A priority scheduling method for protecting packets in an spn tunnel

Abstract

The application relates to the technical field of communication network scheduling, and discloses a priority scheduling method for protection messages in an SPN tunnel. The method acquires a protection message type set of the SPN tunnel, extracts message scheduling parameters containing permanent priority parameters and temporary priority parameters from the protection message type set, generates priority trigger signals of the protection messages based on the parameters, identifies nodes triggered by permanent priority parameter updating and nodes triggered by temporary priority parameter resetting, classifies and calculates permanent priority offsets and temporary priority offsets based on the nodes, generates comprehensive priority offset parameters, separates general scheduling parameters that can be multiplexed across message types and special scheduling parameters that are only applicable to specific message types according to the parameters, and dynamically adjusts a protection message forwarding queue. The method is suitable for transmission requirements of different protection messages through a refined parameter division and a dynamic adjustment mechanism, and optimizes forwarding queue management.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication network scheduling, in particular to a priority scheduling method for protection packets in an SPN tunnel. BACKGROUND

[0002] As a new type of bearing technology for 5G and future networks, SPN combines the advantages of packet switching and time slot switching, and has unique value in supporting large bandwidth and low latency services. It has become an important support for building new generation information infrastructure. With the rapid development of emerging businesses such as industrial internet, remote medical care and autonomous driving, the transmission quality of protection packets in the network is directly related to the continuity and reliability of the business.

[0003] In the SPN tunnel transmission scenario, protection packets bear key functions such as fault switching and state monitoring, and their forwarding efficiency and order are crucial to network self-healing capability. However, the existing scheduling mechanism has obvious shortcomings in handling protection packets. Most methods use static priority division, which cannot dynamically adjust according to the real-time state of the network. When the tunnel traffic suddenly increases, high-priority protection packets may be delayed due to insufficient resource preemption, and low-priority packets may be accumulated for a long time. Traditional mechanisms do not distinguish the types of protection packets well, and mix packets with different functions and different time efficiency requirements, which leads to key protection instructions and ordinary state packets competing for resources, affecting the network fault response speed.

[0004] The existing technology lacks hierarchical consideration in priority parameter design, either relying only on fixed service levels or adjusting only according to temporary congestion, making it difficult to balance long-term stability and short-term flexibility. For example, some mechanisms only assign priority based on the source address or port number of the packet, ignoring the actual role of the packet in a specific business process. Some dynamic scheduling methods adjust parameters too frequently, causing queue oscillation and reducing forwarding efficiency. These problems make it difficult for SPN tunnels to fully realize their technical advantages when facing complex business scenarios, and the transmission reliability of protection packets cannot be effectively guaranteed. SUMMARY

[0005] The purpose of the present application is to provide a priority scheduling method for protection packets in an SPN tunnel to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides a priority scheduling method for protection packets in an SPN tunnel, which comprises:

[0007] S1: Obtain a set of protection packet types for the SPN tunnel, and extract packet scheduling parameters based on the set of protection packet types; the packet scheduling parameters include permanent priority parameters and temporary priority parameters;

[0008] S2: generating a priority trigger signal of the protection packet based on the packet scheduling parameter; identifying a priority change node according to the priority trigger signal; the priority change node includes a node triggered by a permanent priority parameter update and a node triggered by a temporary priority parameter reset;

[0009] S3: based on the priority change node, classifying and calculating a permanent priority offset and a temporary priority offset, and generating a comprehensive priority offset parameter;

[0010] S4: based on the comprehensive priority offset parameter, separating general scheduling parameters that can be multiplexed across packet types and special scheduling parameters that are only applicable to specific packet types, and dynamically adjusting a protection packet forwarding queue according to the separation result;

[0011] The specific packet type is a packet type whose integrity and accuracy requirements for transmission are greater than a preset threshold.

[0012] Preferably, the set of protection packet types of the SPN tunnel includes:

[0013] Based on the packet transmission delay requirement, the permanent priority parameter is divided to generate an initial priority parameter set and a corresponding effective period;

[0014] Based on the burstiness requirement of the service, the temporary priority parameter is divided to generate a temporary parameter set and a corresponding effective period;

[0015] Based on the initial priority parameter set and the corresponding effective period, a corrected temporary parameter set is obtained through a parameter verification rule;

[0016] The initial priority parameter set and the corrected temporary parameter set are combined into a packet scheduling parameter.

[0017] Preferably, the corrected temporary parameter set obtained through the parameter verification rule includes:

[0018] Calculate the priority deviation of each packet transmission stage, which is determined by the time sequence relationship of the permanent priority parameter weight coefficient, the parameter change timestamp and the initial parameter setting timestamp;

[0019] Based on the proportion of the priority deviation in the total deviation, the parameter values of each stage in the temporary parameter set are weighted and corrected.

[0020] Preferably, the priority trigger signal of the protection packet includes:

[0021] A first trigger signal is generated according to the initial priority parameter set;

[0022] A second trigger signal is generated according to the corrected temporary parameter set;

[0023] calculating a priority change trend strength based on the timing correlation between the first trigger signal and the second trigger signal;

[0024] when the priority change trend strength exceeds a set threshold, generating a priority trigger signal in advance or delaying the generation of the priority trigger signal.

[0025] Preferably, the identification of the priority change node comprises:

[0026] calculating the trend correlation degree of the priority trigger signal and the historical permanent priority change node, and marking the node as a permanent priority update node when the trend correlation degree is greater than a set correlation threshold;

[0027] calculating the trend deviation degree of the priority trigger signal and the historical temporary priority reset node, and marking the node as a temporary priority reset node when the trend deviation degree is greater than a set deviation threshold.

[0028] Preferably, the generation of the comprehensive priority offset parameter comprises:

[0029] extracting the permanent priority update node, calculating the offset amount of the permanent priority parameter in the adjacent service period, and recording it as a first offset component;

[0030] extracting the temporary priority reset node, calculating the temporary priority offset amount at the service type switching, and recording it as a second offset component;

[0031] linearly combining the first offset component and the second offset component as the comprehensive priority offset parameter.

[0032] Preferably, the separation of the general scheduling parameter applicable across message types and the special scheduling parameter applicable only to a specific message type comprises:

[0033] analyzing the common characteristics in the comprehensive priority offset parameter based on the message transmission delay and the packet loss rate as the reference dimensions, extracting a parameter subset applicable across message types as the general scheduling parameter;

[0034] analyzing the characteristics strongly related to the specific service in the comprehensive priority offset parameter, and extracting a parameter subset acting only on the current message type as the special scheduling parameter;

[0035] The characteristics strongly related to the specific service are characteristics whose Manhattan distance value between the current message type feature vector and the cluster center vector exceeds the similarity threshold.

[0036] Preferably, the dynamic adjustment of the protection message forwarding queue comprises:

[0037] applying the general scheduling parameter as the reference rule to all protection message forwarding queues;

[0038] generate a real-time scheduling strategy based on the special scheduling parameters, and remap the queue priority of the specific message type according to the benchmark rule;

[0039] update the value range of the special scheduling parameters through policy execution feedback data.

[0040] Preferably, the remapping process comprises:

[0041] construct a priority mapping matrix, wherein the row dimension of the priority mapping matrix corresponds to the message type, and the column dimension corresponds to the forwarding queue level;

[0042] update the weight distribution of the specific message type row in the matrix based on the real-time scheduling strategy;

[0043] when detecting a queue conflict, reassign the message forwarding time slot according to the matrix weight distribution.

[0044] Preferably, the method further comprises a closed-loop optimization process:

[0045] collect message transmission quality data after priority adjustment;

[0046] calculate the verification deviation of the current scheduling strategy based on the message transmission quality data;

[0047] when the verification deviation exceeds the set threshold, return to the step of generating the comprehensive priority offset parameter for recalculation;

[0048] when the verification deviation is lower than the set threshold, synchronize the current scheduling parameter to the historical priority strategy library.

[0049] Compared with the prior art, the present application has the following beneficial effects:

[0050] In the stage of obtaining the protection message type set of the SPN tunnel and extracting the message scheduling parameters, the message scheduling parameters are subdivided into permanent priority parameters and temporary priority parameters by comprehensively sorting the protection message types. This fine parameter division makes the priority judgment of different types of messages more accurate. Compared with the previous general priority setting method, it can better meet the actual needs of various businesses for message transmission priority. For example, in the industrial automation scene, the permanent priority of the protection message corresponding to some key control instructions can be set to a higher value to ensure that it is transmitted preferentially in any case; and the temporary priority of the protection message of some auxiliary monitoring data can be flexibly adjusted according to the real-time network status, appropriately reducing its priority when the network is busy to free up resources for critical businesses, and restoring normal when the network is idle, thereby realizing reasonable allocation of resources and avoiding resource waste or excessive concentration.

[0051] The priority trigger signal of the protection packet is generated based on the packet scheduling parameter, and the priority change node is identified, which greatly enhances the response sensitivity of the system to network changes. Whether it is a node triggered by permanent priority parameter update or a node triggered by temporary priority parameter reset, the system can quickly capture the changes. Taking the financial transaction system as an example, when the market fluctuates sharply and the number of transaction instructions increases sharply, the system can quickly improve the priority of the transaction instruction protection packet according to the preset rules through the permanent priority parameter update trigger node, ensure the timely and accurate transmission of transaction information, and avoid transaction errors caused by information delay. At the same time, when the network is suddenly congested, the temporary priority parameter reset trigger node can timely adjust the priority of some non-critical business protection packets, relieve the network pressure, and ensure the stable operation of the core business.

[0052] In terms of generating comprehensive priority offset parameters, the permanent priority offset and the temporary priority offset are calculated by classification, which can comprehensively consider the influence of various factors on the priority of the packet. Compared with a single priority calculation method, this method is more comprehensive and flexible. For example, in the video live streaming scenario, the key frame protection packet of the live streaming picture can set the permanent priority offset according to the picture quality requirement and the importance of live streaming, to ensure smooth and clear picture; and the temporary priority offset of the audio packet and other relatively secondary protection packets can be dynamically adjusted according to the real-time bandwidth of the network, to give more resources to the transmission of video key frames and improve the overall live streaming effect on the premise of ensuring basic audio quality.

[0053] In the step of dynamically adjusting the forwarding queue of the protection packet, the general scheduling parameter that can be multiplexed across packet types and the special scheduling parameter that is only applicable to a specific packet type are separated, so that the adjustment of the forwarding queue is more scientific and efficient. For the general scheduling parameter, it can be uniformly applied in multiple packet types, reducing repeated calculation and configuration and improving system operation efficiency. For example, in most businesses, the general rule of high-priority packet priority forwarding can be uniformly set through the general scheduling parameter. For specific packet types, such as DICOM format packets in medical image transmission, the special scheduling parameter can develop more stringent forwarding strategies for such packets according to their characteristics, to ensure the safe and reliable transmission of medical image data and avoid data loss or errors, providing strong support for medical diagnosis. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 The working principle diagram of the priority scheduling method of the protection packet in the SPN tunnel described in the present application;

[0055] Figure 2 The flowchart for obtaining and merging parameters for the protection packet type set;

[0056] Figure 3 Flow chart for priority trigger signal generation;

[0057] Figure 4 Flow chart for separation of general and special scheduling parameters;

[0058] Figure 5 Flow chart for closed loop optimization. DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than 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 the present application.

[0060] Referring to Figure 1 The present application provides a priority scheduling method for protecting packets in an SPN tunnel, which comprises:

[0061] A set of protected packet types of the SPN tunnel is obtained, and packet scheduling parameters including permanent priority parameters and temporary priority parameters are extracted. A priority trigger signal is generated based on the packet scheduling parameters, and priority change nodes are identified, including nodes triggered by permanent priority parameter updates and nodes triggered by temporary priority parameter resets. Permanent priority offsets and temporary priority offsets are calculated and classified, and comprehensive priority offset parameters are generated. General scheduling parameters and special scheduling parameters are separated, and the protected packet forwarding queue is dynamically adjusted. The general scheduling parameters are applicable to all packet types, and the special scheduling parameters are applicable to packet types with transmission integrity and accuracy requirements greater than a preset threshold, such as transmission integrity and accuracy requirements higher than 90%. Queue remapping is achieved through a priority mapping matrix, and the scheduling strategy is continuously adjusted in combination with a closed loop optimization mechanism.

[0062] Embodiment 1: Referring to Figure 2 The initial stage of the priority scheduling method for protected packets involves the generation and optimization of packet scheduling parameters. This process strictly follows the dual constraints of transmission delay requirements and traffic burstiness characteristics. Permanent priority parameters are divided based on the tolerance of traffic transmission delay. For example, voice traffic requires a transmission delay of less than 50 ms, and its permanent priority is set to the highest level; video monitoring traffic can accept a 100 ms delay and is assigned a medium priority; data backup traffic has a delay threshold greater than 500 ms, and is classified as low priority. Independent initial priority parameter entries are generated for each type of traffic, and each entry is associated with a clear effective period. The effective period is set according to the characteristics of the business cycle, such as setting high priority from 9:00 to 18:00 on weekdays, and using the basic priority during the rest of the period.

[0063] The temporary priority parameter is designed for burst demand of services. When the network monitors a burst congestion event in a specific area, a temporary priority parameter set is generated for the affected services. The set contains three key fields: burst service type identification, temporary priority value, and valid time window. The valid time window is dynamically set according to the network's predicted congestion duration, for example, if the link failure recovery is estimated to take 120 seconds, the temporary priority will take effect within this window. When the temporary parameter is generated, the current network topology state needs to be scanned, and if the node load rate exceeds 85%, an additional weight coefficient is given to the services passing through the node.

[0064] The parameter verification rule performs secondary correction on the temporary parameter set. The core basis of the correction is the quantitative analysis of the priority deviation amount. The deviation amount calculation is divided into two levels: the first level detects the numerical difference between the permanent priority and the temporary priority. Taking voice services as an example, if the permanent priority is level 5 and the temporary priority is raised to level 3, the absolute deviation amount is 2; the second level introduces time dimension analysis to calculate the interval length between the parameter change timestamp and the initial setting timestamp. The shorter the time interval, the more urgent the change requirement, and the greater the deviation amount weighting coefficient. The weight coefficient setting rule is: the coefficient is 1.5 when the interval is less than 1 minute, 1.2 when the interval is between 1 and 5 minutes, and 1.0 when the interval is greater than 5 minutes.

[0065] The aggregation calculation of the deviation amount adopts a phased processing mode. The packet transmission process is divided into queue waiting stage, node forwarding stage, and link transmission stage. The deviation amount of each stage is calculated independently: the queue waiting stage mainly focuses on delay deviation, the node forwarding stage focuses on packet loss rate deviation, and the link transmission stage calculates bandwidth utilization rate deviation. The total deviation amount is obtained by adding the weights of the deviation amounts of each stage, and the weight value is dynamically configured according to the service type. For example, the packet loss rate deviation weight accounts for 60% and the delay deviation accounts for 30% in high reliability services; and vice versa in real-time services.

[0066] Based on the proportional distribution of the total deviation amount, the weighted correction of the temporary parameter is performed. The specific operation includes: when the deviation amount of a certain stage accounts for more than 40% of the total deviation amount, the temporary parameter value associated with that stage is compressed and adjusted. The compression amplitude follows a linear decay rule, and the parameter value is lowered by 5% for every 10% increase in deviation ratio. At the same time, boundary checking is performed to ensure that the temporary priority after correction does not exceed the upper limit of the preset floating range. For multi-service superposition scenarios, a conflict resolution strategy is adopted: scan the priority parameters of all affected services, and if there is a logical conflict (such as two services requiring the highest priority at the same time), downgrade the temporary parameter value of the secondary service according to the service value level.

[0067] The modified temporary parameter set and the initial permanent parameter set are merged through a three-dimensional matrix. The X-axis of the matrix represents the service type, the Y-axis represents the time dimension, and the Z-axis corresponds to the network area. The matrix elements contain four-tuple data: base priority, temporary priority, effective timestamp, and scope identifier. During merging, data fusion verification is performed: superposition detection is performed on parameters with the same space-time dimension, and if the parameter value difference is less than the threshold, the parameters are directly merged; if the difference exceeds the limit, an iterative approximation algorithm is enabled, with a step size of 0.5 to gradually converge the priority value until the difference threshold requirement is met. Finally, the packet scheduling parameter library is output, containing complete priority configuration of all service types in the space-time dimension.

[0068] The parameter library implements a dynamic maintenance mechanism. The network state perception module continuously collects three key indicators: link load rate, node cache depth, and service request frequency. When the change rate of any of the indicators exceeds 15%, the parameter recalculation process is triggered. Recalculation uses an incremental update mode, adjusting only the parameters of the affected service types and time periods, avoiding resource loss caused by global refresh. During the adjustment process, the effective time period parameters are updated synchronously: if the predicted business peak ends early, the temporary priority window is shortened; if new business burst requests are detected, the time window is expanded and a modified parameter entry is inserted. The entire parameter generation system forms a closed loop control, ensuring that the scheduling parameters are always synchronized with the real-time network state.

[0069] Embodiment 2: refer to Figure 3 The generation of the priority trigger signal is based on dynamic monitoring of packet scheduling parameters. When a change in the initial priority parameter set is detected, the system generates a first trigger signal. The signal carries three groups of core data: the changed service type identifier, the new permanent priority value, and the effective start timestamp. The signal generation logic includes a double verification mechanism: first, verify whether the new priority breaks the pre-set value range (e.g., 0-7 levels), and second, verify whether the timestamp matches the business cycle. If the effective period spans multiple business cycles, it is automatically split into multiple discrete signals, each signal acting on a single cycle window.

[0070] The change in the temporary parameter set triggers the generation of a second trigger signal. Before generation, an effectiveness test is performed: the current transmission quality indicators (delay jitter, packet loss rate) of the service to be changed are extracted, and if any indicator deviates from the baseline value by more than 20%, the signal is allowed to be generated. The signal content includes the temporary priority increment value, the duration window, and the effective business scenario code. The scenario code distinguishes between network failure, burst traffic, and link switching, for example, automatically adding a path redundancy identifier in the link switching scenario. After the signal is generated, it enters the buffer queue and waits for collaborative analysis with the first trigger signal.

[0071] The interaction analysis of the two types of trigger signals uses a time sequence correlation algorithm. The time sequence correlation strength calculation formula is as follows:

[0072]

[0073] wherein: represents the priority change trend intensity, is the signal type weight factor (0.7 for permanent signal, 0.3 for temporary signal), is the absolute value of the timestamp difference of the two types of signals (unit: seconds), is the historical signal pattern matching degree (calculated by cosine similarity).

[0074] The timing window is set as a dynamic value, and the base window is 1.5 times the minimum period of the service. When the value exceeds the set threshold (default 0.65), the signal generation timing adjustment mechanism is triggered: if > 0.8, the signal is generated in advance (the amount of advance is × 0.4), if < 0.3, the signal is generated with delay (the amount of delay is × 0.6).

[0075] The identification of the priority change node relies on the historical signal feature library. The feature library stores the trigger signal records of the past 30 days, and each record contains the signal sequence fingerprint and context metadata. The identification of the permanent priority update node performs a three-step process: calculating the Euclidean distance between the current trigger signal and the historical permanent change node, the distance value is processed by time axis regularization; extract the change slope of the continuous three points in the signal sequence, calculate the Pearson correlation coefficient of the current signal and the historical signal slope; generate the trend correlation index by combining the distance and the correlation coefficient. When the index exceeds the correlation threshold (default 0.7), a permanent priority update node marker is inserted in the signal stream. The node marker contains topological position information to ensure accurate association with the network device processing unit.

[0076] The identification of the temporary priority reset node needs to solve the signal drift problem. The analysis process establishes a double-layer filtering mechanism: the first layer scans the time distribution law of the historical temporary reset node, and calculates the probability of the current signal appearing in the historical distribution window; the second layer extracts the signal intensity change direction feature (increment / decrement), and calculates the direction consistency index. The trend deviation is generated by reverse deduction:

[0077]

[0078] wherein: represents the trend deviation, is the current signal feature vector (including priority change amount, service type weight), is the average feature vector of the historical node, is the anti-zero protection constant (0.001), when >0.75 is marked as a temporary reset node. For direction conflict signals (such as a rising signal should be promoted but a falling signal is detected), a secondary verification process is forcibly triggered: the cache depth data of the service at the last 3 forwarding nodes is re-collected, and if the depth value is out of limits, the signal direction is reversed.

[0079] After the node is marked, topology matching calibration is performed. The current tunnel path node sequence is extracted from the network management plane, and spatial mapping is performed with the topology information of the node marking. The mapping process solves two types of conflicts: cross-node signal conflicts use a path weighted voting mechanism (core node weight x 1.5, edge node x 0.8); timing conflicts enable a logical clock synchronization algorithm, and the calibrated node information is packaged as a control instruction. The instruction transmission uses a datagram protocol with a retransmission mechanism to ensure a delivery success rate of ≥99.5% in a network congestion environment. The system maintains a node state mirror table, which is updated every 200ms, providing real-time context for subsequent offset calculation.

[0080] The signal abnormality processing mechanism runs throughout the entire process. Four types of abnormal scenarios are defined: signal loss (no new signal for more than 150ms), signal oscillation (direction change ≥3 times within 5 seconds), value out of range (priority exceeding the device support range), and timing chaos (new signal timestamp earlier than existing signal). When an abnormality is detected, a three-level processing strategy is activated: the primary strategy discards illegal signals and reissues requests; the intermediate strategy backtracks 30 seconds of valid signals for interpolation repair; the advanced strategy freezes the scheduling parameter changes of the service type and waits for manual intervention. All abnormal events are recorded in a time-stamped audit log for subsequent strategy optimization. The entire signal processing process forms an adaptive control closed loop, converting the original scheduling parameters into executable network node instruction sequences.

[0081] Example 3: The generation of the comprehensive priority offset parameter is based on the accurate identification of priority change nodes. The extraction of permanent priority update nodes uses a sliding window detection algorithm, with the window size dynamically adjusted to 1 / 4 of the service period. Three types of features are scanned within each window: priority value jump amount, service type distribution density, and timestamp interval standard deviation. When the value jump is detected to exceed 2 levels and the service density is higher than the threshold, it is marked as an effective update node. The correlation between nodes is verified by the transmission path hash value to ensure that only consecutive nodes of the same service flow are processed. The determination of adjacent service periods is based on the network topology time synchronization protocol, with a period boundary error controlled within ±50ms.

[0082] The calculation of the permanent priority offset adopts a differential accumulation mode. The calculation process of the first offset component ΔP contains three processing stages: a basic differential stage calculates the arithmetic difference value of the priority parameters of the current period and the last period; a time domain compensation stage introduces a period length scaling factor to amplify and correct the offset in the short period; and a service weight adjustment stage multiplies a type importance coefficient. The service type importance coefficient is divided into five levels, increasing by 0.25 from 0.5 to 1.5. The final first offset component ΔP is expressed as:

[0083]

[0084] wherein: P k represents the permanent priority value of the kth period; is a period length compensation factor (current period / base period); is a service type weight coefficient; is the number of continuously updated nodes.

[0085] The processing of the temporary priority reset node adopts an event-driven mechanism. The system maintains a ring buffer with a length of 8 to store the recent temporary priority change events. Each event record contains four dimensions of data: trigger time, service type identifier, original priority, and target priority. Service type switching detection is achieved by comparing the type identifiers of consecutive events, and a type switching event is determined when non-consecutive identifiers appear. The calculation of the second offset component ΔT introduces a moving average filter, taking the average value of the priority changes of the last three type switching events as the base value, and then multiplying by the current network load factor α. The load factor is calculated by monitoring the standard deviation of the link utilization rate in the period, and the value range is 0.8-1.2.

[0086] The linear combination of the offset components implements dynamic weight distribution. The weight calculation module monitors two types of indicators in real time: service flow stability index and network environment fluctuation coefficient. The service flow stability index is represented by the reciprocal of the variance of the priority changes in the last 10 periods; the network environment fluctuation coefficient is calculated based on the link quality probe data, including the packet loss rate change gradient and the delay jitter amplitude. The combined weight is dynamically adjusted according to the following rules: when the service flow stability index is higher than 0.7, the weight of the first offset component is increased by 15%; when the network environment fluctuation coefficient exceeds the warning value, the weight of the second offset component is increased by 20%. The weight adjustment process sets an anti-shock mechanism, and the single adjustment amplitude does not exceed 5%, and the adjacent adjustment interval is at least 300ms.

[0087] ​The generation of the comprehensive priority offset parameter Φ implements three-level verification. The primary verification checks the consistency of the offset sign to prevent permanent and temporary offset direction conflicts; the intermediate verification compares the deviation of the parameter value from the historical average value, and triggers manual review when the deviation exceeds three times the standard deviation; and the final verification verifies the effectiveness of the parameter within the device support range. The verified parameter is encoded as a control instruction, and the instruction format includes: 8-bit service type identifier, 16-bit offset value, 8-bit weight combination identifier, and 32-bit effective timestamp. The instruction transmission uses a layered encapsulation protocol, the physical layer uses forward error correction encoding, and the data link layer implements CRC-32 verification.

[0088] The parameter application stage implements a gradual adjustment strategy. The comprehensive priority offset parameter is divided into multiple fine-tuning steps, and the adjustment amount of each step is 1 / 5-1 / 3 of the total offset. The step interval time is automatically configured according to the service sensitivity: 50ms short interval for voice services, 100ms medium interval for video services, and 200ms long interval for data services. Three key indicators are collected after each fine-tuning: queue waiting time change rate, node processing throughput, and link bandwidth utilization. When any of the indicators deteriorates by more than 10%, the system automatically rolls back to the last stable version and starts the offset recalculation process.

[0089] The update of the historical parameter library implements differentiated management. For the permanent priority offset component, the complete record of the last 20 calculation periods is retained, and the old data is eliminated using the LRU algorithm; for the temporary priority offset component, only abnormal values (more than twice the average value) and mode transition points (continuous three times of the same direction change) are stored. Parameter library retrieval uses a multi-level index structure, the first level is hashed partitioned according to service type, the second level is indexed using a B+ tree according to time range, and the third level is constructed using a skip list according to offset size. An operation log is generated synchronously every time the parameter is updated, and the log entry includes operator identity, modification time, pre-modification value, post-modification value, and audit status, etc. metadata, meeting the network operation and maintenance audit requirements.

[0090] The abnormal handling mechanism is designed for six typical scenarios: offset overflow (exceeding the device processing range), weight combination failure (not equal to 1), service type conflict (multiple services competing for the same queue), time sequence out of sync (node clock deviation > 100ms), transmission packet loss (continuous 3 times of confirmation timeout), and calculation timeout (single processing exceeding 500ms). Each abnormality corresponds to a recovery strategy, including parameter truncation, weight rebalancing, service priority degradation, clock resynchronization, data retransmission, and calculation task migration. All abnormal events trigger associated analysis, and an abnormal pattern recognition model is established for prediction and prevention of similar problems. The system maintains an abnormal handling knowledge base, accumulates more than 200 experience rules, and supports fuzzy matching and similarity retrieval.

[0091] Embodiment 4: see Figure 4The separation and dynamic adjustment of the priority offset parameter is achieved through a hierarchical processing mechanism. In the analysis of common characteristics phase, the system establishes a multi-dimensional feature matrix, taking packet transmission delay and packet loss rate as the core dimensions, and expanding seven derived dimensions: delay fluctuation rate, packet loss event interval, bandwidth occupancy ratio, queue depth change rate, node processing delay, path hop count, and service period stability. Each dimension sets an independent feature extractor, and uses a sliding window method to calculate the statistics. The delay dimension extracts the maximum value, minimum value, and 90th percentile within the window. The packet loss rate dimension calculates the weighted moving average, with recent packet loss events given higher weights. The rows of the feature matrix represent different packet types, and the columns correspond to the feature values of each dimension, forming a data structure that can be compared horizontally.

[0092] The extraction of cross-packet type reusable parameters is completed through a feature clustering algorithm. The system presets three clustering centers: strict real-time type, elastic reliable type, and best effort type. Each clustering center defines a range of feature values, such as a strict real-time type with a delay upper limit of 50ms and a packet loss rate threshold of 0.1%. The Euclidean distance between the feature vector of all packet types and the clustering center is calculated, and the one with the closest distance and meeting the threshold condition is assigned to the cluster. The universal scheduling parameters are generated from the common features in the cluster, including three types of basic rules: queue scheduling period (strict real-time type uses a 10ms period, and other types use a 20-50ms period), packet loss retransmission strategy (elastic reliable type allows two retransmissions, and other types allow one), and bandwidth reservation ratio (strict real-time type reserves 30%, and other types reserve 15%).

[0093] The identification of specific service-related features uses a difference analysis method. The Manhattan distance between the feature vector of the current packet type and the clustering center vector is calculated, and the dimensions with a distance value exceeding the similarity threshold are marked as specific dimensions, indicating that the feature is strongly related to a specific service. The dedicated scheduling parameters are generated from the specific dimension combination, such as a video monitoring service that is classified as a strict real-time type but has a significantly higher path hop count than the clustering center value. Therefore, a dedicated parameter "cross-node direct pass flag" is generated, allowing the service to bypass the regular queue detection of intermediate nodes. The dedicated parameters are stored in the form of key-value pairs, with the key being the concatenation of the service type code and the feature dimension code, and the value containing the parameter content and the effective condition.

[0094] Table 1: Parameter separation results for three typical services.

[0095]

[0096] The dynamic adjustment of the protection packet forwarding queue implements a hierarchical control strategy. The general scheduling parameters are loaded as basic rules into the policy template library of all network nodes, and the update period is fixed at 15 minutes. After receiving the new parameters, the node first simulates the operation in the check area to verify the compatibility of the parameters and the hardware configuration, and only after passing the verification will it be activated to the running area. The real-time strategy generator of the special scheduling parameters includes a strategy synthesis engine, and the engine inputs include: the current network topology state, the quality of service monitoring data, and the historical strategy execution record. The synthesis process goes through four stages: the demand analysis stage converts the service characteristics into QoS demand items; the resource matching stage finds the available bandwidth and queue resources; the conflict detection stage identifies the resource competition between strategies; and the optimization output stage generates a strategy combination with a weight coefficient.

[0097] The queue priority remapping implements a three-step processing flow. First, a priority mapping matrix is established, with the rows corresponding to the separated service types and the columns representing the four standard queue levels (emergency, priority, regular, and best effort). The matrix elements include initial weights and dynamic adjustment factors, for example, the initial weight of emergency voice service in the emergency queue is 1.0, and the dynamic factor fluctuates in the range of 0.8-1.2 according to real-time network load. Second, real-time strategy injection is performed, which converts special scheduling parameters into matrix update instructions, including target row and column positions, weight increments, and effective time length. Finally, queue conflicts are handled when multiple services compete for the same queue level, and the weighted round-robin method is used to allocate resources by comparing their strategy weights and service criticality levels.

[0098] The strategy execution feedback mechanism builds a double-layer monitoring system. The data collection layer deploys probes at three locations: the entrance node records the original scheduling parameter effect, the intermediate node monitors the strategy execution deviation, and the exit node evaluates the final service quality. The analysis layer calculates four key indicators: parameter matching degree (similarity between expected and actual effects), strategy coverage rate (proportion of services affected by special parameters), resource utilization rate (ratio of allocation to consumption), and conflict resolution rate (proportion of successfully handled competition events). Feedback data is returned through the control plane channel, and the update period can be configured, with a default setting of 5 minutes. The range adjustment uses a gradual method, allowing only a 10% parameter value change per update, and must pass three consecutive sampling verifications before being confirmed effective.

[0099] The closed-loop optimization process pays special attention to the life cycle management of dedicated parameters. Newly generated dedicated parameters have a default validity period of 2 hours, and each time the feedback data is verified to be valid, the validity period is extended by 50%. The system maintains a parameter value evaluation model, and calculates a comprehensive score based on the parameter usage frequency, effect improvement amplitude, and resource consumption cost. Parameters with scores below the threshold are automatically downgraded to observation mode and no longer applied in practice; parameters with scores exceeding the threshold for three consecutive periods may be promoted to general parameter candidates. All parameter change records form a version chain, supporting quick rollback to any historical version. Smooth transition is performed when switching versions, and the new and old parameters run in parallel for a period of time before gradually fading out the old parameters.

[0100] The abnormal scenario processing sets up a special state recovery module. When a policy execution anomaly is detected, the system automatically triggers a three-level response: the primary response suspends the execution of related parameters, the intermediate response reverts to the last stable configuration, and the advanced response starts cross-node coordinated adjustment. Each abnormality processing generates a diagnostic report, marking the possible cause chain and repair suggestions. Common abnormal patterns are classified and coded, forming a response knowledge base containing 32 standard scenarios, supporting fuzzy matching and similar case retrieval. The knowledge base is automatically updated every 24 hours, collecting the latest processing experience from all network nodes.

[0101] Embodiment 5: refer to Figure 5 The construction of the priority mapping matrix adopts a three-dimensional dynamic structure. The first dimension corresponds to the separated business types, which are divided into eight levels according to business criticality, from the highest real-time voice business to the lowest priority data backup business. The second dimension represents the forwarding queue hierarchy, designed as a four-level structure: emergency queue, guarantee queue, regular queue, and elastic queue. The third dimension introduces a time axis, slicing the matrix snapshots in business cycles. The matrix elements store a composite data structure, including a basic weight value, a dynamic adjustment factor, and a historical trend record. The basic weight value is pre-set according to business level, such as setting the basic weight of voice business in the emergency queue to 0.95; the dynamic adjustment factor floats within ±0.15 according to real-time network load; the historical trend record saves the value sequence of the last five adjustments.

[0102] The injection of real-time scheduling strategies implements an update mechanism with conflict detection. The update instruction contains target coordinate locators, composed of business type code, queue level code, and time slice index. The strategy engine generates an incremental instruction package after analyzing dedicated scheduling parameters, such as adding 0.2 weight to industrial control business in the guarantee queue level. Before updating, conflict pre-detection is performed: scanning the three business types and two queue levels adjacent to the target position to detect whether there is weight competition. When a competition conflict is detected, the weight redistribution algorithm is started: calculating the priority difference of competing businesses, and prioritizing high-priority businesses; for businesses of the same level, time slice rotation distribution is used, and each business obtains a fixed time window with high weight.

[0103] Queue conflict resolution adopts time slot dynamic division technology. Time slot allocation unit maintains a timeline model of physical port sending cycle, which divides each cycle into 256 micro-slots. When a conflict is detected, the service time slot proportion is calculated according to the weight distribution of the mapping matrix. When the voice service weight is 0.95, 48% of the total time slot is allocated; when the industrial control service weight is 0.8, 38% of the total time slot is allocated; the remaining time slot is used as a buffer pool. The allocation process implements sliding window optimization: the actual queue usage rate is counted every 10 cycles, and if a service has a usage rate of less than 80% for three consecutive cycles, part of its time slot is temporarily borrowed to a high demand service. Time slot reallocation instructions are issued through an in-band control channel, which shares the transmission channel with data packets but uses a higher priority marker.

[0104] Message transmission quality monitoring implements a full-path tracking mechanism. A marker header is implanted at the service flow entry node, containing initial parameter fingerprints and sampling point identifiers. Each forwarding node in the path adds local processing records: queue waiting time, scheduling decision timestamp, and actual forwarding quality. The exit node generates a quality report after collecting complete path records, with core indicators including: end-to-end delay and jitter value, effective throughput, out-of-order packet ratio, and marker loss rate. The quality data reporting period is dynamically adjusted: a 5-minute regular cycle is used when the network is stable, and a 30-second emergency cycle is used when an anomaly is detected.

[0105] The check deviation amount calculation adopts a multi-dimensional comparison model. A target quality template is established, including delay upper limit, throughput lower limit, and packet loss rate threshold. The actual quality data is compared with the template to generate the original deviation value. The deviation items related to permanent priority offset are multiplied by a time decay coefficient, with higher weight for recent data; the deviation items related to temporary priority offset consider the burst intensity factor. The final check deviation amount is calculated by weighted root mean square, with delay deviation weight accounting for 40%, throughput deviation accounting for 35%, and packet loss rate deviation accounting for 25%. Environmental calibration is added to the calculation process: when the overall network load exceeds 75%, all deviation amounts are multiplied by a tolerance coefficient of 0.8.

[0106] The policy closed-loop optimization executes a dual-mode judgment. The check deviation threshold is set as a dynamic floating value, with a basic threshold of 0.15 that automatically increases to 0.2 during late peak hours. When the deviation exceeds the threshold, the parameter recalculation process is triggered: the current queue scheduling policy is frozen, and the last stable configuration is rolled back; a recalculation request is sent to the control center, carrying detailed quality analysis reports; the system prioritizes the recalculation of comprehensive priority offset parameters, and if the threshold is still exceeded after two consecutive recalculation, the general / specialized parameter separation is extended. When the deviation is below the threshold, policy solidification is performed: a policy snapshot is extracted from the current effective parameters; the snapshot is matched with the historical policy library for similarity, and if the matching degree is higher than 90%, only the timestamp is updated; if the matching degree is lower than 70%, it is stored as a new entry in the policy library, with the associated service type marked.

[0107] The historical priority policy library implements intelligent compressed storage. The library structure is partitioned by service type, and each partition contains three sub-libraries: the permanent offset policy library stores typical patterns for nearly three months; the temporary offset policy library retains high-frequency use (weekly frequency > 5) policies; and the exception handling library accumulates processing solutions for events with check deviation exceeding the limit. Storage optimization uses pattern recognition technology: similar policies are merged, and parameter value ranges are retained instead of fixed values; obsolete policies (unused for 30 consecutive days) are automatically archived to secondary storage; and conflicting policies are marked with mutual exclusion to prevent simultaneous activation. Policy retrieval supports multi-condition combined queries, including service features, network status, time range, and other dimensions.

[0108] The exception handling mechanism focuses on six typical scenarios. In the clock desynchronization scenario, a nanosecond-level time synchronization protocol is started to calibrate node clocks within 50 ms. When the queue overflows, three-level responses are executed: immediately discard the lowest priority messages, dynamically expand the queue cache, and send a backpressure signal to the upstream node. When policy oscillation detects three consecutive direction reversals, the parameter adjustment function is locked and an alarm is issued. Transmission interruption processing uses a backup control channel to resend instructions, while recording the interruption position for path optimization. When computing resources are overloaded, processing tasks are migrated to adjacent nodes, and the migration process maintains scheduling state consistency. Policy conflict resolution calls the historical case library to find a solution template for similar scenarios.

[0109] The self-healing module is built on an event bus, which listens to four types of core events: quality exceeding the standard triggers deviation recalculation; node failure events activate standby path switching; configuration change events start policy compatibility checks; and business growth events trigger resource pre-allocation. Each event corresponds to a standard processing pipeline: the event filtering layer filters out non-critical information; the pattern matching layer associates historical processing solutions; the execution layer calls pre-installed processing programs; and the feedback layer collects processing effect data. The processing process generates an operation trace chain to support full-link audit tracking. The system automatically generates an optimization suggestion report every week, pointing out weak links and improvement directions in the policy library.

[0110] It should be noted that, in this text, relational terms such as first and second are used merely to distinguish one entity or action from another, and do not necessarily require or imply that there is any such actual relationship or order between these entities or actions. Moreover, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such a process, method, article or device.

[0111] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. A priority scheduling method for protecting packets in an SPN tunnel, characterized in that, The method comprises the following steps: S1: obtaining a protection message type set of an SPN tunnel, extracting a message scheduling parameter based on the protection message type set; the message scheduling parameter comprises a permanent priority parameter and a temporary priority parameter; S2: generating a priority trigger signal of the protection message based on the message scheduling parameter; identifying a priority change node according to the priority trigger signal; the priority change node comprises a node triggered by permanent priority parameter updating and a node triggered by temporary priority parameter resetting; S3: classifying and calculating a permanent priority offset and a temporary priority offset based on the priority change node, and generating a comprehensive priority offset parameter; S4: separating general scheduling parameters that can be multiplexed across message types and special scheduling parameters that are only applicable to specific message types based on the comprehensive priority offset parameter, and dynamically adjusting a protection message forwarding queue according to the separation result; The specific message type is a message type whose requirements for integrity and accuracy of transmission are greater than a preset threshold; The obtaining of the protection message type set of the SPN tunnel comprises: dividing the permanent priority parameter based on message transmission delay requirements, generating an initial priority parameter set and a corresponding effective period; dividing the temporary priority parameter based on service burstiness requirements, generating a temporary parameter set and a corresponding effective period; obtaining a corrected temporary parameter set through a parameter verification rule based on the initial priority parameter set and the corresponding effective period; merging the initial priority parameter set and the corrected temporary parameter set into the message scheduling parameter; The identification of the priority change node comprises: calculating the trend correlation degree of the priority trigger signal and the historical permanent priority change node, and marking it as a permanent priority updating node when the trend correlation degree is greater than a set correlation threshold; calculating the trend deviation degree of the priority trigger signal and the historical temporary priority resetting node, and marking it as a temporary priority resetting node when the trend deviation degree is greater than a set deviation threshold; The separation of the general scheduling parameters that can be multiplexed across message types and the special scheduling parameters that are only applicable to specific message types comprises: analyzing the common features in the comprehensive priority offset parameter based on message transmission delay and packet loss rate as the reference dimension, and extracting a parameter subset that can be multiplexed across message types as the general scheduling parameter; analyzing the features strongly related to specific services in the comprehensive priority offset parameter, and extracting a parameter subset that only acts on the current message type as the special scheduling parameter; The feature strongly related to the specific service is a feature whose Manhattan distance value between the feature vector of the current message type and the cluster center vector exceeds the similarity threshold.

2. The priority scheduling method for protection messages in an SPN tunnel according to claim 1, characterized in that, The obtaining of the corrected temporary parameter set through the parameter verification rule comprises: calculating the priority deviation amount of each message transmission stage, which is determined by the time sequence relationship among the permanent priority parameter weight coefficient, the parameter change timestamp, and the initial parameter setting timestamp; weighting and correcting the parameter values of each stage in the temporary parameter set based on the proportion of the priority deviation amount in the total deviation amount.

3. The priority scheduling method for protection messages in an SPN tunnel according to claim 1, characterized in that, The generation of the priority trigger signal of the protection message comprises: generating a first trigger signal according to the initial priority parameter set; generating a second trigger signal according to the modified temporary parameter set; calculating a priority change trend strength based on the timing correlation between the first trigger signal and the second trigger signal; generating a priority trigger signal in advance or delaying the generation of the priority trigger signal when the priority change trend strength exceeds a set threshold.

4. The method of claim 1, wherein the priority of the packet is determined according to a priority of a service of the packet. The generating of the comprehensive priority offset parameter includes: extracting a permanent priority update node and calculating the offset of the permanent priority parameter in the adjacent service period, denoted as a first offset component; extracting a temporary priority reset node and calculating the temporary priority offset when the service type is switched, denoted as a second offset component; linearly combining the first offset component and the second offset component as the comprehensive priority offset parameter.

5. The method of claim 1, wherein the priority of the packet is determined according to a priority of a service associated with the packet. The dynamic adjustment of the protection packet forwarding queue includes: applying a general scheduling parameter as a reference rule to all protection packet forwarding queues; generating a real-time scheduling strategy based on a special scheduling parameter, and remapping the queue priority of the specific packet type in combination with the reference rule; updating the value range of the special scheduling parameter through policy execution feedback data.

6. The method of claim 5, wherein the priority of the packet is determined according to the following table: Table 1: Priority of the packet in the SPN tunnel The remapping process includes: constructing a priority mapping matrix, the row dimension of the priority mapping matrix corresponding to the packet type, and the column dimension corresponding to the forwarding queue level; updating the weight distribution of the specific packet type row in the matrix based on the real-time scheduling strategy; when a queue conflict is detected, reallocating the packet forwarding time slot according to the matrix weight distribution.

7. The method of claim 1, wherein the priority of the packet is determined according to a priority of the SPN tunnel. It also includes a closed-loop optimization process: collecting packet transmission quality data after priority adjustment; calculating the verification deviation of the current scheduling strategy based on the packet transmission quality data; when the verification deviation exceeds the set threshold, returning to the step of generating the comprehensive priority offset parameter for recalculation; when the verification deviation is lower than the set threshold, synchronizing the current scheduling parameter to the historical priority strategy library.

Images