, causing the network to appear normal on the surface while hiding risks that affect service quality. For example, after a routine software version upgrade, a specific batch of reconfigurable optical add-drop multiplexers (ROADM) devices deployed in the network begin to exhibit abnormal behavior that is difficult to detect due to a small timing compatibility problem between the new version firmware and some early production hardware modules. This abnormality is manifested as extremely weak and intermittent fluctuations in optical power loss when the internal optical switch matrix performs wavelength routing switching under certain load conditions, such as when the ROADM needs to handle a large number of wavelength channels for routing switching, or when the environmental temperature reaches a certain critical value. The fluctuation amplitude is usually between 0.1 dB and 0.3 dB, which is much lower than the system's preset alarm threshold (usually 0.5 dB or higher), so it does not trigger any alarm information, and these ROADM devices still display as "normal operation" on the network management system's monitoring interface.

[0004] Further, due to such intermittent fluctuations, the affected ROADM devices also produce slight deviations in the data collected by their internal optical power monitoring modules when reporting their own state data to the network management system. For example, after a certain wavelength channel is allocated and activated, the actual output optical power value of the channel reported by the ROADM may be slightly lower than its nominal value for a short time, or the internal optical signal-to-noise ratio (OSNR) monitoring value may also have a slight instantaneous drop at some time. These deviations are also very subtle and are not enough to be identified as error data by the data checking mechanism of the network management system, as they are still within the "normal" measurement error range and are considered as normal measurement noise or system jitter. The data preprocessing module of the scheduling system has a certain ability to filter abnormal data, but its design goal is to eliminate obvious outliers or format errors, and it is powerless for such "legal" but "inaccurate" weak fluctuations.

[0005] The prior art needs to be improved in view of the above problems. SUMMARY

[0006] The present application provides an optical communication network resource intelligent scheduling method and system, aiming to solve the problem that the existing optical communication network intelligent scheduling system cannot effectively perceive, quantify and avoid potential risk paths when facing weak and intermittent physical layer performance fluctuations, resulting in a decline in the quality of high-value business services and difficulty in diagnosis.

[0007] In a first aspect, to solve the above technical problems, the present application provides an optical communication network resource intelligent scheduling method, comprising:

[0008] Obtaining device operation data, scheduling decision data and service performance data and performing time synchronization;

[0009] According to the device operation data after time synchronization, identifying weak fluctuations below a preset alarm threshold and performing risk labeling;

[0010] When the performance of the service is monitored to decline, obtaining the service identifier and the time point of performance decline, backtracking the optical path of the service, and according to the optical path, the risk label, the scheduling decision data and the service performance data, correlatively analyzing the causal relationship between the weak fluctuation record and the performance decline;

[0011] According to the causal relationship, adjusting the path selection preference of the service.

[0012] Preferably, the correlatively analyzing the causal relationship between the weak fluctuation record and the performance decline according to the optical path, the risk label, the scheduling decision data and the service performance data comprises:

[0013] acquire optical power and optical signal-to-noise ratio data and radio frequency interference intensity data of the device on the optical path within the time period of the performance decline;

[0014] evaluate whether the device performs zero-point calibration;

[0015] if yes, issue an instruction to the device so that the device switches to an idle state or a standby bypass within a preset time window;

[0016] within the preset time window, collect optical power and optical signal-to-noise ratio data of the device, and synchronously collect radio frequency interference intensity data;

[0017] quantify the contribution proportion of the weak fluctuation according to the device data collected in the zero-point calibration and the radio frequency interference intensity data;

[0018] associate the cause-effect correlation according to the optical path, the risk label, the scheduling decision data, the service performance data, and the contribution proportion.

[0019] Preferably, when the performance of the service is monitored to decline, the service identifier and the occurrence time point of the performance decline are acquired, and the optical path passed by the service is traced back, comprising:

[0020] when the performance of the service is monitored to decline, the path tracing program is triggered according to the service identifier and the occurrence time point;

[0021] acquire the allocation record of the service, the allocation record comprising an allocation path, a wavelength channel, and an accurate time stamp;

[0022] acquire path distribution information of actual transmission traffic of the service within the preset time window, the path distribution information comprising a traffic proportion and a time stamp of each path;

[0023] compare the allocation record with the path distribution information of the actual transmission traffic to identify a switching time point and a switched path of the traffic between multiple optical paths within the preset time window;

[0024] construct a time series distribution diagram of the traffic on each optical path within the preset time window according to the switching time point and the switched path;

[0025] according to the time series distribution diagram, in combination with the duration of the performance decline of the service, an optical path carrying main traffic of the service is taken as the optical path passed by the service.

[0026] Preferably, the method further comprises:

[0027] During the performance degradation of the service, the traffic proportion of each optical path in the time series distribution diagram is divided into fine-grained time windows, and the traffic proportion in each time window is counted;

[0028] All optical paths overlapping in the performance degradation duration of the service are identified, and for each overlapping optical path, a corresponding weak fluctuation record is obtained;

[0029] According to the counting result and the weak fluctuation record, the potential contribution value of each overlapping optical path to the performance degradation of the service is calculated;

[0030] The overlapping optical path with the highest potential contribution value is identified as the optical path.

[0031] Preferably, the method further comprises:

[0032] According to the type of the service, a sensitivity weight of the service is obtained;

[0033] Based on the sensitivity weight, the weights of the traffic proportion and the weak fluctuation record in the potential contribution value are dynamically adjusted;

[0034] The adjusted weights are applied to the traffic proportion and the weak fluctuation record to calculate the potential contribution value of each overlapping optical path to the performance degradation of the service.

[0035] Preferably, the method further comprises:

[0036] Obtaining a service characteristic parameter of an emerging service or a customized service;

[0037] In a controlled network environment, the transmission of the emerging service or the customized service under different combinations of bit error rate, delay, and packet loss rate is simulated, and the performance and user experience feedback of the service under each combination are recorded;

[0038] According to the service characteristic parameter, the performance, and the user experience feedback, a sensitivity curve of the emerging service or the customized service is constructed;

[0039] According to the sensitivity curve, the sensitivity weight of the emerging service or the customized service is extracted.

[0040] Preferably, the extracting the sensitivity weight of the emerging service or the customized service according to the sensitivity curve comprises:

[0041] In the sensitivity curve, a performance degradation point corresponding to a specific performance degradation threshold is identified;

[0042] Starting from the performance degradation point, a local interval containing the performance degradation point and having relatively stable curve morphology is identified by extending to both sides of the sensitivity curve;

[0043] In the local interval, an inflection point or a plateau starting point of the curve is identified according to the slope and curvature change trend of the sensitivity curve;

[0044] According to the inflection point or the plateau starting point, the local interval is divided into a plurality of sub-intervals;

[0045] For each sub-interval, the corresponding average sensitivity or dominant sensitivity is calculated;

[0046] According to the average sensitivity or the dominant sensitivity of the sub-interval and the position of the specific performance degradation threshold in the local interval, the sensitivity weight is weighted calculated.

[0047] Preferably, the identifying the inflection point or the plateau starting point of the curve in the local interval according to the slope and curvature change trend of the sensitivity curve comprises:

[0048] In the local interval, the sensitivity curve is subjected to a moving average processing;

[0049] The first-order difference and the second-order difference of the sensitivity curve after the moving average processing are calculated;

[0050] According to the continuous change trend of the first-order difference and the second-order difference, a point with maximum slope change rate or continuous change of the second-order difference sign is identified as the inflection point;

[0051] A region in which the first-order difference continuously and stably stays in a preset small range and the continuous time exceeds a preset threshold is identified as the plateau starting point.

[0052] Preferably, after the region in which the first-order difference continuously and stably stays in a preset small range and the continuous time exceeds a preset threshold is identified as the plateau starting point, the method further comprises:

[0053] The identified inflection point or the plateau starting point is subjected to consistency verification, and when a plurality of similar points meet the identification condition, a point with the largest change amplitude is selected as the final inflection point or plateau starting point.

[0054] In a second aspect, the present application provides an optical communication network resource intelligent scheduling system, comprising:

[0055] an input end configured to acquire device operation data, scheduling decision data, and service performance data and perform time synchronization, identify weak fluctuations below a preset alarm threshold from the device operation data after time synchronization, and perform risk marking;

[0056] an analysis end configured to, when performance degradation of a service is monitored, acquire a service identifier and a time point of performance degradation, trace an optical path of the service, and analyze a cause-effect relationship between the weak fluctuations and the performance degradation according to the optical path, the risk marking, the scheduling decision data, and the service performance data;

[0057] an adjustment end configured to adjust path selection preference of the service according to the cause-effect relationship.

[0058] Compared with the prior art, the present application has the following beneficial effects:

[0059] By acquiring device operation data, scheduling decision data, and service performance data and performing time synchronization, the network operation state can be comprehensively mastered. On this basis, the method can identify weak fluctuations below a preset alarm threshold and perform risk marking, effectively solving the problem that weak fluctuations are difficult to be found in the prior art. When performance degradation of a service is monitored, the method can acquire a service identifier and a time point of performance degradation, trace an optical path of the service, and then analyze a cause-effect relationship between weak fluctuation records and performance degradation according to the optical path, the risk marking, the scheduling decision data, and the service performance data. This innovative cause-effect relationship analysis mechanism breaks through the limitation of the prior scheduling system that only relies on surface “normal” data for decision-making, and can reveal potential risks hidden behind weak fluctuations. Finally, the method adjusts path selection preference of the service according to the cause-effect relationship, so as to actively avoid optical paths with potential risks and ensure stable transmission of high-value services. Through the above technical solutions, the present application effectively solves the problem that the intelligent scheduling system in the prior art cannot perceive and quantify the cumulative impact of weak physical layer performance fluctuations on optical path transmission quality, leading to suboptimal resource allocation and hidden risks, and significantly improves the precision, reliability, and service quality of optical communication network resource scheduling. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 is a flowchart of an optical communication network resource intelligent scheduling method provided by an embodiment of the present application;

[0061] Figure 2 is a flowchart of a method for analyzing weak fluctuation records and cause-effect relationships provided by an embodiment of the present application;

[0062] Figure 3A schematic diagram of an optical communication network resource intelligent scheduling system structure is provided by an embodiment of the present application. DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0064] Reference Figure 1 , Figure 1 A flowchart of an optical communication network resource intelligent scheduling method is provided by an embodiment of the present application, comprising the following steps:

[0065] S1, obtaining device running data, scheduling decision data and service performance data and performing time synchronization;

[0066] S2, identifying weak fluctuations below a preset alarm threshold and performing risk marking according to the time-synchronized device running data;

[0067] S3, when the performance of a service is monitored to decrease, obtaining the service identifier and the time point of performance decrease, backtracking the optical path passed by the service, and according to the optical path, the risk marking, the scheduling decision data and the service performance data, associating and analyzing the causal relationship between the weak fluctuation record and the performance decrease;

[0068] S4, adjusting the path selection preference of the service according to the causal relationship.

[0069] The conventional existing optical communication network intelligent scheduling system mainly relies on the real-time running state data reported by the device when performing resource allocation. However, these data may have weak and intermittent physical layer performance fluctuations, such as slight changes in optical power loss or instantaneous decrease in signal-to-noise ratio. These fluctuations are often below the preset alarm threshold, causing the scheduling system to be unable to perceive and quantify the potential impact on service transmission quality, so that high-value services may be allocated to paths with potential risks, ultimately leading to service performance decrease and difficulty in diagnosis.

[0070] To this end, the present application proposes an intelligent scheduling method for optical communication network resources. By acquiring device operation data, scheduling decision data, and service performance data and performing time synchronization, the method can identify weak fluctuations below the preset alarm threshold and mark them as risks. When a decrease in service performance is detected, the method can backtrack the optical path taken by the service and comprehensively analyze the causal relationship between the weak fluctuation records and the performance decline based on the optical path, risk markers, scheduling decision data, and service performance data. In this way, the subsequent path selection preferences of the service can be adjusted according to the causal relationship, thereby effectively avoiding potential risks and improving the stability and reliability of service transmission.

[0071] In order to better understand the intelligent scheduling method for optical communication network resources proposed by the present application, some key terms involved in the method are first explained.

[0072] "Device operation data" refers to real-time or historical data generated by various devices in the optical communication network during operation, such as optical transponders, optical amplifiers, reconfigurable optical add-drop multiplexers (ROADM), etc. These data include but are not limited to physical layer parameters such as optical power, optical signal-to-noise ratio (OSNR), bit error rate (BER), temperature, voltage, current, as well as device operating mode, port state, alarm information, etc. These data are the basis for evaluating the health status and performance of network devices.

[0073] "Scheduling decision data" refers to the strategies, rules, and historical decision records that the intelligent scheduling system relies on when allocating network resources. This includes but is not limited to path selection algorithm parameters, wavelength allocation strategies, service priority settings, network topology information, resource capacity limitations, and historical scheduling results, etc. These data provide guidance and reference for future scheduling by the system.

[0074] "Service performance data" refers to the performance indicators of various services carried in the network during transmission, such as delay, jitter, packet loss rate, throughput, bit error rate, and user experience feedback, etc. These data directly reflect the quality of service (QoS) and user experience (QoE) of the service, and are an important basis for evaluating the service level of the network.

[0075] "Weak fluctuation" refers to abnormal changes in device operation data that have small amplitude and may be intermittent. These changes are usually below the conventional alarm threshold, so they are not easily detected by traditional monitoring systems. For example, a transient decrease in optical power within the range of 0.1dB to 0.3dB, or a slight fluctuation in OSNR.

[0076] "Risk marker" refers to the labeling of identified weak fluctuations, indicating their potential impact on service performance in terms of risk level or type.

[0077] "Lightpath path" refers to a series of fiber links and optical network devices that service data passes through from a source point to a destination point in an optical communication network.

[0078] "Causal correlation" refers to determining that weak fluctuations are the cause or significant influence of service performance degradation by analyzing the time, space and logical relationship between weak fluctuation records and service performance degradation.

[0079] The implementation environment of the present application is generally an intelligent scheduling system in an optical communication network, which has data acquisition, processing, analysis, decision and control capabilities, and can interact with various types of optical equipment in the network and issue instructions.

[0080] The core of the optical communication network resource intelligent scheduling method proposed in the present application is the perception, analysis and utilization of weak fluctuations in the network, so as to optimize the path selection of services.

[0081] First, device operation data, scheduling decision data and service performance data need to be obtained and time-synchronized. Device operation data can be periodically collected from optical network equipment through a network management system (NMS) or a special performance monitoring agent. For example, a ROADM device can be configured to report optical power and OSNR data of its ports every 100 milliseconds. Scheduling decision data can be directly read from the database of the intelligent scheduling system, including the current effective scheduling strategy and historical scheduling records. Service performance data can be monitored in real time by deploying performance probes at the network edge or service servers, such as sampling the delay, packet loss rate and bit error rate of specific service flows. In order to ensure the accuracy of subsequent analysis, all obtained data need to be time-synchronized. This can be achieved by using high-precision time protocols (such as NTP or PTP) to calibrate the timestamps of different data sources, ensuring that all data points can accurately correspond to the same time axis.

[0082] Secondly, according to the time-synchronized device operation data, the weak fluctuations below the preset alarm threshold are identified and marked with risk. Traditional alarm systems usually only focus on significant abnormalities that exceed a certain threshold. However, the present application focuses on those "legal" but "inaccurate" weak fluctuations. For example, the optical power data can be processed with a sliding window average, and the deviation between each data point and the sliding average is calculated. If the deviation of a certain data point continuously exceeds a small preset threshold (e.g. 0.2 dB), but is lower than the conventional alarm threshold (e.g. 0.5 dB), it can be identified as a weak fluctuation. For the identified weak fluctuations, different risk labels can be assigned according to factors such as their duration, amplitude, frequency of occurrence, and the type of equipment involved. For example, weak fluctuations with longer duration and larger amplitude can be labeled as "high risk", while short and small amplitude fluctuations can be labeled as "low risk". These risk labels will serve as important inputs for subsequent causal correlation analysis.

[0083] Furthermore, when a performance degradation of a service is monitored, the service identifier and the time point of occurrence of the performance degradation are obtained, the optical path path through which the service passes is traced back, and the causal relationship between the weak fluctuation record and the performance degradation is analyzed according to the optical path path, the risk label, the scheduling decision data and the service performance data. The monitoring of the performance degradation of the service can be realized by continuously analyzing the service performance data. For example, when the bit error rate of a certain service significantly increases in a short period of time, or the delay exceeds the preset service level agreement (SLA) threshold, it is determined that the performance has degraded. Once the performance degradation is monitored, the system immediately records the unique identifier of the service and the precise time point of the performance degradation.

[0084] Subsequently, the system needs to trace back the optical path path through which the service passes during the performance degradation. This can be achieved by querying the historical path allocation records of the scheduling system. For example, if a service is allocated to path A in a certain period of time, and to path B in another period of time, it is necessary to identify which path the service is mainly carried on when the performance degradation occurs. After determining the optical path path, the system will combine the weak fluctuation records (with risk labels) of all devices on the path, the historical scheduling decision data and the performance data of the service itself for comprehensive correlation analysis. For example, if there is a weak fluctuation marked as "high risk" on the path through which the performance-degraded service passes at a certain time point, and the fluctuation highly coincides with the time point of the performance degradation of the service, it can be preliminarily judged that there is a causal relationship between the two.

[0085] Finally, according to the causal relationship, the path selection preference of the subsequent business is adjusted. Once the causal relationship between the weak fluctuation and the performance decline of the business is determined, the system will use this information to optimize future scheduling decisions. For example, if it is found that the weak fluctuation on a certain ROADM device leads to the performance decline of a certain business, the system will reduce the priority of the path containing the ROADM device or completely avoid using the device when selecting the path for the business or similar sensitive businesses in the future. This can be achieved by introducing a "risk weight" or "blacklist" mechanism in the scheduling algorithm. For example, the cost factor of the device or link with weak fluctuation can be increased, so that the scheduling algorithm tends to choose other more stable paths when optimizing the objective function. This adjustment can be dynamic and continuously updated as the network state changes and new weak fluctuation records appear.

[0086] The optical communication network resource intelligent scheduling method proposed in the present application effectively solves the problem that the traditional scheduling system cannot perceive and avoid hidden network risks by introducing the identification and risk marking of weak fluctuations below the preset alarm threshold.

[0087] The core innovation of the present application is that it breaks through the limitation of traditional scheduling systems that only focus on explicit alarms, and incorporates the perception and quantification of "weak fluctuations" into the scheduling decision-making process. By performing fine-grained analysis on device operation data, the present application can identify and mark weak fluctuations below the preset alarm threshold, revealing potential and hidden risk points in the network. Further, when the performance of a business declines, the present application can backtrack the optical path of the business and perform causal relationship analysis by combining risk marking, scheduling decision data and business performance data, thereby accurately locating the weak fluctuation that caused the performance decline.

[0088] In some embodiments of the present application described above, a causal relationship analysis of weak fluctuation records and performance decline is proposed based on optical path, risk marking, scheduling decision data and business performance data. However, in actual application, direct causal relationship analysis based on these data may not completely rule out the influence of device measurement errors or environmental factors (such as radio frequency interference) on optical signal quality, resulting in inaccurate judgment of the causal relationship between weak fluctuations and performance decline, or even misjudgment. If the above problems are not solved, it may lead to incorrect scheduling decisions, affecting the stability of the business and user experience. To this end, the present application further proposes a more accurate causal relationship analysis method, which introduces a zero-point calibration mechanism to quantify the real contribution ratio of weak fluctuations, thereby improving the accuracy of causal analysis.

[0089] Reference Figure 2 , Figure 2A method flowchart for correlating weak fluctuation records and causal relationships is provided, and specifically comprises the following steps:

[0090] S31, acquiring optical power, optical signal-to-noise ratio data and radio frequency interference intensity data of the device on the optical path within the performance decline period;

[0091] S32, evaluating whether the device can perform zero-point calibration;

[0092] S33, if yes, issuing an instruction to the device to switch to an idle state or a standby bypass within a preset time window;

[0093] S34, within the preset time window, collecting optical power and optical signal-to-noise ratio data of the device, and synchronously collecting environmental radio frequency interference intensity data;

[0094] S35, quantifying the contribution ratio of the weak fluctuation according to the device data collected during the zero-point calibration and the environmental radio frequency interference intensity data;

[0095] S36, correlating the causal relationship according to the optical path, the risk label, the scheduling decision data, the service performance data and the contribution ratio.

[0096] Specifically, when the service performance is monitored to decline, first, the optical power data, the optical signal-to-noise ratio data and the radio frequency interference intensity data of the related device on the optical path within the performance decline period are acquired. These data are used for preliminary analysis of the device running state and the external interference. Among them, the optical power data reflects the intensity of the optical signal, the optical signal-to-noise ratio data reflects the quality of the optical signal, and the radio frequency interference intensity data is used to evaluate the potential influence of the external electromagnetic environment on the optical communication link.

[0097] Further, it is necessary to evaluate whether the device has the ability to perform zero-point calibration. Zero-point calibration refers to measuring the output of a device under no signal input or known stable input to determine the inherent bias or background noise of the device. If the device supports zero-point calibration, an instruction is issued to the device to switch to an idle state or a standby bypass within a preset time window. The idle state means that the device does not carry any service traffic, and the standby bypass means that the service traffic is temporarily switched to other paths, so that the target device can be tested independently. The preset time window is a pre-set time period to ensure the integrity of the calibration process and not to affect the normal service operation.

[0098] Within the preset time window, the optical power and optical signal-to-noise ratio data of the device are collected, and the environmental radio frequency interference intensity data are collected synchronously. The optical power and optical signal-to-noise ratio data collected at this time reflect the baseline performance and internal noise level of the device in the absence of service traffic or external optical signal input. The environmental radio frequency interference intensity data are collected synchronously to exclude or quantify the influence of environmental radio frequency interference on the measurement results of the device during calibration.

[0099] Therefore, according to the device data (i.e., baseline data) collected during the zero-point calibration and the environmental radio frequency interference intensity data, the contribution ratio of the weak fluctuation can be quantified. The contribution ratio represents the degree of signal quality decline actually caused by the weak fluctuation on the optical path after excluding the inherent bias of the device itself and the environmental radio frequency interference. For example, by taking the measurement value during calibration as a reference and comparing it with the measurement value during service operation, the real change caused by the weak fluctuation can be stripped.

[0100] Finally, the cause-effect association is associated according to the optical path, the risk label, the scheduling decision data, the service performance data, and the contribution ratio. By introducing the quantified contribution ratio, the cause-effect relationship between the weak fluctuation record and the service performance decline can be more accurately judged, and the performance problem caused by the weak fluctuation is avoided to be misjudged as the inherent error of the device or the environmental interference.

[0101] The scheme of the present application effectively solves the problem of insufficient precision in the traditional method when analyzing the cause-effect relationship between the weak fluctuation and the performance decline. Specifically, after monitoring the service performance decline, the optical power, optical signal-to-noise ratio, and radio frequency interference intensity data of the device on the optical path in the performance decline period are first obtained, which provides a basis for preliminary analysis. Subsequently, whether the device supports zero-point calibration is evaluated, and in the case of support, the device is switched to an idle state or a standby bypass, so that the baseline optical power and optical signal-to-noise ratio data of the device are collected within a preset time window, and the environmental radio frequency interference intensity data are collected synchronously. It is because these baseline data are collected in the absence of service or in a known stable state that the influence of the inherent bias of the device itself, internal noise, and environmental radio frequency interference can be stripped from the measurement data during service operation. In this way, the real contribution ratio of the weak fluctuation on the optical path to the service performance decline can be accurately quantified, so that non-fluctuation factors (such as device aging and environmental interference) are avoided to be misjudged as the cause of the weak fluctuation. The quantified contribution ratio makes the cause-effect association analysis more accurate and reliable.

[0102] By the technical solution, the accuracy of the causal correlation analysis between the weak fluctuation and the service performance decline in the optical communication network can be improved. Specifically, by the zero-point calibration and the synchronous collection of the environmental radio frequency interference data, the measurement error of the device itself and the external environmental interference can be effectively eliminated, the confusion of the causal judgment can be avoided, and the quantification of the real contribution of the weak fluctuation is more accurate. This avoids the misjudgment of the non-fluctuation factors as the reason for the performance decline, thereby reducing the false alarm and unnecessary resource scheduling. Thus, based on the more accurate causal correlation, the subsequent service path selection preference adjustment is more targeted and effective, the optical path with the real weak fluctuation can be more accurately avoided, and the overall stability and service quality of the optical communication network are improved, and the operation cost is reduced.

[0103] In some preferred embodiments, the following is described by a specific example. Assuming that an optical path in an optical communication network carries high-priority services, and the performance of the services is intermittently declined in recent days. Preliminary analysis shows that there are weak fluctuation records on the optical path which are lower than the preset alarm threshold. In order to accurately determine whether these weak fluctuations are the real reason for the performance decline of the services, the scheme of the present application is enabled.

[0104] Firstly, the system obtains the optical power, optical signal-to-noise ratio and radio frequency interference intensity data of the relevant optical transmission device on the optical path during the performance decline. Then, the system finds that the optical transmission device supports the zero-point calibration function. Therefore, the system issues an instruction to the device to switch the service traffic to the standby bypass in the preset time window (for example, 2:00 to 2:15 in the early morning) of the service low peak period, so that the device enters the idle state.

[0105] During the preset time window, the device automatically collects the optical power and optical signal-to-noise ratio baseline data when there is no service traffic, and the environmental sensor deployed near the device synchronously collects the environmental radio frequency interference intensity data in the time period. After the calibration is completed, the system corrects and analyzes the data collected during the performance decline of the services by using the baseline data and the environmental radio frequency interference data. For example, if the calibration data shows that the device itself has a 0.5 dBm optical power measurement error, and the environmental radio frequency interference causes a 0.2 dB signal-to-noise ratio decline in a specific frequency band, these factors will be deducted when quantifying the contribution proportion of the weak fluctuation.

[0106] In this way, the system can accurately calculate the proportion of performance degradation actually caused by the slight fluctuation in the optical path after excluding the inherent bias of the device and environmental interference. For example, if the correction shows that the slight fluctuation contributes up to 80% to the performance degradation, it can be concluded that the slight fluctuation is the main cause; if the contribution is low, further investigation of other factors may be needed. Ultimately, based on this more accurate contribution ratio, the system can more accurately associate the causal relationship between the slight fluctuation and the business performance degradation, and accordingly adjust the subsequent path selection preference of the business, for example, permanently or temporarily switch the business to another more stable optical path, thereby effectively avoiding resource waste or business interruption caused by misjudgment.

[0107] In some embodiments of the application described above, when the performance degradation of the business is monitored, the optical path passed by the business needs to be traced back for subsequent causal correlation analysis. However, in actual optical communication networks, business traffic may not always be transmitted along a single preset path, especially when the network is dynamically adjusted, load balanced or fault recovered, the business traffic may be switched between multiple optical paths. If the main optical path actually carrying the business traffic during performance degradation cannot be accurately identified, the accuracy of causal correlation analysis may be reduced, thereby affecting the effectiveness of subsequent scheduling decisions.

[0108] To this end, the application further proposes that the step of tracing back the optical path passed by the business when the performance degradation of the business is monitored, comprises:

[0109] When the performance degradation of the business is monitored, triggering a path tracing program according to the business identifier and the occurrence time point;

[0110] Obtaining allocation records of each path of the business, the allocation records including allocated paths, wavelength channels and accurate time stamps;

[0111] Obtaining path distribution information of actual transmission traffic of the business within the preset time window, the path distribution information including traffic proportion and time stamp of each path;

[0112] Comparing the allocation records with the path distribution information of the actual transmission traffic to identify switching time points and switched paths of the traffic between multiple optical paths within the preset time window;

[0113] According to the switching time points and the switched paths, constructing a time series distribution diagram of the traffic on each optical path within the preset time window;

[0114] According to the time series distribution diagram, in combination with the duration of the performance degradation of the service, an optical path carrying main traffic of the service is identified as an optical path actually passed by the service.

[0115] Specifically, when the performance of a service in an optical communication network is monitored to be degraded, the system immediately acquires the unique identifier of the service and the exact time point of the performance degradation. Based on this information, a special path backtracking program is triggered to accurately determine the optical path actually passed by the service during the performance degradation.

[0116] The path backtracking program first acquires the historical path allocation records of the service from the network management system. The allocation records detail the optical path to which the service is allocated at different time points, the wavelength channel used, and the corresponding exact time stamp. These records reflect the "planned" path of the service in the network.

[0117] Meanwhile, the system also acquires the path distribution information of the actual transmission traffic of the service within the preset time window. The path distribution information is obtained through real-time monitoring or historical data analysis, and contains the distribution proportion of service traffic on different optical paths and the corresponding collection time stamp, reflecting the "actual" transmission situation of service traffic.

[0118] Subsequently, by comparing the allocation records with the path distribution information of the actual transmission traffic, the system can identify the specific time points at which the traffic of the service switches between multiple optical paths within the preset time window and the optical paths used after the switch. This comparison helps to find the difference between the actual traffic and the allocation records, especially the path deviation that may occur during network dynamic adjustment or fault recovery.

[0119] Based on the identified switching time points and switched paths, the system will construct a time series distribution diagram of the traffic of the service on each optical path within the preset time window. The distribution diagram intuitively presents the dynamic distribution of service traffic on different optical paths, for example, the traffic of the service may be mainly concentrated on path A in a certain time period, and switched to path B in another time period.

[0120] Finally, according to the time series distribution diagram, in combination with the duration of the performance degradation of the service, the system will identify the optical path carrying the main traffic of the service during the performance degradation as the optical path actually passed by the service. This ensures that the optical path based on the subsequent causal correlation analysis is the most consistent with the actual traffic situation.

[0121] The scheme of the present application effectively solves the problem that the traditional method cannot accurately identify the actual transmission path of the service in a dynamic network environment by introducing a refined path backtracking mechanism. Specifically, when the service performance declines, the system not only considers the preset path allocation record, but more importantly, it obtains the path distribution information of the actual transmission traffic and compares it with the allocation record, thereby accurately identifying the switching behavior of the service traffic between multiple optical path paths. Thus, by constructing a time series distribution graph of the traffic on each optical path path and combining the duration of the performance decline, the scheme can identify the optical path path that mainly carries the traffic during the critical period. This method avoids inaccurate path identification due to path switching or traffic dispersion, and provides a more accurate and reliable basis for subsequent causal association analysis of weak fluctuations and performance decline.

[0122] Through the above technical scheme, the present application can overcome the limitation of inaccurate path backtracking caused by service traffic path switching or dispersion in a complex dynamic network environment. The present scheme accurately identifies the traffic switching point by comprehensively analyzing the allocation record and actual traffic distribution of the service, and constructs a time series distribution graph, thereby accurately determining the optical path path that actually carries the main traffic during the performance decline of the service. This significantly improves the accuracy and reliability of optical path path identification, provides a solid foundation for subsequent causal association analysis of weak fluctuations and performance decline of the service, and enables the intelligent scheduling system to make more accurate and effective path selection preference adjustments, ultimately improving the overall operation efficiency and service quality of the optical communication network.

[0123] In some preferred embodiments, the following is described by a specific example. Assume that in a certain optical communication network, a high-bandwidth video conference service is monitored to have serious video stuttering and packet loss from 10:30 to 10:45 in the morning, i.e., the service performance declines.

[0124] First, the system obtains the identification of the video conference service (e.g., service ID: VC_20231027_001) and the occurrence time point of the performance decline (10:30). Subsequently, the path backtracking program is triggered.

[0125] The program queries the historical record and finds that the service was allocated to optical path path A using wavelength channel 1550nm at 10:00 in the morning. However, through real-time traffic monitoring data, the system finds that the actual traffic distribution of the service within the preset time window of 10:00 to 10:45 is as follows:

[0126] - 10:00-10:15: 100% traffic is transmitted on optical path path A.

[0127] - 10:15-10:30: Due to network load balancing strategy, 70% traffic is transmitted on light path A, and 30% traffic is shunted to light path B.

[0128] - 10:30-10:45: Due to slight jitter of a device on light path A, the system automatically switches all traffic of the service to light path B.

[0129] By comparing the allocation records and the actual traffic distribution, the system identifies that traffic switching occurs at 10:15 and 10:30. Based on these switching points, the system constructs a time series distribution diagram of the traffic on light path A and light path B during 10:00-10:45 for the service.

[0130] In combination with the duration of performance degradation of the service (10:30-10:45), the system analyzes the time series distribution diagram. It is found that during the entire duration of performance degradation, light path B carries all or most of the traffic of the service. Therefore, the system finally identifies light path B as the main light path through which the video conference service passes during performance degradation. In this way, the subsequent causal correlation analysis will focus on the weak fluctuation records on light path B, so as to more accurately locate the root cause of the problem.

[0131] In some embodiments of the above application, by constructing a time series distribution diagram of service traffic on each light path and identifying the light path carrying the main traffic according to the distribution diagram, the light path passed by the service is traced back. However, in actual application, the cause of service performance degradation may not simply be caused by the light path carrying the main traffic, especially when multiple light paths carry service traffic and a certain path has weak fluctuations. Only by the judgment of the main traffic, the root cause of performance degradation may not be accurately identified. This may lead to misjudgment of the fault root cause, thereby affecting the accuracy of subsequent scheduling decisions.

[0132] In view of this, the application further provides the above method of identifying, according to the time series distribution diagram, in combination with the duration of performance degradation of the service, the light path carrying the main traffic of the service as the light path passed by the service, comprising:

[0133] During the performance degradation of the service, the traffic proportion of each light path in the time series distribution diagram is divided into fine-grained time windows, and the traffic proportion in each time window is counted;

[0134] All overlapping light paths in the duration of performance degradation of the service are identified, and for each overlapping light path, the corresponding weak fluctuation record is obtained;

[0135] According to the statistical result and the weak fluctuation record, a potential contribution value of each overlapping light path to the performance degradation of the service is calculated;

[0136] The light path with the highest potential contribution value is identified as the light path of the service.

[0137] Specifically, during the performance degradation of the service, the system performs fine-grained time window division on the traffic proportion of each light path in the time series distribution diagram. The fine-grained time window can be understood as further subdividing the entire performance degradation duration into smaller time periods, such as every second, every minute, or shorter time intervals, in order to more accurately capture the dynamic changes of traffic on different paths. Within each fine-grained time window, the traffic proportion carried by each light path is counted, thereby obtaining a detailed dynamic view of the traffic distribution.

[0138] Among them, identifying all overlapping light paths during the performance degradation duration of the service means that all light paths that have ever carried or are currently carrying the traffic of the service will be identified within the entire duration of the performance degradation of the service. For each identified overlapping light path, the system obtains its corresponding weak fluctuation record. These weak fluctuation records are weak fluctuations below the preset alarm threshold identified from the time-synchronized device operation data and have been risk-labeled.

[0139] In practical applications, according to the statistical result and the weak fluctuation record, a potential contribution value of each overlapping light path to the performance degradation of the service is calculated. The potential contribution value aims to quantify the possible influence degree of each path in the performance degradation event of the service, which considers the traffic proportion of the path during the performance degradation (statistical result) and the weak fluctuation existing on the path. For example, the higher the traffic proportion, the more frequent or more serious the weak fluctuation, the higher the potential contribution value.

[0140] Finally, the system identifies the overlapping light path with the highest potential contribution value as the light path of the service. This path is considered to be the most main or most direct light path that causes the performance degradation of the service.

[0141] The scheme of the present application solves the problem of possible misjudgment caused by only judging the main traffic by introducing fine-grained time window division, identifying overlapping light path paths, and combining with weak fluctuation records. Specifically, by performing fine-grained time window division on the traffic proportion during the service performance decline period, the dynamic distribution of traffic on different paths can be captured more finely, avoiding the information loss that may be caused by coarse-grained statistics. At the same time, all overlapping light path paths in the performance decline duration are identified to ensure that even the paths carrying non-main traffic but having potential problems can be included in the consideration range. More importantly, by comprehensively analyzing the traffic proportion statistical results of each overlapping light path path and its corresponding weak fluctuation record, and calculating the potential contribution value, the system can quantify the actual impact of each path on the performance decline. This quantification method can more accurately evaluate which paths have a stronger causal relationship between the weak fluctuation and the service performance decline, thereby avoiding the limitation of only focusing on traffic size and ignoring potential risks. Thus, the root light path path causing the service performance decline can be more accurately located.

[0142] Through the above technical scheme, the present application can significantly improve the accuracy of service performance decline fault positioning in optical communication networks. By carefully analyzing the traffic distribution and weak fluctuation records, even in multi-path transmission or complex fault scenarios, the light path path with the highest potential contribution to the service performance decline can be accurately identified. This not only avoids the possible misjudgment of traditional methods, but also provides a more reliable and fine basis for subsequent causal relationship analysis and path selection preference adjustment, thereby effectively improving the efficiency and accuracy of network resource intelligent scheduling, and ensuring the stable operation of services and user experience.

[0143] In some preferred embodiments, the following is described by a specific example. Assume that a video conference service has a performance decline phenomenon of video freezing and audio interruption during 10:00-10:05 am. According to the above method, the system has constructed a time series distribution diagram of the traffic of the service on light path path A, light path path B, and light path path C during 10:00-10:05. Among them, light path path A carries about 70% of the traffic, light path path B carries 20%, and light path path C carries 10%.

[0144] In some embodiments of the present application, in calculating the potential contribution of each overlapping path to the performance degradation of the service, the traffic proportion statistics and the weak fluctuation record are mainly used. However, in practical applications, different types of services have significant differences in sensitivity to network performance degradation. For example, real-time audio and video services are extremely sensitive to delay and packet loss rate, while offline data transmission services are relatively insensitive. If the type of service is not distinguished and the potential contribution value is simply calculated based on a unified weight, the analysis of the performance degradation of some key or sensitive services may not be accurate enough, thereby affecting the adjustment effect of the subsequent path selection preference, and even the quality of service of high-priority services may not be effectively guaranteed in a timely manner.

[0145] To this end, the present application further provides that the above-mentioned calculation of the potential contribution of each overlapping path to the performance degradation of the service according to the statistics and the weak fluctuation record comprises:

[0146] According to the type of the service, a sensitivity weight of the service is obtained;

[0147] Based on the sensitivity weight, the weights of the traffic proportion and the weak fluctuation record in the potential contribution value are dynamically adjusted;

[0148] The adjusted weights are applied to the traffic proportion and the weak fluctuation record to calculate the potential contribution of each overlapping path to the performance degradation of the service.

[0149] Specifically, the above-mentioned obtaining of the sensitivity weight of the service according to the type of the service refers to that the system queries or calculates the sensitivity of the service to network performance degradation according to the specific type of the service currently experiencing performance degradation, such as video conference, online game, VoIP call, data backup or web browsing, etc. The sensitivity weight can be a quantitative value reflecting the degree of influence of the service on service quality or user experience when facing specific performance degradation (such as delay increase, packet loss rate increase, bandwidth fluctuation, etc.). For example, the sensitivity weight of real-time interactive services is usually higher than that of non-real-time data transmission services.

[0150] Among them, the above-mentioned dynamic adjustment of the weights of the traffic proportion and the weak fluctuation record in the potential contribution value based on the sensitivity weight refers to that, in calculating the potential contribution value, a fixed weight factor is no longer used, but the influence of the traffic proportion and the weak fluctuation record in the final contribution value calculation is dynamically adjusted according to the obtained sensitivity weight of the service. For example, for high-sensitivity services, the weight of the weak fluctuation record may be increased to highlight its influence on performance degradation; and for low-sensitivity services, the weight of the traffic proportion may remain unchanged or be slightly adjusted. This dynamic adjustment ensures that the contribution value calculation can better reflect the characteristics of the service.

[0151] In practical applications, the application of the adjusted weight to the traffic proportion and the weak fluctuation record to calculate the potential contribution value of each overlapping optical path to the performance degradation of the service refers to multiplying the dynamically adjusted weight coefficient to the traffic proportion statistical value and the weak fluctuation record value of the corresponding optical path, respectively, and then comprehensively calculating the weighted values to obtain the final potential contribution value of each overlapping optical path to the current performance degradation of the service. The calculation result can more accurately reflect the actual influence degree of the weak fluctuation on the different optical paths on the performance degradation of the specific service.

[0152] The scheme of the present application effectively solves the limitation of the traditional method that fails to fully consider the difference in business types when evaluating the cause of service performance degradation by introducing a service sensitivity weight and dynamically adjusting the weight of the traffic proportion and the weak fluctuation record in the calculation of the potential contribution value. Specifically, when monitoring the performance degradation of the service, the sensitivity weight of the service type is first obtained, and the weight can quantify the tolerance of different services to network fluctuations. Then, the influence of the traffic proportion and the weak fluctuation record in the calculation of the potential contribution value is dynamically adjusted using the sensitivity weight. For example, for a service (such as real-time video) that is highly sensitive to delay and packet loss rate, even if the absolute value of the weak fluctuation record is not large, the weight of the weak fluctuation record in the potential contribution value will be correspondingly increased to highlight its potential influence on the service performance. Conversely, for a service (such as background data synchronization) that has low performance requirements, the weight of the weak fluctuation record may be relatively low. Thus, through the differentiated weight adjustment mechanism, the finally calculated potential contribution value can more accurately reflect the actual influence degree of the weak fluctuation on the specific optical path on the performance degradation of the current service, thereby avoiding the evaluation bias of all services.

[0153] Through the above technical scheme, the present application can realize more refined and personalized analysis of the cause of service performance degradation in optical communication networks. By introducing a service sensitivity weight and dynamically adjusting the calculation parameters, the system can fully consider the differentiated needs of different services for network quality when identifying the causal relationship between weak fluctuations and service performance degradation. This not only improves the accuracy of the calculation of the potential contribution value, making the root cause positioning more accurate, but also helps the scheduling system to more intelligently identify and prioritize the problems that have the greatest impact on high-sensitivity services, thereby optimizing the resource scheduling strategy, effectively ensuring the quality of service and user experience of critical services, and avoiding resource waste or service degradation caused by general evaluation models.

[0154] In some preferred embodiments, the following is described by a specific example. Assume that there are two services: service A is real-time video conference, and service B is large file download. When monitoring the performance degradation of service A and service B, the system will obtain their sensitivity weights, respectively.

[0155] For service A (real-time video conference), its sensitivity weight is set to a high value, e.g. 0.8, because it is very sensitive to delay and packet loss. At this time, in calculating the potential contribution value, the weight of the weak fluctuation record will be dynamically adjusted upwards based on 0.8, while the weight of the traffic proportion may be relatively adjusted downwards, to ensure that even a small weak fluctuation, as long as it occurs on the optical path carrying service A, its potential contribution value to the performance degradation of service A will be significantly amplified.

[0156] For service B (large file download), its sensitivity weight is set to a low value, e.g. 0.3, because it has a high tolerance to short delay and small packet loss. At this time, in calculating the potential contribution value, the weight of the weak fluctuation record will be dynamically adjusted downwards based on 0.3, while the weight of the traffic proportion may be relatively adjusted upwards, to reflect its dependence on continuous large traffic transmission.

[0157] For example, if there are the same weak fluctuation record and traffic proportion on a certain optical path, in calculating the potential contribution value to service A, due to its high sensitivity weight, the contribution value of this optical path will be calculated to be higher; while in calculating the potential contribution value to service B, due to its low sensitivity weight, the contribution value of this optical path will be relatively lower. Thus, the system can more accurately identify the optical path problem that has the greatest impact on real-time video conference, and prioritize processing or adjusting its path selection preference, thereby effectively improving the user experience of high sensitivity services.

[0158] In some embodiments of the above-mentioned embodiments of the present application, although the sensitivity weight of the service is obtained according to the type of the service, and the weights of the traffic proportion and the weak fluctuation record in the potential contribution value are dynamically adjusted based on the sensitivity weight, to calculate the potential contribution value of each overlapping optical path to the performance degradation of the service, for emerging services or customized services, their sensitivity characteristics may lack historical data or standard model support, resulting in inaccurate sensitivity weight acquisition, and further affecting the accuracy of causal association analysis.

[0159] In view of this, the present application further provides that the sensitivity weight of the service is obtained according to the type of the service, specifically comprising:

[0160] obtaining a service characteristic parameter of an emerging service or a customized service;

[0161] simulating the transmission of the emerging service or the customized service under different combinations of bit error rate, delay and packet loss rate in a controlled network environment, and recording the performance and user experience feedback of the service under each combination;

[0162] According to the service characteristic parameters, the performance and the user experience feedback, a sensitivity curve of the emerging service or the customized service is constructed;

[0163] According to the sensitivity curve, a sensitivity weight of the emerging service or the customized service is extracted.

[0164] Specifically, the service characteristic parameters of the emerging service or the customized service are obtained, which means that the key attributes defined in the early stage of design, deployment or operation of these services are collected. For example, for the emerging real-time interactive VR service, its characteristic parameters may include strict requirements on bandwidth, delay, jitter and packet loss rate, as well as its user experience sensitivity to visual fluency and interactive response speed. These parameters provide basic input for subsequent simulation and sensitivity curve construction.

[0165] In a controlled network environment, the transmission of the emerging service or the customized service under different combinations of error rate, delay and packet loss rate is simulated, and the performance and user experience feedback of the service under each combination are recorded, aiming to quantify the response of the service to network quality changes through systematic experiments. The controlled network environment can be a simulation platform or an isolated test network, in which network parameters such as error rate, delay and packet loss rate can be accurately controlled. Performance can include objective indicators such as throughput, success rate and response time, while user experience feedback can be obtained through questionnaire surveys, user behavior analysis or expert evaluation, such as user tolerance to video lag and voice interruption.

[0166] In practical applications, according to the service characteristic parameters, the performance and the user experience feedback, a sensitivity curve of the emerging service or the customized service is constructed, which aims to visualize and quantify the response of the service to network quality changes. The sensitivity curve can be a multi-dimensional function or chart, which depicts the trend of changes in business performance and user experience feedback under different network quality combinations. For example, a surface or curve cluster can be constructed with error rate, delay and packet loss rate as independent variables and business performance decline or user satisfaction as dependent variables.

[0167] Further, according to the sensitivity curve, a sensitivity weight of the emerging service or the customized service is extracted, which aims to simplify the complex sensitivity characteristics into a numerical value that can be used for weight adjustment. The sensitivity weight can be a single numerical value or a set of weight vectors, which reflects the sensitivity of the service to changes in a specific network parameter. For example, the weight can be calculated according to the slope, inflection point or performance decline amplitude at a specific threshold of the curve, to ensure that in the subsequent calculation of potential contribution value, the true sensitivity of the service to weak fluctuations can be accurately reflected.

[0168] The scheme of the present application solves the problem that the traditional method is difficult to accurately obtain the sensitivity weight when lacking historical data by establishing a set of sensitivity evaluation mechanism for emerging or customized services. Specifically, first, the characteristic parameters of the service are obtained to provide basic information for subsequent simulation; second, the transmission of the service under various network quality conditions is simulated in a controlled environment, and its performance and user experience feedback are recorded comprehensively, thereby obtaining the real response data of the service under different network conditions; then, a curve reflecting the sensitivity characteristics of the service is constructed based on these data, and the complex service sensitivity is quantified; finally, the accurate sensitivity weight is extracted from the sensitivity curve. Due to this data-driven and systematic sensitivity weight acquisition method, even for emerging or customized services lacking historical data, the sensitivity to network fluctuations can be accurately evaluated, thereby providing more reliable input for subsequent causal correlation analysis.

[0169] Through the above technical scheme, the accuracy and robustness of the optical communication network resource intelligent scheduling method in processing emerging or customized services can be significantly improved. Specifically, by finely obtaining the sensitivity weight of these special services, the potential contribution of slight fluctuations to the performance decline of the service can be more accurately evaluated, avoiding misjudgment or omission due to inaccurate sensitivity weight. Thus, in the causal correlation analysis, the root cause of the performance decline of the service can be more accurately identified, thereby making the subsequent path selection preference adjustment more targeted and effective, ultimately improving the resource scheduling efficiency and service quality of the entire optical communication network, especially in the face of evolving and diversified business demands.

[0170] In some preferred embodiments, the following is described by a specific example. Assume that there is an emerging "remote high-precision surgery" service, which is extremely sensitive to network delay and packet loss rate.

[0171] First, the service characteristic parameters of the remote high-precision surgery service are obtained, such as the requirement that the end-to-end delay is less than 5 milliseconds, the packet loss rate is less than 0.001%, and the synchronization of video stream and control signal is extremely high.

[0172] Next, in a controlled network environment, the transmission of the service under different combinations of bit error rate, delay, and packet loss rate is simulated. For example, multiple experiments are set up to simulate the change of delay from 1 millisecond to 50 milliseconds and packet loss rate from 0% to 1%, and the real-time feedback delay, video frame freezing frequency, control instruction loss rate, and other performance data of the surgery operation are recorded. At the same time, professional surgeons are invited to perform simulated operations, and their user experience feedback on operation fluency, feedback timeliness, and visual clarity is collected.

[0173] Then, according to these service characteristic parameters, performance data and user experience feedback, a sensitivity curve of the remote high-precision surgery service is constructed. The curve can intuitively show that when the delay exceeds 10 milliseconds or the packet loss rate exceeds 0.01%, the service performance and user experience will decrease sharply.

[0174] Finally, according to the constructed sensitivity curve, the sensitivity weight of the remote high-precision surgery service is extracted. For example, it can be identified that when the delay and packet loss rate reach a certain threshold, the slope of the service performance decrease is very steep, indicating that its sensitivity to these parameters is extremely high, so it is given a higher sensitivity weight. This weight is then used to adjust the traffic ratio and the weight of the weak fluctuation record in the potential contribution value, ensuring that the special sensitivity of the service to network weak fluctuations can be fully considered when analyzing the causal relationship of the service performance decrease.

[0175] In some embodiments of the above-mentioned application, a method for obtaining service sensitivity weight according to service type is proposed, in which the sensitivity curve of the service is constructed and the sensitivity weight is extracted therefrom. However, in actual application, simply extracting the weight from the sensitivity curve may not fully capture the sensitivity change characteristics of the service under different performance decreases, especially when the sensitivity curve presents nonlinearity or has multiple key turning points, which may lead to the extracted weight failing to accurately reflect the real sensitivity of the service to performance decrease, thereby affecting the adjustment accuracy of the subsequent path selection preference.

[0176] To this end, the application further proposes the above-mentioned method for extracting the sensitivity weight of emerging services or customized services according to the sensitivity curve, the steps of which include:

[0177] In the sensitivity curve, a performance decrease point corresponding to a certain performance decrease threshold is identified;

[0178] Starting from the performance decrease point, the sensitivity curve is expanded to both sides, and a local interval containing the performance decrease point and having relatively stable curve morphology is identified;

[0179] In the local interval, according to the slope and curvature change trend of the sensitivity curve, an inflection point or a platform period starting point of the curve is identified;

[0180] According to the inflection point or the platform period starting point, the local interval is divided into multiple subintervals;

[0181] For each subinterval, the corresponding average sensitivity or dominant sensitivity is calculated;

[0182] According to the average sensitivity or the dominant sensitivity of each subinterval and the position of the certain performance decrease threshold in the local interval, the sensitivity weight is weighted calculated.

[0183] Specifically, identifying the performance degradation point corresponding to a specific performance degradation threshold refers to determining a preset performance degradation degree on the constructed sensitivity curve (which generally reflects the relationship between service performance and user experience feedback), such as a certain critical value of bit error rate, a delay exceeding a certain upper limit, etc., and finding the point on the curve corresponding to the threshold. This point is usually considered as the starting point of the significant deterioration of service performance or the obvious damage to user experience.

[0184] Further, from the performance degradation point, expand to both sides along the sensitivity curve, and identify a local interval containing the performance degradation point and having relatively stable curve morphology. This local interval aims to focus on the area near the performance degradation point where the service sensitivity changes most critically. By expanding to both sides, the trend of service sensitivity change before and after the performance degradation point can be captured, and "relatively stable curve morphology" means that the sensitivity change in this interval has a certain regularity, which is convenient for subsequent analysis.

[0185] Among the local interval, according to the slope, curvature change trend of the sensitivity curve, the inflection point or the starting point of the platform period is identified. The inflection point represents the point where the sensitivity change rate changes significantly, such as from slow decline to rapid decline, or vice versa. The starting point of the platform period represents that the sensitivity change tends to be flat and enters a relatively stable stage. These points are crucial for understanding the internal mechanism of service sensitivity, as they mark the transition of service response mode to performance degradation. For example, these points can be identified by calculating the first derivative (slope) and second derivative (curvature) of the sensitivity curve.

[0186] Thus, according to the inflection point or the starting point of the platform period, the local interval is divided into multiple sub-intervals. Each sub-interval represents different sensitivity characteristics of the service within a specific performance degradation range. This division helps to analyze the sensitivity of the service more finely and avoids considering the entire local interval as a single sensitivity mode.

[0187] For each sub-interval, the corresponding average sensitivity or dominant sensitivity is calculated. The average sensitivity can be obtained by averaging the sensitivity values of all points in the sub-interval, while the dominant sensitivity can be determined by identifying the point with the highest sensitivity value or the most drastic change in the sub-interval. Both of these methods aim to provide a representative sensitivity quantitative value for each sub-interval.

[0188] Finally, the sensitivity weight is calculated by weighting the average sensitivity of each sub-interval or the dominant sensitivity, and the position of the specific performance degradation threshold in the local interval. This means that the sensitivity contribution of different sub-intervals will be weighted according to their importance, such as the distance from the performance degradation threshold, the length of the sub-interval, or the size of its sensitivity value, so as to obtain a comprehensive and accurate weight value that reflects the overall sensitivity of the service.

[0189] The scheme of the present application overcomes the limitations that may exist in simple extraction of weight by fine analysis of the sensitivity curve. Specifically, first, the performance degradation point corresponding to the specific performance degradation threshold is identified, so that the analysis can focus on the key area where the service performance begins to be significantly affected. Then, by extending the identification of the local interval, the overall consideration of the sensitivity change before and after the performance degradation point is ensured. In the local interval, by analyzing the slope and curvature change trend of the sensitivity curve to identify the inflection point or the starting point of the plateau period, the transition of the service sensitivity mode can be accurately captured, such as from linear response to nonlinear response, or from rapid deterioration to stable deterioration. These inflection points and starting points of the plateau period serve as key division points to divide the local interval into multiple sub-intervals, so that each sub-interval can represent the unique sensitivity characteristics of the service in a specific performance degradation range. By calculating the average sensitivity or the dominant sensitivity of each sub-interval and combining the position of the specific performance degradation threshold in the local interval for weighted calculation, the scheme can generate a more fine, accurate and representative sensitivity weight. This weighted calculation ensures that the area with the greatest impact on service performance is given a higher weight, so that the final sensitivity weight can more truly reflect the sensitivity of the service to performance degradation.

[0190] Through the above technical scheme, the present application can provide a more accurate and detailed sensitivity weight extraction method. Compared with simply extracting weight from the sensitivity curve, the present application identifies the key performance degradation point, the local interval, the inflection point and the starting point of the plateau period, and performs sub-interval division and weighted calculation, which significantly improves the accuracy and representativeness of the sensitivity weight. This fine analysis makes the obtained sensitivity weight more truly reflect the sensitivity characteristics of emerging services or customized services under different performance degradation levels, especially in the key areas where performance deteriorates rapidly or tends to be stable. Therefore, in the subsequent path selection preference adjustment, decisions can be made based on more accurate sensitivity weights, so as to realize more intelligent and efficient optical communication network resource scheduling, effectively avoid improper resource allocation or further deterioration of service performance due to inaccurate sensitivity weight, and ultimately improve user experience and network service quality.

[0191] Specifically, the above-mentioned identification of the inflection point or the starting point of the plateau period of the curve in the local interval according to the slope and curvature change trend of the sensitivity curve comprises:

[0192] performing a moving average processing on the sensitivity curve within the local interval;

[0193] calculating a first-order difference and a second-order difference of the sensitivity curve after the moving average processing;

[0194] identifying a point with the maximum slope change rate or a point with a continuous change in the sign of the second-order difference as the inflection point according to the continuous change trend of the first-order difference and the second-order difference;

[0195] identifying a region with a continuous stability of the first-order difference within a preset small range and a duration exceeding a preset threshold as the starting point of the plateau period.

[0196] In the local interval, the moving average processing is performed on the sensitivity curve, aiming to smooth the curve data and reduce the influence of random noise on subsequent difference calculation, thereby improving the accuracy and stability of the identification of the inflection point and the starting point of the plateau period. The moving average processing can be achieved by selecting a suitable sliding window size, for example, a moving average filter can be used to perform a weighted average on the data points on the sensitivity curve.

[0197] Further, the first-order difference and the second-order difference of the sensitivity curve after the moving average processing are calculated, which is a common method for mathematically analyzing the change trend of the curve. The first-order difference reflects the slope of the curve, i.e., the change rate; the second-order difference reflects the change rate of the slope, i.e., the curvature of the curve. These difference data provide a quantitative basis for accurately identifying the features of the curve. For example, the first-order difference can be approximately calculated by the difference between adjacent data points, and the second-order difference can be approximately calculated by the difference between the first-order differences.

[0198] Specifically, according to the continuous change trend of the first-order difference and the second-order difference, a point with the maximum slope change rate or a point with a continuous change in the sign of the second-order difference is identified as the inflection point. The inflection point is usually the place where the slope of the curve changes most sharply, which is manifested as a change in the sign of the second-order difference or a local extreme value of the first-order difference. By monitoring these mathematical features, the inflection point can be accurately located.

[0199] In addition, a region with a continuous stability of the first-order difference within a preset small range and a duration exceeding a preset threshold is identified as the starting point of the plateau period. The plateau period represents a stage where the business performance is no longer sensitive to the change of a certain parameter or the sensitivity change trend tends to be flat. On the first-order difference, this is manifested as a slope close to zero and stable. By setting a preset small range (for example, a threshold close to zero) and a preset duration threshold, the starting point of the plateau period can be effectively identified.

[0200] The scheme of the present application effectively suppresses the noise in the sensitivity curve by introducing a moving average process, providing a more smooth and reliable data basis for subsequent mathematical analysis. It is due to the data smoothing process that the calculation of the first-order difference and the second-order difference can more accurately reflect the inherent trend of the curve, rather than being disturbed by random fluctuations. Through accurate calculation of the first-order difference (slope) and the second-order difference (curvature) and analysis of the continuous change trend, the inflection point of the curve, i.e. the point with the maximum slope change rate or the point where the curvature sign changes, can be mathematically defined and identified. These points usually correspond to the key transition of the performance response mode in business sensitivity analysis. At the same time, by monitoring the region where the first-order difference is stable within a predetermined small range, the present application can accurately locate the starting point of the plateau, which is crucial to understanding when the business performance enters the saturation or insensitive state. Thus, the scheme overcomes the limitations of traditional methods in noisy environments, such as inaccurate identification or reliance on subjective judgment, and realizes the automation and high-precision identification of key feature points of the sensitivity curve.

[0201] Through the above technical scheme, the present application can significantly improve the identification accuracy and robustness of the inflection point and the starting point of the plateau in the sensitivity curve. Compared with the method of relying only on visual observation or simple trend judgment, the present application uses mathematical tools for quantitative analysis of the curve, effectively filtering out the influence of data noise, and ensuring the objectivity and consistency of the identification results. This precise identification capability makes the subsequent division of local intervals more reasonable, so that the sensitivity weight of the business can be calculated more accurately, providing a more reliable decision basis for intelligent scheduling of optical communication network resources. Ultimately, this helps to optimize the configuration of network resources, improve business performance, and improve user experience.

[0202] In some preferred embodiments, assuming that when constructing the sensitivity curve of a new video service, the curve shows a trend of rapid decline followed by flattening in a certain local interval, but there is certain measurement noise in the data. In order to accurately identify the inflection point of performance decline and the starting point of the plateau entering the plateau, the method proposed by the present application can be used.

[0203] First, the sensitivity curve data of the local interval is subjected to a moving average process, for example, a 5-point moving average filter is used to eliminate high-frequency noise.

[0204] Next, based on the smoothed data, the first-order difference and the second-order difference are calculated. By analyzing the change rate of the first-order difference, it can be found that it reaches a maximum value at a certain point, and the sign of the second-order difference changes near this point, so that the point is accurately identified as the inflection point of performance decline.

[0205] Subsequently, the first-order difference is continuously observed, and when the value thereof remains stable within a preset small range, for example, -0.01 to 0.01, and the duration of the stable state is longer than, for example, 10 data points, the starting point of the region is identified as the starting point of the plateau period. In this way, the key turning points and stable periods of the video service in the performance decline process can be automatically and accurately identified, and fine sensitivity weights can be provided for subsequent adjustment of the resource scheduling strategy.

[0206] In some embodiments of the present application, a method for identifying the inflection point or the starting point of the plateau period according to the slope and curvature change trend of the sensitivity curve in a local interval is proposed. However, in actual applications, due to the noise or local small fluctuations in network data, multiple similar data points may simultaneously meet the identification conditions of the inflection point or the starting point of the plateau period, thereby introducing the uncertainty of the identification result and affecting the accuracy of the subsequent sensitivity weight calculation. In this regard, the present application further proposes an optimization scheme to improve the accuracy and robustness of the identification of the inflection point or the starting point of the plateau period, and to ensure that the identified key points can more accurately reflect the essential changes of the service sensitivity curve.

[0207] After identifying the region in which the first-order difference remains stable within a preset small range and the duration is longer than a preset threshold as the starting point of the plateau period, the method further includes: performing consistency verification on the identified inflection point or the starting point of the plateau period, and when multiple similar points meet the identification conditions, selecting the point with the largest change amplitude as the final inflection point or the starting point of the plateau period.

[0208] Specifically, the consistency verification refers to further evaluation and screening of all points preliminarily identified as meeting the conditions of the inflection point or the starting point of the plateau period. The purpose is to exclude the misidentification caused by data noise or local small fluctuations and to ensure the reliability of the selected points. Among them, the multiple similar points can be understood as multiple points within a certain local interval of the sensitivity curve, which meet the identification standards of the inflection point or the starting point of the plateau period within a preset distance threshold or time window.

[0209] For example, in a certain sharp change region of the curve, there can be multiple adjacent data points whose first or second order difference reaches or approaches the identification threshold. In practical applications, the point with the largest change amplitude refers to the point among these similar candidate points whose slope change rate (for inflection points) or sensitivity value change (for plateau starting points) of the corresponding sensitivity curve is the most significant. For example, for an inflection point, the absolute value of its second order difference can be calculated, and the point with the largest absolute value is selected; for a plateau starting point, the change amount of its sensitivity values before and after can be investigated, and the point with the largest change amount is selected. The purpose is to identify the key turning point that has the most significant impact on business performance. Through the above verification and selection process, a most representative and influential point is finally determined as the inflection point or plateau starting point of the local interval, which is used for subsequent sensitivity weight calculation.

[0210] The scheme of the present application effectively solves the problem that multiple similar points can simultaneously meet the identification conditions of inflection points or plateau starting points in a complex data environment by introducing a consistency verification mechanism. Specifically, when multiple potential inflection points or plateau starting points are initially identified, the verification mechanism further evaluates the "importance" or "significance" of these points. By selecting the point with the largest change amplitude as the final inflection point or plateau starting point, it is ensured that the identified key point is the turning point that has the most significant impact on the change of business sensitivity in the local interval. Thus, false positives caused by noise or minor fluctuations are avoided, and the subsequent division of the sensitivity curve and the calculation of the sensitivity weight can be based on more accurate and representative key points.

[0211] Through the above technical scheme, the present application can significantly improve the accuracy and robustness of inflection point or plateau starting point identification. Specifically, by performing consistency verification on the identified candidate points and selecting the point with the largest change amplitude, the interference of data noise and local minor fluctuations on the identification result is effectively eliminated, ensuring that the identified key point can more accurately reflect the essential characteristics of the business sensitivity curve. This makes the subsequent sensitivity curve division more reasonable, so that more accurate sensitivity weights can be extracted, providing more reliable decision basis for intelligent scheduling of optical communication network resources, further optimizing the accuracy of resource allocation and business path selection.

[0212] In some preferred embodiments, it is assumed that when analyzing the sensitivity curve of a new business, three adjacent data points A, B and C in a local interval are preliminarily identified to meet the identification conditions of inflection points by calculating the first-order difference and the second-order difference. The absolute value of the second-order difference of point A is 0.05, the absolute value of the second-order difference of point B is 0.12, and the absolute value of the second-order difference of point C is 0.06. According to the scheme of the present application, consistency verification will be performed on the three points. Since the absolute value of the second-order difference of point B is the largest (0.12), it indicates that the corresponding curve slope changes most sharply, and therefore, point B will be selected as the final inflection point of the local interval. In this way, even if there are multiple potential inflection points, the most representative key turning point can be accurately identified, and the sensitivity weight calculation deviation caused by selecting a secondary point can be avoided.

[0213] Reference Figure 3 , Figure 3 is a schematic diagram of an optical communication network resource intelligent scheduling system provided by an embodiment of the present application, comprising:

[0214] an input end for acquiring device running data, scheduling decision data and service performance data and performing time synchronization; identifying weak fluctuations below a preset alarm threshold and performing risk marking according to the device running data after time synchronization;

[0215] an analysis end for acquiring the service identifier and the time point of performance degradation when monitoring the performance degradation of the service, backtracking the optical path passed by the service, and associating and analyzing the causal relationship between the weak fluctuation record and the performance degradation according to the optical path, the risk marking, the scheduling decision data and the service performance data;

[0216] an adjustment end for adjusting the subsequent path selection preference of the service according to the causal relationship.

[0217] The conventional existing optical communication network intelligent scheduling system mainly relies on the real-time running state data reported by the device when performing resource allocation. However, these data may have weak and intermittent physical layer performance fluctuations. These fluctuations are often below the preset alarm threshold due to their small amplitude and short duration, which leads to the fact that the scheduling system cannot perceive and quantify their potential impact on service transmission quality, so that high-value services may be allocated to paths with potential risks, ultimately leading to service performance degradation and difficulty in diagnosis.

[0218] To this end, the present application proposes an optical communication network resource intelligent scheduling system. Through its input end, device operation data, scheduling decision data and service performance data are obtained and time synchronization is performed. According to the time-synchronized device operation data, weak fluctuations below the preset alarm threshold are identified and risk labels are added. When the service performance is monitored to decline, the analysis end can obtain the service identifier and the time point of performance decline, backtrack the optical path path passed by the service, and comprehensively analyze the optical path path, risk label, scheduling decision data and service performance data to associate and analyze the causal relationship between the weak fluctuation record and the performance decline. Thus, the adjustment end can adjust the subsequent path selection preference of the service according to the causal association, thereby effectively avoiding potential risks and improving the stability and reliability of service transmission. Through the modular design, the data acquisition and preliminary processing, the core analysis and causal association, and the final scheduling strategy adjustment function are clearly divided, ensuring the professionalism and synergy of each functional module, and realizing the intelligent and risk-avoiding scheduling of optical communication network resources.

[0219] In order to better understand the optical communication network resource intelligent scheduling system proposed in the present application, some key components involved therein are described in detail below.

[0220] The input end is configured to obtain device operation data, scheduling decision data and service performance data and perform time synchronization. Specifically, the input end can be a data acquisition module that periodically acquires various device operation data such as optical power, optical signal-to-noise ratio (OSNR), bit error rate (BER), etc. through interface communication with the network management system (NMS), performance monitoring agent or directly with the optical network equipment. At the same time, the input end can also read scheduling decision data from the database of the scheduling system and obtain service performance data from the service performance probe or service server. In order to ensure the accuracy of data analysis, the input end is also responsible for time synchronization of all the obtained data, such as calibrating the time stamps of different data sources through high-precision time protocol (such as NTP or PTP).

[0221] Further, the input end is also configured to identify weak fluctuations below the preset alarm threshold and add risk labels according to the time-synchronized device operation data. For example, the input end can include a data preprocessing unit that performs sliding window average processing on the collected optical power data and calculates the deviation between each data point and the sliding average. When the deviation continuously exceeds a small preset threshold (e.g. 0.2 dB) but is below the regular alarm threshold (e.g. 0.5 dB), the unit identifies it as a weak fluctuation. For the identified weak fluctuations, the input end can assign different risk labels such as "high risk" or "low risk" according to factors such as their duration, amplitude, occurrence frequency and device type involved, and store these weak fluctuation records with risk labels for subsequent analysis.

[0222] Analysis end: configured to obtain the performance degradation of the business identity and the time point when the performance of the business is monitored to decline, trace back the optical path path passed by the business, and analyze the cause-effect relationship between the weak fluctuation record and the performance decline according to the optical path path, the risk mark, the scheduling decision data and the business performance data. Specifically, the analysis end can include a performance monitoring module, which continuously analyzes the business performance data. For example, when the bit error rate of a certain business significantly increases in a short time or the delay exceeds the preset service level agreement (SLA) threshold, the module can determine that the performance has declined, and record the unique identity of the business and the accurate time point of the performance decline.

[0223] Further, the analysis end also includes a path backtracking module, which identifies the optical path path passed by the business during the performance decline by querying the historical path allocation record of the scheduling system. For example, the module can query the specific optical path path to which the business is allocated in different time periods according to the business identity and the time point. After determining the optical path path, the core function of the analysis end is its cause-effect correlation analysis module. The module will combine the weak fluctuation record (with risk mark) of all devices on the path, the historical scheduling decision data and the performance data of the business itself for comprehensive correlation analysis. For example, if there is a weak fluctuation marked as “high risk” on the path passed by the performance-declined business at a certain time point, and the fluctuation is highly coincident with the time point of the business performance decline, the module can preliminarily determine that there is a cause-effect relationship between the two. The module can use statistical analysis, machine learning model or expert system rule to identify the cause-effect correlation.

[0224] Adjustment end: configured to adjust the path selection preference of the business in the future according to the cause-effect correlation. Once the analysis end determines the cause-effect relationship between the weak fluctuation and the performance decline of the business, the adjustment end will use this information to optimize future scheduling decisions. Specifically, the adjustment end can include a strategy updating module, which dynamically updates the path selection preference in the scheduling algorithm according to the results of the cause-effect correlation analysis. For example, if it is found that a weak fluctuation on a ROADM device leads to the performance decline of a specific business, then when selecting a path for the business or similar sensitive businesses in the future, the strategy updating module will reduce the priority of the path containing the ROADM device, or temporarily list it in the “blacklist”. This can be achieved by introducing a “risk weight” or “cost factor” into the scheduling algorithm, so that the scheduling algorithm tends to select other more stable paths when optimizing the objective function. This adjustment can be dynamic, continuously updated as the network state changes and new weak fluctuation records appear, ensuring that the scheduling system can continuously learn and adapt to changes in the network environment.

[0225] The optical communication network resource intelligent scheduling system provided in the present application effectively solves the problem that the traditional scheduling system cannot perceive and avoid hidden network risks by introducing the identification of weak fluctuations below the preset alarm threshold, risk marking, and scheduling adjustment mechanism based on causal association.

[0226] Compared with the prior art, the system of the present application has significant progress. The traditional optical communication network scheduling system usually relies on explicit alarms and conventional performance indicators reported by devices for resource allocation, and its system architecture often lacks a special module to process and analyze those weak fluctuations below the alarm threshold. When these hidden risks occur in the network, the traditional system cannot identify them, resulting in scheduling decisions based on incomplete or biased data, which may allocate high-value services to paths with potential performance risks. This not only leads to a decline in service performance, but also makes fault location and resolution extremely difficult, seriously affecting the quality of service and operation and maintenance efficiency of the network.

[0227] The system of the present application realizes fine management of weak fluctuations and causal association analysis through its clearly divided input end, analysis end and adjustment end. The input end is responsible for comprehensive and detailed data collection and identification of weak fluctuations, providing accurate risk information for subsequent analysis. The analysis end focuses on complex causal association analysis and can accurately locate the root cause of the decline in service performance. The adjustment end dynamically and intelligently adjusts the scheduling strategy based on the analysis results to actively avoid risky paths. The innovation of this system architecture lies in its ability to perceive and quantify the "physical reality" embedded in the scheduling decision-making process, enabling the system to fundamentally avoid allocating services to paths with potential risks. In this way, the system of the present application can significantly improve the utilization efficiency of network resources and the stability of service transmission, reduce the complexity of manual intervention and fault troubleshooting, and thus provide users with higher quality and more reliable optical communication services.

[0228] The above only describes the embodiments of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for intelligent scheduling of resources in an optical communication network, characterized in that, The method comprises: acquiring device operation data, scheduling decision data, and service performance data and performing time synchronization; identifying weak fluctuations below a preset alarm threshold and performing risk marking according to the time-synchronized device operation data; when a performance decline of a service is monitored, acquiring a service identifier and a time point of the performance decline, backtracking an optical path of the service, and correlatively analyzing a causal relationship between weak fluctuation records and the performance decline according to the optical path, the risk marking, the scheduling decision data, and the service performance data; adjusting path selection preferences of the service according to the causal relationship.

2. The method of claim 1, wherein, The correlatively analyzing the causal relationship between the weak fluctuation records and the performance decline according to the optical path, the risk marking, the scheduling decision data, and the service performance data comprises: acquiring optical power and optical signal-to-noise ratio data and radio frequency interference intensity data of a device on the optical path within a time period of the performance decline; evaluating whether the device performs zero-point calibration; if yes, issuing an instruction to the device to switch to an idle state or a standby bypass within a preset time window; within the preset time window, collecting the optical power and the optical signal-to-noise ratio data of the device, and synchronously collecting the radio frequency interference intensity data; quantifying a contribution proportion of the weak fluctuation according to the device data collected within the zero-point calibration and the radio frequency interference intensity data; correlatively analyzing the causal relationship according to the optical path, the risk marking, the scheduling decision data, the service performance data, and the contribution proportion.

3. The method of claim 2, wherein, The acquiring a service identifier and a time point of a performance decline of a service when a performance decline of the service is monitored and backtracking an optical path of the service comprises: when the performance decline of the service is monitored, triggering a path backtracking program according to the service identifier and the time point; acquiring an allocation record of the service, the allocation record comprising an allocated path, a wavelength channel, and an accurate time stamp; acquiring path distribution information of actual transmission traffic of the service within a preset time window, the path distribution information comprising a traffic proportion and a time stamp of each path; comparing the allocation record with the path distribution information of the actual transmission traffic to identify a switching time point and a switched path of traffic between multiple optical paths of the service within the preset time window; constructing a time series distribution diagram of the traffic on each optical path of the service within the preset time window according to the switching time point and the switched path; according to the time series distribution diagram and a duration of the performance decline of the service, taking an optical path carrying main traffic of the service as the optical path of the service.

4. The method of claim 3, wherein, The according to the time series distribution diagram and the duration of the performance decline of the service, taking an optical path carrying main traffic of the service as the optical path of the service comprises: during the performance decline of the service, performing fine-grained time window division on the traffic proportion of each optical path in the time series distribution diagram and performing statistics on the traffic proportion in each time window. Identify all optical path paths overlapping in the performance degradation duration of the service, for each overlapping optical path path, obtain the corresponding weak fluctuation record; According to the statistical results and the weak fluctuation record, calculate the potential contribution value of each overlapping optical path path to the performance degradation of the service; Identify the highest potential contribution value of the overlapping optical path path as the optical path path.

5. The method of claim 4, wherein, According to the statistical results and the weak fluctuation record, calculating the potential contribution value of each overlapping optical path path to the performance degradation of the service, comprising: According to the type of the service, obtain the sensitivity weight of the service; Based on the sensitivity weight, dynamically adjust the weight of the traffic ratio and the weak fluctuation record in the potential contribution value; Apply the adjusted weight to the traffic ratio and the weak fluctuation record to calculate the potential contribution value of each overlapping optical path path to the performance degradation of the service.

6. The method of claim 5, wherein, According to the type of the service, obtain the sensitivity weight of the service, comprising: Obtain the service characteristic parameters of emerging services or customized services; In a controlled network environment, simulate the transmission of the emerging services or the customized services under different combinations of error rates, delays and packet loss rates, and record the performance and user experience feedback of the services under each combination; According to the service characteristic parameters, the performance and the user experience feedback, construct the sensitivity curve of the emerging services or the customized services; According to the sensitivity curve, extract the sensitivity weight of the emerging services or the customized services.

7. The method of claim 6, wherein, According to the sensitivity curve, extract the sensitivity weight of the emerging services or the customized services, comprising: In the sensitivity curve, identify the performance degradation point corresponding to a specific performance degradation threshold; From the performance degradation point, expand along the sensitivity curve to both sides to identify a local interval containing the performance degradation point and having relatively stable curve form; In the local interval, according to the slope and curvature change trend of the sensitivity curve, identify the inflection point or the starting point of the platform period of the curve; According to the inflection point or the starting point of the platform period, divide the local interval into multiple subintervals; For each subinterval, calculate the corresponding average sensitivity or dominant sensitivity; According to the average sensitivity or the dominant sensitivity of the subinterval and the position of the specific performance degradation threshold in the local interval, calculate the sensitivity weight by weighting.

8. The method of claim 7, wherein, In the local interval, according to the slope and curvature change trend of the sensitivity curve, identify the inflection point or the starting point of the platform period of the curve, comprising: In the local interval, perform sliding average processing on the sensitivity curve; Calculate the first order difference and the second order difference of the sensitivity curve after the sliding average processing; According to the continuous change trend of the first order difference and the second order difference, identify the point with the maximum slope change rate or the point with the continuous change of the second order difference sign as the inflection point; Identify the region with continuous stability of the first order difference in a preset small range and the duration exceeding a preset threshold as the starting point of the platform period.

9. The method of claim 8, wherein, After the identification of the region where the first-order difference is continuously stable within a preset small range and the duration exceeds a preset threshold as the starting point of the plateau, the method further comprises: Consistency verification is performed on the identified inflection point or the starting point of the plateau, and when multiple similar points meet the identification condition, the point with the largest change amplitude is selected as the final inflection point or the starting point of the plateau.

10. An optical communication network resource intelligent scheduling system, characterized in that, The method comprises: an input end configured to acquire device operation data, scheduling decision data, and service performance data and perform time synchronization; According to the device operation data synchronized in time, the weak fluctuations below the preset alarm threshold are identified and marked as risks; When the performance of a service is found to be degraded, the service identifier and the time point of the performance degradation are acquired, the optical path of the service is traced back, and the causal relationship between the weak fluctuations and the performance degradation is analyzed according to the optical path, the risk mark, the scheduling decision data, and the service performance data; An adjustment end is configured to adjust the path selection preference of the service according to the causal relationship.

Images