A network slice control method, system and medium

Abstract

The present application relates to the technical field of network control, and especially relates to a network slice control method, system and medium. The method comprises the following steps: deploying temperature sensors on cabinets where each network element of a network slice passes, opening a precision time protocol offset measurement output port on each network element, and sequentially numbering the network elements in the service flow direction to form a slice path sequence; within a calibration window, synchronously collecting a temperature time sequence of the cabinet where each network element is located and a precision time protocol offset sequence of the network element based on the slice path sequence; and performing linear fitting with the temperature time sequence as the independent variable and the precision time protocol offset sequence as the dependent variable to calculate a temperature sensitivity coefficient of each network element. The present application realizes clock drift prediction and active control based on environmental temperature changes by introducing temperature field propagation modeling, and improves the time synchronization accuracy of network slices and the stability of service operation.

Description

Technical Field

[0001] This invention relates to the field of network control technology, and in particular to a network slicing control method, system, and medium. Background Technology

[0002] Network slicing is a key capability in the 5G core architecture, originating from the 3GPP TS 23.501 (system architecture) and TS 28.530 / 28.531 / 28.533 (management and orchestration) series of specifications. It refers to building multiple end-to-end logically independent "slices" on top of a shared physical infrastructure (RAN + bearer + core network) through virtualization and SDN / NFV technologies. Each slice tailors resources and functions for a specific service type (eMBB, URLLC, mMTC) to meet differentiated service level agreements.

[0003] In existing network slicing operations, the Precision Time Protocol (PTP) is typically used to achieve high-precision time synchronization between different network elements. However, in actual deployment environments, the switches, routers, edge computing nodes, and time-sensitive bridges that the network slice passes through are often distributed in different cabinets and different data center areas. Their operating environment is affected by factors such as air conditioning systems, cooling airflow organization, equipment load changes, and data center heat distribution, which cause changes in the frequency of the internal oscillators of the equipment, thus causing clock drift. Summary of the Invention

[0004] Therefore, the present invention needs to provide a network slicing control method, system and medium to solve at least one of the above-mentioned technical problems.

[0005] To achieve the above objectives, a network slicing control method includes the following steps: Step S1: Deploy temperature sensors in the racks of the network elements that the network slice passes through, enable the precise time protocol offset measurement output port on each network element, and number the network elements sequentially according to the direction of service flow to form a slice path sequence. Step S2: Within the calibration window, based on the slice path sequence, synchronously collect the temperature time series of the cabinet where each network element is located and the precise time protocol offset sequence of that network element. Step S3: Perform linear fitting with the temperature time series as the independent variable and the precise time protocol offset series as the dependent variable to calculate the temperature sensitivity coefficient of each network element; Step S4: Perform cross-correlation on the temperature time series of any two network elements in the slice path sequence, and take the time offset corresponding to the maximum value of the cross-correlation function as the temperature phase delay between the two network elements; construct the slice temperature response signature by combining the temperature sensitivity coefficient and the temperature phase delay. Step S5: Calculate the predicted time drift of each network element based on the slice temperature response signature, and perform phase-sensitive aggregation along the slice path to obtain the end-to-end predicted drift; invert the end-to-end predicted drift as the time offset compensation amount, and trigger cross-domain collaboration when the end-to-end predicted drift exceeds the preset service level protocol threshold.

[0006] The present invention also provides a network slicing control system for executing the network slicing control method described above, the network slicing control system comprising: The deployment numbering module is used to deploy temperature sensors in the cabinets of each network element that the network slice passes through, enable the precise time protocol offset measurement output port on each network element, and sequentially number the network elements according to the direction of service flow to form a slice path sequence. The synchronous acquisition module is used to synchronously acquire the temperature time series of the cabinet where each network element is located and the precise time protocol offset sequence of the network element, based on the slice path sequence within the calibration window. The coefficient fitting module is used to perform linear fitting with the temperature time series as the independent variable and the precise time protocol offset series as the dependent variable to calculate the temperature sensitivity coefficient of each network element. The signature construction module is used to cross-correlate the temperature time series of any two network elements in the slice path sequence, and take the time offset corresponding to the maximum value of the cross-correlation function as the temperature phase delay between the two network elements; the temperature sensitivity coefficient and the temperature phase delay constitute the slice temperature response signature. The aggregation compensation module is used to calculate the predicted time drift of each network element based on the slice temperature response signature, and perform phase-sensitive aggregation along the slice path to obtain the end-to-end predicted drift; the end-to-end predicted drift is inverted as the time offset compensation amount, and cross-domain collaboration is triggered when the end-to-end predicted drift exceeds the preset service level protocol threshold.

[0007] The present invention also provides a computer medium storing a computer program, wherein when the computer program is executed, it implements the network slicing control method described in any one of the above claims.

[0008] This invention establishes a direct correlation between temperature field changes in the physical environment of a network slice and PTP clock offset changes, realizing a closed-loop management mechanism from environmental temperature sensing and temperature propagation modeling to clock drift prediction and active control. Compared with the traditional method of relying on actual clock error detection and then compensation, this invention utilizes rack-level temperature acquisition and PTP cumulative offset acquisition to construct temperature sensitivity coefficients for each network element. This enables the system to quantify the response of different network elements to temperature disturbances, thereby transforming future clock drift, which is difficult to observe directly, into calculable and predictable temperature-driven parameters.

[0009] By performing cross-correlation analysis on the temperature time series of different cabinets, the propagation law of temperature disturbances in the slice path is identified, and the temperature field coupling strength and temperature phase delay are obtained. Then, a slice temperature response signature containing the temperature response characteristics of the network element itself and the temperature propagation relationship between network elements is constructed. This allows the system to not only grasp the degree of temperature influence on individual network elements, but also to grasp the propagation process of the temperature field in the entire slice path. On this basis, by dividing the temperature field into subdomains and identifying the temperature phase-leading network elements, the effective temperature is generated for the temperature phase-lag network elements using the historical temperature backtracking mechanism of the leading network elements. This enables the prediction model to reflect the temporal relationship of temperature propagation, avoids prediction bias caused by different network elements directly using the current temperature, and improves the consistency between drift prediction results and the actual thermal environment evolution process. Subsequently, the predicted time series drift of each network element is calculated by combining the temperature sensitivity coefficient, effective temperature, and PTP calibration cycle, and phase-sensitive aggregation is performed along the slice path to obtain the end-to-end predicted drift. This allows the future drift trend to be obtained in advance before the clock error accumulates and causes business impact.

[0010] By inverting the end-to-end predicted drift to form a time offset compensation and applying it to the packet transmission time of the egress network element, the anticipated synchronization deviation can be offset in advance, reducing the accumulation of synchronization errors in time-sensitive services. Simultaneously, when the predicted drift exceeds the service level agreement threshold, the system can automatically trigger cross-domain collaborative control measures such as protection bandwidth expansion, service rate rollback, and path switching. This extends time synchronization control from a single clock domain to the transmission resource domain, service control domain, and path orchestration domain, achieving multi-level joint regulation. This endows network slicing with proactive clock management capabilities based on temperature propagation patterns, improving time synchronization stability, drift control accuracy, and service assurance capabilities in complex environments. It is particularly suitable for applications with high clock synchronization accuracy requirements, such as time-sensitive networks, industrial internet, smart manufacturing, and ultra-low latency communication. Attached Figure Description

[0011] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1A flowchart illustrating the steps of a network slicing control method; Figure 2 This is a schematic diagram of the fitting of the temperature sensitivity coefficient of a network element and the removal of abnormal samples in one embodiment. Figure 3 This is a schematic diagram of a cross-domain collaborative control process according to one embodiment; Figure 4 This is a schematic diagram of a network slicing control system module. Detailed Implementation

[0012] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0013] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0014] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0015] To achieve the above objectives, please refer to Figures 1 to 4 This invention provides a network slicing control method, the method comprising the following steps: Step S1: Deploy temperature sensors in the racks of the network elements that the network slice passes through, enable the precise time protocol offset measurement output port on each network element, and number the network elements sequentially according to the direction of service flow to form a slice path sequence. Step S2: Within the calibration window, based on the slice path sequence, synchronously collect the temperature time series of the cabinet where each network element is located and the precise time protocol offset sequence of that network element. Step S3: Perform linear fitting with the temperature time series as the independent variable and the precise time protocol offset series as the dependent variable to calculate the temperature sensitivity coefficient of each network element; Step S4: Perform cross-correlation on the temperature time series of any two network elements in the slice path sequence, and take the time offset corresponding to the maximum value of the cross-correlation function as the temperature phase delay between the two network elements; construct the slice temperature response signature by combining the temperature sensitivity coefficient and the temperature phase delay. Step S5: Calculate the predicted time drift of each network element based on the slice temperature response signature, and perform phase-sensitive aggregation along the slice path to obtain the end-to-end predicted drift; invert the end-to-end predicted drift as the time offset compensation amount, and trigger cross-domain collaboration when the end-to-end predicted drift exceeds the preset service level protocol threshold.

[0016] Furthermore, step S1 includes the following steps: Step S11: Identify each network element traversed by the network slice and record it as the rack identifier sequence of the rack where each network element is located; In one embodiment, the slice orchestration controller obtains the service forwarding information of the target network slice. Based on the actual forwarding path of the service flow in the bearer network, it extracts all network element devices that the service flow passes through sequentially, and further queries the asset management information or data center resource management information corresponding to each network element to determine the rack location where each network element is installed. Subsequently, the rack identifiers corresponding to each network element are associated and recorded according to the order of service flow passage, forming a rack identifier sequence for subsequent environmental information mapping. The key point of this process is to establish the correspondence between the service path and the physical deployment environment, so that the subsequently collected ambient temperature can be accurately associated with the specific network element on the slice path.

[0017] For example, the data flow of a certain network slice starts from access switch A, passes through edge switch B, aggregation switch C, and core switch D in sequence, and finally reaches the service server. Switch A is deployed in rack R01, switch B in rack R03, switch C in rack R05, and switch D in rack R08. The system records a rack identification sequence "R01→R03→R05→R08", which serves as the physical environment mapping path corresponding to this network slice.

[0018] Step S12: Determine the cooling airflow inlet direction of the cabinet in the cabinet identification sequence, and install a temperature sensor at the inlet position along the cooling airflow inlet direction on the cabinet; In one embodiment, after obtaining the rack identification sequence, the heat dissipation structure of each rack is identified to determine the direction of cooling air entering the rack, and the location that best reflects changes in the external thermal environment is selected as the temperature acquisition point. Since the heat generated by the equipment inside the rack causes the exhaust temperature to be affected by the equipment load, while the intake temperature better reflects the ambient temperature of the computer room and the temperature propagation trend, the temperature sensor is preferentially installed in the intake area. During installation, the specific installation location can be determined according to the rack structure, such as a front door intake structure, a bottom air supply structure, or a side intake structure, to ensure the consistency and comparability of the collected temperature data.

[0019] For example, a data center uses a cold aisle and hot aisle isolation design, with cold air entering from the front door of the server rack and exiting from the rear door. After the system identifies that the server rack adopts a front-to-back airflow pattern, it installs a temperature sensor 10 centimeters above the air inlet on the front door of the rack to continuously monitor the temperature changes of the cold air entering the rack.

[0020] Step S13: Establish a one-to-one binding relationship between the temperature sensor and the cabinet identification sequence, so that the readings of each temperature sensor can be reversed to all network elements in the cabinet through the cabinet identification.

[0021] In one embodiment, after the temperature sensors are installed, a unique sensor identifier is assigned to each sensor, and a mapping table between sensors and cabinets is established in the management system. This mapping table records information such as cabinet identifier, sensor identifier, and sensor communication address. When it is necessary to obtain the ambient temperature corresponding to a certain network element, it is not necessary to directly query the network element itself. Instead, the corresponding temperature sensor is retrieved from the cabinet to which the network element belongs, thereby obtaining the temperature data of the environment where the network element is located. In this way, it is possible to avoid deploying a separate temperature monitoring module on each network element device, reducing system complexity.

[0022] For example, a temperature sensor T03 is installed in rack R03, and switch B is deployed inside rack R03. When the system needs to obtain the ambient temperature of switch B, it first queries the rack identifier R03 corresponding to switch B, then locates the sensor T03 according to the binding relationship, and reads its current temperature value as the temperature information of the environment where switch B is located.

[0023] Furthermore, step S1 also includes the following steps: Step S14: Enable the offset measurement value output register in the precision time protocol stack through the network management interface of each network element. The output register is configured to output the current cumulative offset measurement value before each clock correction event is executed by the precision time protocol stack, and establish a one-to-one binding relationship between the access handle of the output register and the device identifier of the corresponding network element. In one embodiment, the Precision Time Protocol (PTP) function module of each network element is accessed through the network management system to configure its internal offset monitoring mechanism. This allows the device to automatically output the current accumulated time offset measurement result before performing clock synchronization correction. The output data reflects the accumulated deviation between the device's local clock and the reference clock within the current correction cycle. The system saves the access method of the corresponding register for each network element, including the register address, access path, and call interface information, and establishes a correspondence between this information and the device identifier for subsequent automatic acquisition of offset data.

[0024] For example, switch B supports the standard PTP protocol, and its internal register address 0x8201 is used to store the current cumulative offset value. The system enables the periodic output function of this register through the network management interface and records the mapping relationship between "device B - register 0x8201". When it is necessary to read the clock offset information of switch B later, the corresponding data can be obtained directly through this access path.

[0025] Step S15: Identify the service flow ingress network element and service flow egress network element of the network slice in the cabinet identification sequence, traverse all network elements passed by the network slice along the service flow direction, and assign an incremental path position number to each network element according to the traversal order. In one embodiment, to subsequently analyze the accumulation process of temperature disturbances and clock drift according to the service flow propagation direction, the system first identifies the service entry node and service exit node in the network slice, and traverses each network element hop-by-hop along the actual forwarding path, starting from the entry node. During the traversal, a unique and continuously increasing position number is assigned to each network element, which indicates the order of the network element in the service flow path. All subsequent drift calculations, temperature propagation analyses, and end-to-end aggregation operations use this order as a unified reference.

[0026] For example, a network slice passes through network elements A, B, C, and D. The system numbers A as 1, B as 2, C as 3, and D as 4. At this time, A is the ingress network element and D is the egress network element, forming the corresponding path position sequence set {1,2,3,4}.

[0027] Step S16: Combine the device identifier, cabinet identifier, access handle of the offset measurement value output register of each network element with the path bit number into a path record, and sort all path records in ascending order of path bit number to obtain the slice path sequence.

[0028] In one embodiment, after obtaining the device identifier, rack identifier, offset register access information, and path bit sequence number, the above information is integrated into a unified data structure to describe a single network element node on the network slice path. Each data structure completely records the network element's identity information, environmental mapping information, and clock offset acquisition information, thereby providing a unified index for subsequent temperature and offset acquisition. Subsequently, the data structures corresponding to all network elements are sorted in ascending order according to the path bit sequence number, and organized sequentially to form a slice path sequence, which is stored in the control system for subsequent steps to call.

[0029] For example, for switch B in the path, whose device identifier is DEV_B, whose rack is R03, whose offset register access handle is HANDLE_03, and whose path bit number is 2, then a path record is generated: "{DEV_B, R03, HANDLE_03, 2}".

[0030] After all network elements in the path have generated corresponding records, the system arranges them in ascending order according to the path position number, forming: "{DEV_A, R01, HANDLE_01, 1} → {DEV_B, R03, HANDLE_03, 2} → {DEV_C, R05, HANDLE_05, 3} → {DEV_D, R08, HANDLE_08, 4}".

[0031] This ordered set of records constitutes the slice path sequence used in subsequent steps.

[0032] Of particular importance, step S2 includes: The temperature sensor is located based on the cabinet identifier recorded in the path sequence of the slice path, and the offset measurement value output register is located based on the access handle of the offset measurement value output register recorded in the path sequence of the slice path. Within the calibration window, monitor the clock correction actions of the PTP protocol stack of the network element and record the occurrence time of each clock correction action as the alignment time. The current temperature value is read from the temperature sensor based on the alignment time, and the current cumulative offset measurement value is read from the offset measurement value output register. Arrange the current temperature value and current cumulative offset measurement value read at multiple consecutive alignment times in chronological order to obtain the temperature time series and PTP offset series of the corresponding network element; When the calibration window ends, stop listening and reading.

[0033] Furthermore, step S3 includes the following steps: Step S31: Pair the temperature time series with the precise time protocol offset series according to the timestamp to obtain a data point sequence consisting of temperature values ​​and cumulative offset measurements; In one embodiment, the temperature time series and PTP offset series corresponding to the same network element are read, and a correspondence is established based on the timestamps generated during acquisition. Since the temperature value and cumulative offset measurement value in step S2 are acquired synchronously at the same alignment time, the system can directly use the timestamp as the association index to combine the temperature observation result at each time moment with the cumulative offset measurement result at the same time moment to form a sample point. Subsequently, all sample points are organized into a data point sequence according to the chronological order, thereby constructing a basic dataset reflecting the correspondence between changes in ambient temperature and changes in clock offset, providing input data for the subsequent establishment of a temperature influence model.

[0034] For example, data from five alignment moments are acquired within the calibration window, corresponding to temperature values ​​of 25.1℃, 25.8℃, 26.3℃, 27.0℃, and 27.5℃, and cumulative offset measurements of 10ns, 18ns, 25ns, 36ns, and 43ns, respectively. After associating the data points according to the same timestamp, the system forms a data point sequence [(25.1,10), (25.8,18), (26.3,25), (27.0,36), (27.5,43)].

[0035] Step S32: Using the temperature value of each data point in the data point pair sequence as the independent variable and the cumulative offset measurement value of each data point as the dependent variable, perform least squares linear fitting to obtain the slope and intercept of the fitted line. In one embodiment, the system treats temperature values ​​in the data point sequence as explanatory variables and the corresponding cumulative offset measurements as response variables, and uses the least squares method to construct a linear relationship model between them. During the fitting process, a fitting line is found that minimizes the sum of the squares of the vertical distances from all sample points to this line, thereby obtaining a mathematical expression describing the relationship between temperature change and clock offset change. After fitting, the system obtains the slope and intercept of the fitting line, where the slope represents the cumulative offset change trend caused by each unit change in temperature, and the intercept represents the initial offset baseline at the theoretical reference temperature.

[0036] For example, for the data point sequence [(25.1,10), (25.8,18), (26.3,25), (27.0,36), (27.5,43)], the expression obtained after least squares fitting is: Cumulative Offset = 11.8 × Temperature - 286.5. Here, the slope of 11.8 indicates that for every 1°C increase in temperature, the cumulative offset increases by an average of approximately 11.8 nanoseconds.

[0037] Step S33: Substitute the slope and intercept into each data point in the data point sequence, calculate the predicted cumulative offset measurement, subtract the predicted cumulative offset measurement from the actual cumulative offset measurement to obtain the fitting residual of the data point pair, and form the fitting residual sequence. In one embodiment, after obtaining the fitted model, the theoretically corresponding offset value is recalculated for each temperature sample in the data point sequence using the fitted expression, thereby obtaining the predicted cumulative offset measurement value. The predicted value is then compared with the actual observed value, and the difference between the two measures the degree to which the sample point deviates from the overall fitted trend. For all sample points, the corresponding differences are calculated sequentially, and a fitted residual sequence is formed in chronological order. This residual sequence is used to characterize the degree to which each sample conforms to the overall temperature response pattern, providing a basis for subsequent identification of abnormal samples.

[0038] For example, when the temperature is 27.0℃, the fitted model predicts a corresponding offset of 32.1 ns, while the actual observed value is 36 ns, so the fitting residual for this sample is 3.9 ns; when the temperature is 25.8℃, the predicted value is 17.9 ns, while the actual value is 18 ns, so the corresponding residual is 0.1 ns. The residuals for each sample are recorded sequentially to form a residual sequence.

[0039] Step S34: Calculate the standard deviation of the fitted residual sequence, and remove data points whose absolute value of the fitted residual is greater than the product of the preset residual multiple and the standard deviation from the data point sequence to obtain the converged data point sequence; In one embodiment, a dispersion analysis is performed on the fitted residual sequence. The standard deviation of all residuals is used to characterize the range of fluctuation of normal samples around the fitted model. Then, an anomaly detection boundary is constructed based on a pre-set residual multiple, and data points exceeding this boundary are identified as anomalous samples. Anomalous samples typically originate from factors such as temporary network jitter, device restarts, synchronization link disturbances, or instantaneous sensor anomalies. Their variation patterns are not directly related to the temperature influence mechanism, and therefore need to be removed from the modeling data. After anomaly removal, the remaining samples constitute a convergent data point sequence.

[0040] For example, if the calculated residual standard deviation is 2 ns and the preset residual factor is set to 3, then the anomaly detection threshold is 6 ns. When the residual corresponding to a data point is 9 ns, it is identified as an anomaly sample and deleted because it exceeds 6 ns; while data points with a residual of 4 ns are retained.

[0041] Step S35: Re-execute least squares linear fitting and residual abnormality removal with the converged data point sequence until the number of data points removed in this round is zero or the number of removal iterations reaches the preset iteration limit. Use the slope obtained from the last round of fitting as the temperature sensitivity coefficient of the network element.

[0042] In one embodiment, the converged data point sequence is used as the new modeling dataset, and the fitting calculation, residual calculation, and outlier identification process are re-executed. As outliers are continuously removed, the fitted model gradually stabilizes, and the residual distribution gradually converges. When no new outliers are found after a certain round of processing, or when the preset maximum number of iterations has been reached, the iteration process is terminated, and the slope in the final fitted model is extracted as the temperature sensitivity coefficient of the network element. This parameter is used to quantify the impact of temperature changes on the clock offset of the network element and is the core basis for subsequent prediction of timing drift calculation.

[0043] For example, in the first round of fitting, 5 outlier samples were found and removed; in the second round of fitting, 1 more outlier sample was found; in the third round of fitting, no new outlier samples were found, so the system ended the iteration. The final fitting expression is: Cumulative offset = 12.4 × temperature - 301.7. At this point, the slope of 12.4 is recorded as the temperature sensitivity coefficient of the network element.

[0044] See Figure 2 The synchronously acquired temperature values ​​and PTP cumulative offset values ​​are used to form a data point sequence, and the least squares method is used for linear fitting to obtain the fitting relationship between temperature and offset. The slope of the fitted line in the figure is 12.4 ns / ℃, which corresponds to the temperature sensitivity coefficient of the network element. Then, the residual of each data point relative to the fitted line is calculated, and data points with residuals exceeding a preset threshold are removed as outliers. Finally, a stable temperature sensitivity coefficient is obtained for subsequent prediction of time series drift calculation.

[0045] Furthermore, step S4 includes the following steps: Step S41: Extract the pairwise combination of each pair of network elements from the slice path sequence according to the path position number. For each pair of network elements, extract the temperature time series of the cabinet where the preceding network element is located and the temperature time series of the cabinet where the following network element is located. In one embodiment, after obtaining the slice path sequence, the path position number is used as a unified sorting basis to establish an analysis object set for the network elements in the path according to their chronological order. To identify the propagation relationship of temperature disturbances between different network elements, the system does not only analyze adjacent network elements, but also constructs all possible combinations of preceding and following network elements in chronological order. For each combination, the corresponding temperature sensor is located based on the cabinet identifier corresponding to the network element, and the temperature time series formed during the previous calibration phase is read. In this way, the correspondence between the temperature change processes of multiple network elements can be obtained, providing basic data for subsequent determination of the temperature propagation direction and propagation delay.

[0046] For example, if a slice path contains network element A, network element B, network element C, and network element D, then multiple network element pairs such as (A,B), (A,C), (A,D), (B,C), (B,D), and (C,D) can be constructed. For the network element pair (A,C), the temperature time series corresponding to cabinet A and the temperature time series corresponding to cabinet C are extracted respectively, and these two series are used as inputs for subsequent analysis.

[0047] Step S42: Within the preset time offset interval, perform normalized cross-correlation calculation on the two temperature time series to obtain a cross-correlation function sequence indexed by the time offset; In one embodiment, for each pair of network elements, normalized cross-correlation analysis is performed on the two temperature time series within a preset time offset range. During the analysis, one of the temperature series is progressively shifted along the time axis, and the similarity between the two series is calculated at each shift position. Due to the normalization process, the influence of differences in the absolute values ​​and fluctuation amplitudes of temperatures from different cabinets can be eliminated, ensuring that the analysis results primarily reflect the consistency of temperature change trends. As the time offset changes, a set of corresponding correlation values ​​is obtained, thus forming a cross-correlation function sequence.

[0048] For example, the temperature time series of both racks A and C show periodic fluctuations caused by the daily start and stop of the air conditioning. The analysis range was set to ±120 minutes, with a calculation step of 1 minute. The correlation coefficient was 0.62 when the time offset was 0 minutes; 0.81 when the offset was 20 minutes; 0.95 when the offset was 35 minutes; and decreased to 0.70 when the offset was 60 minutes. The corresponding cross-correlation function sequence was then generated.

[0049] Step S43: Read the global maximum value and its corresponding time offset of the cross-correlation function sequence within the preset time offset interval. Use the global maximum value as the temperature field coupling strength between the preceding and following network elements, and use the time offset as the temperature phase delay between the preceding and following network elements.

[0050] In one embodiment, the cross-correlation function sequence is iterated to find the maximum correlation value across the entire analysis range, and the time offset corresponding to the occurrence of this maximum value is recorded. The maximum correlation value reflects the consistency of the temperature change patterns of the two racks, and therefore can be used to measure the strength of their influence by the same temperature field; while the time offset for reaching the maximum correlation reflects the time difference experienced by the temperature disturbance from the preceding rack to the following rack. Both are extracted and used as important parameters for subsequent temperature field modeling.

[0051] For example, for the cross-correlation function sequence corresponding to the network element pair (A,C), it was found that the correlation coefficient reaches its maximum value of 0.95 when the time offset is 35 minutes. Therefore, 0.95 is recorded as the temperature field coupling strength of the network element pair, and 35 minutes is recorded as the temperature phase delay of the network element pair.

[0052] Furthermore, step S4 also includes the following steps: Step S44: Compare the temperature field coupling strength with a preset coupling threshold. If the temperature field coupling strength is greater than the preset coupling threshold, mark the preceding and subsequent network elements as the same temperature field subdomain; otherwise, mark the preceding and subsequent network elements as different temperature field subdomains. In one embodiment, the temperature field coupling strength is used to determine whether different network elements are located in the same thermal environment influence area. When the temperature change trends of two network elements are highly consistent, it indicates that they are likely to be jointly affected by the same cooling airflow path, the same regional air conditioning system, or the same local environment of the computer room, and therefore can be classified into the same temperature field subdomain; otherwise, they are considered to be affected by different temperature fields. The division and marking of temperature field subdomains are completed by comparing the temperature field coupling strength with a pre-set coupling threshold.

[0053] For example, the preset coupling threshold is set to 0.85. When the temperature field coupling strength between network element A and network element C is 0.95, the system marks them as the same temperature field subdomain; while when the temperature field coupling strength between network element A and network element D is only 0.41, they are marked as different temperature field subdomains.

[0054] Step S45: Establish a set of network element-level entries. For each network element in the slice path sequence, generate network element-level entries, where each network element-level entry contains the device identifier and corresponding temperature sensitivity coefficient of the network element. In one embodiment, a set of network element-level information is established to describe the temperature response characteristics of individual network elements. For each network element in the slice path, its device identifier and the temperature sensitivity coefficient obtained in previous steps are read, and the two pieces of information are encapsulated into a network element-level entry in a uniform format. This entry is used to describe the clock drift sensitivity exhibited by the network element when affected by temperature changes, thereby providing basic parameters for subsequent drift prediction.

[0055] For example, network element A is calculated to have a temperature sensitivity coefficient of 11.8 ns / ℃, while network element B has a temperature sensitivity coefficient of 7.6 ns / ℃. The system generates the following entries: Network element A: {DEV_A, 11.8} Network element B: {DEV_B, 7.6} Then all network element entries are stored in a unified network element-level entry set.

[0056] It should be noted that this set mainly describes the temperature response capability of each network element itself, and therefore does not include information on the relationship between network elements.

[0057] Step S46: Establish a set of network element pair level entries. For each network element pair, generate a network element pair level entry. Each network element pair level entry includes the device identifier of the preceding network element, the device identifier of the following network element, the temperature phase delay between the network element pairs, the temperature field coupling strength, and the temperature field subdomain label. In one embodiment, a set of network element pair-level information is established to describe the temperature propagation relationship between network elements. For each network element pair obtained from the previous analysis, the preceding network element identifier, the following network element identifier, the temperature phase delay, the temperature field coupling strength, and the temperature field subdomain label are uniformly encapsulated to form a corresponding network element pair-level entry. This entry is used to record the propagation characteristics of temperature disturbances between different network elements and their environmental correlations, thereby supporting subsequent temperature propagation path analysis and phase-sensitive prediction calculations.

[0058] For example, for a network element pair (A,C), if a temperature phase delay of 35 minutes, a temperature field coupling strength of 0.95, and the temperature field subdomain is marked as co-domain, then the following entry is generated: {DEV_A, DEV_C, 35min, 0.95, same domain} For the network element pair (B,D), if the corresponding parameters are 50 minutes, 0.42, and different domains, then the following is generated: {DEV_B, DEV_D, 50min, 0.42, foreign} And uniformly store them in the network element level entry set.

[0059] It should be noted that a network element can appear in multiple network element pair entries at the same time, because it may have different degrees of temperature propagation relationship with multiple subsequent network elements.

[0060] Step S47: Combine the network element-level entry set with the network element pair-level entry set to form a slice temperature response signature.

[0061] In one embodiment, a set of network element-level entries describing the temperature response capability of each network element and a set of network element pair-level entries describing the temperature propagation relationship between network elements are integrated to construct a temperature response feature model for the network slice. This model includes both the sensitivity of each network element to temperature changes and the propagation law of temperature disturbances in the slice path, thus reflecting both the local and global characteristics of temperature impact. This integrated feature model is defined as the slice temperature response signature and serves as the basis for subsequent prediction of time-series drift and execution of compensation control.

[0062] For example, the set of network element-level entries corresponding to a certain slice includes: {DEV_A, 11.8} {DEV_B, 7.6} {DEV_C, 10.2} The set of network element-level entries includes: {DEV_A, DEV_B, 12min, 0.91, same domain} {DEV_A, DEV_C, 35min, 0.95, same domain} {DEV_B, DEV_C, 18min, 0.88, same domain} The two pieces of information are combined and stored to form the temperature response signature corresponding to the slice.

[0063] Furthermore, step S5 includes the following steps: Step 51: Based on the temperature field subdomain markers of each network element pair in the slice temperature response signature, the slice path sequence is divided into several subdomain path segments; wherein the temperature field subdomain markers between any two network elements within a subdomain path segment all point to the same temperature field subdomain. In one embodiment, all element-level entries in the slice temperature response signature are read, and a temperature correlation graph is established based on the temperature field subdomain labels recorded therein. Elements labeled as belonging to the same temperature field subdomain are considered to be in the same thermal environment propagation system, and temperature disturbances exhibit continuous propagation characteristics among these elements. Elements labeled as belonging to different temperature field subdomains are considered to be affected by different thermal environment factors. Subsequently, the slice path sequence is traversed, and elements continuously belonging to the same temperature field subdomain are grouped into the same subdomain path segment according to the temperature field subdomain affiliation, thereby splitting the originally complete slice path into multiple independent temperature propagation analysis units. This process avoids forcibly including elements without temperature propagation correlations into the same prediction model, improving the accuracy of subsequent drift predictions.

[0064] For example, a slice path contains network elements A, B, C, D, E, and F. A, B, and C are all marked as the same temperature field subdomain, while D, E, and F are all marked as another temperature field subdomain. In this case, the slice path is split into subdomain path segments ABC and DEF, and then each is analyzed separately.

[0065] Step S52: Within the subdomain path segment, based on the temperature phase delay of each network element pair in the slice temperature response signature, extract the network element whose temperature phase delay relative to other network elements is non-positive as the temperature phase leading network element of the segment, and the remaining network elements as the temperature phase lagging network elements. In one embodiment, for each subdomain path segment, the temperature phase delay parameters corresponding to all network element pairs are read, and the temperature propagation timing relationship within that subdomain is constructed. Since the temperature phase delay reflects the propagation order of temperature disturbances among different network elements, the location where the temperature change first occurs can be determined based on the phase relationship of each network element relative to other network elements. Specifically, if the temperature phase delay of a network element relative to other network elements within the same subdomain is not greater than zero, it indicates that the temperature change of that network element occurred first or occurred synchronously with other network elements, and is therefore identified as a temperature phase leading network element; the remaining network elements are considered temperature phase lagging network elements.

[0066] For example, in a certain subdomain path segment, the temperature phase delay of network element A relative to network element B is -8 minutes, and its temperature phase delay relative to network element C is -15 minutes, while the temperature phase delay of network element B relative to network element C is -7 minutes. Analysis reveals that network element A is leading relative to all other network elements. Therefore, network element A is identified as the leading network element in temperature phase for this subdomain, while network elements B and C are identified as lagging network elements.

[0067] Step S53: Take the current temperature value of the temperature phase leading network element as the effective temperature of the network element.

[0068] In one embodiment, the temperature phase-leading network element is used as a reference node for the current temperature field change trend, and the temperature value corresponding to the current moment of the network element is directly read. Since this network element is the first to sense the temperature disturbance, its current temperature can reflect the latest temperature change state in the subdomain. This temperature value is recorded as the effective temperature of the network element and used as the temperature input parameter for subsequent prediction of time series drift calculation.

[0069] For example, at the current time 12:00, the temperature of the cabinet where the temperature phase-leading network element A is located is 29.3℃. Then, 29.3℃ is directly recorded as the effective temperature of network element A and stored in the subsequent drift prediction process.

[0070] Furthermore, step S5 also includes the following steps: Step S54: For the temperature phase lagging network element, read a temperature value from the temperature time series of the temperature phase leading network element. The read temperature value is used as the effective temperature of the temperature phase lagging network element. The time corresponding to the read temperature value is the time corresponding to the backtracking time of the current time. The backtracking time is the temperature phase delay of the temperature phase lagging network element relative to the temperature phase leading network element of that time. In one embodiment, for temperature-phase-lagging network elements within the same subdomain, their current temperature value is not directly used for drift prediction. Instead, their corresponding effective temperature is derived using temperature propagation relationships. Specifically, based on the temperature phase delay of the network element relative to the temperature-phase-leading network element, the time required for a temperature disturbance to propagate from the leading network element to the lagging network element is determined. Then, the corresponding time length is traced back from the temperature time series stored in the leading network element, and the temperature value at that time is read as the effective temperature of the lagging network element. In this way, the temporal state of different network elements actually affected by temperature can be reconstructed using temperature propagation laws.

[0071] For example, if the current time is 12:00, the temperature phase of network element A is 29.3℃ ahead of its current temperature, and the temperature phase of network element B is 20 minutes behind that of network element A, then the temperature value corresponding to 11:40 is read from the temperature time series of network element A. If this temperature value is 27.8℃, then 27.8℃ is taken as the effective temperature of network element B, instead of using the temperature value currently measured by network element B.

[0072] Step S55: Multiply the temperature sensitivity coefficient of the network element by the effective temperature of the network element, and then multiply the product by the precise time protocol calibration cycle duration to obtain the predicted time series drift; In one embodiment, the temperature sensitivity coefficient and effective temperature of each network element are read, and the growth trend of the unit time offset under the current temperature conditions is estimated based on their product. Then, combined with the calibration cycle length of the precise time protocol, the cumulative timing drift expected to occur in the network element before the next calibration is calculated. In this way, environmental temperature factors can be directly converted into future clock offset changes, achieving a mapping from temperature monitoring to drift prediction.

[0073] For example, if network element A has a temperature sensitivity coefficient of 12 ns / ℃, an effective temperature of 29.3℃, and a PTP calibration period of 1 second, then the system calculates the predicted timing drift as: 12 × 29.3 × 1 = 351.6 ns. Therefore, it is expected that this network element may experience a timing drift of approximately 351.6 nanoseconds in the next calibration period.

[0074] Step S56: Perform algebraic summation of the prediction time drift of all network elements to obtain the end-to-end prediction drift; In one embodiment, the predicted timing drifts corresponding to all network elements in the slice path are collected uniformly and accumulated according to the slice path direction. Since service packets pass through multiple network elements continuously during transmission, the timing drift generated by each network element will affect the final end-to-end synchronization accuracy. By algebraically summing all predicted timing drifts, the total clock offset expected to occur in the next calibration cycle for the entire slice path is obtained.

[0075] For example, if the predicted timing drifts for network elements A, B, C, and D are 350ns, 180ns, 220ns, and 150ns respectively, then the system calculates: end-to-end predicted drift = 350 + 180 + 220 + 150 = 900ns. Therefore, the entire network slice is expected to generate approximately 900 nanoseconds of end-to-end clock drift in the next calibration cycle.

[0076] Step S57: Invert the end-to-end prediction drift as the time offset compensation amount, and apply the time offset compensation amount to the message transmission time of the last network element in the slice path sequence; In one embodiment, a pre-compensation control quantity is generated based on the end-to-end predicted drift result. Since the predicted drift represents the possible direction and magnitude of future offsets, a compensation quantity with the same magnitude but opposite direction is generated to preemptively offset the impending clock skew. Subsequently, this compensation quantity is loaded into the message transmission scheduling module of the last network element in the slice path. By adjusting the actual message transmission time, the future drift and the compensation quantity cancel each other out, thereby reducing the end-to-end synchronization error.

[0077] For example, when the predicted end-to-end drift is +900ns, a time offset compensation of -900ns is generated, and the transmission time is advanced by 900 nanoseconds before the egress network element sends the message, thereby offsetting the expected synchronization error.

[0078] Step S58: Compare the absolute value of the end-to-end prediction drift with the preset SLA threshold. When the absolute value of the end-to-end prediction drift is greater than the preset SLA threshold, issue a cross-domain coordination command to the cross-domain control interface associated with the network slice.

[0079] In one embodiment, the changes in end-to-end prediction drift are continuously monitored, and its absolute value is compared with a pre-set service level agreement (SLA) threshold. When the prediction drift remains within an acceptable range, adjustment is made solely by a time compensation mechanism. When the prediction drift exceeds the threshold, it is considered that simple time compensation is insufficient to guarantee service synchronization accuracy, and a higher-level resource coordination mechanism needs to be initiated. At this time, cross-domain coordination instructions are sent to relevant controllers through a cross-domain control interface to coordinate multiple dimensions such as network resources, service traffic, and slice paths to participate in the adjustment, thereby preventing the continuous expansion of temperature disturbances from causing a decline in service quality.

[0080] For example, a time-sensitive service specifies an SLA threshold of 1000ns. When the calculated end-to-end predicted drift reaches 1500ns, a cross-domain coordination process is immediately triggered, and one or more of the following commands are sent to the corresponding controller: a protection bandwidth extension command, a rate rollback command, or a path switching command, in order to reduce the subsequent drift growth rate.

[0081] Of particular importance, the cross-domain collaboration instructions include one or more of the following three types of instructions: The protection bandwidth extension instruction includes the device identifier of the target time-sensitive bridge, the slice identifier of the target network slice, and the protection bandwidth extension amount. The protection bandwidth extension amount is equal to the absolute value of the end-to-end predicted drift. The target time-sensitive bridge is the time-sensitive bridge connected downstream of the last network element in the slice path sequence. The rate rollback instruction includes the slice identifier of the target network slice, the device identifier of the target service source, and the rate rollback ratio. The target service source is the upstream service source of the first network element in the slice path sequence. The rate rollback ratio is positively correlated with the absolute value of the end-to-end prediction drift. The path switching command includes the slice identifier of the target network slice, the identifier of the current slice path sequence, and the identifier of the backup slice path sequence; the backup slice path sequence is a pre-configured sequence of alternative network elements that are physically separated from the cabinets traversed by the current slice path sequence. The cross-domain collaboration command is issued to the cross-domain control interface associated with the network slice. The cross-domain control interface is one or more of the time-sensitive network management controller, slice service source management controller, or slice path orchestration controller. The recipient of the cross-domain collaborative instruction executes the corresponding action according to the device identifier and parameters carried in the instruction. The corresponding action is actively canceled by the cross-domain control interface after the absolute value of the end-to-end predicted drift falls below the preset SLA threshold.

[0082] See Figure 3 The effective temperature of each network element is determined based on the slice temperature response signature, and the predicted timing drift is calculated in conjunction with the corresponding temperature sensitivity coefficient. The predicted timing drift of each network element is then algebraically accumulated to obtain the end-to-end predicted drift, and a time offset compensation amount in the opposite direction is generated and applied to the egress network element. When the end-to-end predicted drift exceeds a preset SLA threshold, one or more coordinated control measures among protection bandwidth extension, rate backoff, and path switching are triggered through the cross-domain control interface.

[0083] See Figure 4 The present invention also provides a network slicing control system 100 for executing the network slicing control method described above, the network slicing control system 100 comprising: The deployment numbering module 101 is used to deploy temperature sensors in the cabinets of each network element that the network slice passes through, enable the precise time protocol offset measurement output port on each network element, and sequentially number the network elements according to the direction of service flow to form a slice path sequence. The synchronous acquisition module 102 is used to synchronously acquire the temperature time series of the cabinet where each network element is located and the precise time protocol offset sequence of the network element based on the slice path sequence within the calibration window. The coefficient fitting module 103 is used to perform linear fitting with the temperature time series as the independent variable and the precise time protocol offset series as the dependent variable to calculate the temperature sensitivity coefficient of each network element. The signature construction module 104 is used to cross-correlate the temperature time series of any two network elements in the slice path sequence, and take the time offset corresponding to the maximum value of the cross-correlation function as the temperature phase delay between the two network elements; the slice temperature response signature is composed of the temperature sensitivity coefficient and the temperature phase delay. The aggregation compensation module 105 is used to calculate the predicted time drift of each network element based on the slice temperature response signature, and perform phase-sensitive aggregation along the slice path to obtain the end-to-end predicted drift; the end-to-end predicted drift is inverted as the time offset compensation amount, and cross-domain collaboration is triggered when the end-to-end predicted drift exceeds the preset service level protocol threshold.

[0084] The present invention also provides a computer medium storing a computer program, wherein when the computer program is executed, it implements the network slicing control method described in any one of the above claims.

[0085] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0086] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A network slicing control method, characterized in that, Includes the following steps: Step S1: Deploy temperature sensors in the racks of the network elements that the network slice passes through, enable the precise time protocol offset measurement output port on each network element, and number the network elements sequentially according to the direction of service flow to form a slice path sequence. Step S2: Within the calibration window, based on the slice path sequence, synchronously collect the temperature time series of the cabinet where each network element is located and the precise time protocol offset sequence of that network element. Step S3: Perform linear fitting with the temperature time series as the independent variable and the precise time protocol offset series as the dependent variable to calculate the temperature sensitivity coefficient of each network element; Step S4: Perform cross-correlation on the temperature time series of any two network elements in the slice path sequence, and take the time offset corresponding to the maximum value of the cross-correlation function as the temperature phase delay between the two network elements; construct the slice temperature response signature by combining the temperature sensitivity coefficient and the temperature phase delay. Step S5: Calculate the predicted time drift of each network element based on the slice temperature response signature, and perform phase-sensitive aggregation along the slice path to obtain the end-to-end predicted drift; invert the end-to-end predicted drift as the time offset compensation amount, and trigger cross-domain collaboration when the end-to-end predicted drift exceeds the preset service level protocol threshold.

2. The network slicing control method according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Identify each network element traversed by the network slice and record it as the rack identifier sequence of the rack where each network element is located; Step S12: Determine the cooling airflow inlet direction of the cabinet in the cabinet identification sequence, and install a temperature sensor at the inlet position along the cooling airflow inlet direction on the cabinet; Step S13: Establish a one-to-one binding relationship between the temperature sensor and the cabinet identification sequence, so that the readings of each temperature sensor can be reversed to all network elements in the cabinet through the cabinet identification.

3. The network slicing control method according to claim 2, characterized in that, Step S1 also includes the following steps: Step S14: Enable the offset measurement value output register in the precision time protocol stack through the network management interface of each network element. The output register is configured to output the current cumulative offset measurement value before each clock correction event is executed by the precision time protocol stack, and establish a one-to-one binding relationship between the access handle of the output register and the device identifier of the corresponding network element. Step S15: Identify the service flow ingress network element and service flow egress network element of the network slice in the cabinet identification sequence, traverse all network elements passed by the network slice along the service flow direction, and assign an incremental path position number to each network element according to the traversal order. Step S16: Combine the device identifier, cabinet identifier, access handle of the offset measurement value output register of each network element with the path bit number into a path record, and arrange all path records in ascending order of path bit number to obtain the slice path sequence.

4. The network slicing control method according to claim 3, characterized in that, Step S3 includes the following steps: Step S31: Pair the temperature time series with the precise time protocol offset series according to the timestamp to obtain a data point sequence consisting of temperature values ​​and cumulative offset measurements; Step S32: Using the temperature value of each data point in the data point pair sequence as the independent variable and the cumulative offset measurement value of each data point as the dependent variable, perform least squares linear fitting to obtain the slope and intercept of the fitted line. Step S33: Substitute the slope and intercept into each data point in the data point sequence, calculate the predicted cumulative offset measurement, subtract the predicted cumulative offset measurement from the actual cumulative offset measurement to obtain the fitting residual of the data point pair, and form the fitting residual sequence. Step S34: Calculate the standard deviation of the fitted residual sequence, and remove data points whose absolute value of the fitted residual is greater than the product of the preset residual multiple and the standard deviation from the data point sequence to obtain the converged data point sequence; Step S35: Re-execute least squares linear fitting and residual abnormality removal with the converged data point sequence until the number of data points removed in this round is zero or the number of removal iterations reaches the preset iteration limit. Use the slope obtained from the last round of fitting as the temperature sensitivity coefficient of the network element.

5. The network slicing control method according to claim 4, characterized in that, Step S4 includes the following steps: Step S41: Extract the pairwise combination of each pair of network elements from the slice path sequence according to the path position number. For each pair of network elements, extract the temperature time series of the cabinet where the preceding network element is located and the temperature time series of the cabinet where the following network element is located. Step S42: Within the preset time offset interval, perform normalized cross-correlation calculation on the two temperature time series to obtain a cross-correlation function sequence indexed by the time offset; Step S43: Read the global maximum value and its corresponding time offset of the cross-correlation function sequence within the preset time offset interval. Use the global maximum value as the temperature field coupling strength between the preceding and following network elements, and use the time offset as the temperature phase delay between the preceding and following network elements.

6. The network slicing control method according to claim 5, characterized in that, Step S4 also includes the following steps: Step S44: Compare the temperature field coupling strength with a preset coupling threshold. If the temperature field coupling strength is greater than the preset coupling threshold, mark the preceding and subsequent network elements as the same temperature field subdomain; otherwise, mark the preceding and subsequent network elements as different temperature field subdomains. Step S45: Establish a set of network element-level entries. For each network element in the slice path sequence, generate network element-level entries, where each network element-level entry contains the device identifier and corresponding temperature sensitivity coefficient of the network element. Step S46: Establish a set of network element pair level entries. For each network element pair, generate a network element pair level entry. Each network element pair level entry includes the device identifier of the preceding network element, the device identifier of the following network element, the temperature phase delay between the network element pairs, the temperature field coupling strength, and the temperature field subdomain label. Step S47: Combine the network element-level entry set with the network element pair-level entry set to form a slice temperature response signature.

7. The network slicing control method according to claim 6, characterized in that, Step S5 includes the following steps: Step 51: Based on the temperature field subdomain markers of each network element pair in the slice temperature response signature, the slice path sequence is divided into several subdomain path segments; wherein the temperature field subdomain markers between any two network elements within a subdomain path segment all point to the same temperature field subdomain. Step S52: Within the subdomain path segment, based on the temperature phase delay of each network element pair in the slice temperature response signature, extract the network element whose temperature phase delay relative to other network elements is non-positive as the temperature phase leading network element of the segment, and the remaining network elements as the temperature phase lagging network elements. Step S53: Take the current temperature value of the temperature phase leading network element as the effective temperature of the network element.

8. The network slicing control method according to claim 7, characterized in that, Step S5 also includes the following steps: Step S54: For the temperature phase lagging network element, read a temperature value from the temperature time series of the temperature phase leading network element. The read temperature value is used as the effective temperature of the temperature phase lagging network element. The time corresponding to the read temperature value is the time corresponding to the backtracking time of the current time. The backtracking time is the temperature phase delay of the temperature phase lagging network element relative to the temperature phase leading network element of that time. Step S55: Multiply the temperature sensitivity coefficient of the network element by the effective temperature of the network element, and then multiply the product by the precise time protocol calibration cycle duration to obtain the predicted time series drift; Step S56: Perform algebraic summation of the prediction time drift of all network elements to obtain the end-to-end prediction drift; Step S57: Invert the end-to-end prediction drift as the time offset compensation amount, and apply the time offset compensation amount to the message transmission time of the last network element in the slice path sequence; Step S58: Compare the absolute value of the end-to-end prediction drift with the preset SLA threshold. When the absolute value of the end-to-end prediction drift is greater than the preset SLA threshold, issue a cross-domain coordination command to the cross-domain control interface associated with the network slice.

9. A network slicing control system, characterized in that, For executing the network slicing control method as described in claim 1, the network slicing control system includes: The deployment numbering module is used to deploy temperature sensors in the cabinets of each network element that the network slice passes through, enable the precise time protocol offset measurement output port on each network element, and sequentially number the network elements according to the direction of service flow to form a slice path sequence. The synchronous acquisition module is used to synchronously acquire the temperature time series of the cabinet where each network element is located and the precise time protocol offset sequence of the network element based on the slice path sequence within the calibration window. The coefficient fitting module is used to perform linear fitting with the temperature time series as the independent variable and the precise time protocol offset series as the dependent variable to calculate the temperature sensitivity coefficient of each network element. The signature construction module is used to cross-correlate the temperature time series of any two network elements in the slice path sequence, and take the time offset corresponding to the maximum value of the cross-correlation function as the temperature phase delay between the two network elements; the temperature sensitivity coefficient and the temperature phase delay constitute the slice temperature response signature. The aggregation compensation module is used to calculate the predicted time drift of each network element based on the slice temperature response signature, and perform phase-sensitive aggregation along the slice path to obtain the end-to-end predicted drift; the end-to-end predicted drift is inverted as the time offset compensation amount, and cross-domain collaboration is triggered when the end-to-end predicted drift exceeds the preset service level protocol threshold.

10. A computer medium storing a computer program, characterized in that, When the computer program is executed, it implements the network slicing control method as described in any one of claims 1 to 8.

Images