Time sequence data management method of edge computing gateway

Abstract

The invention discloses a time sequence data management method of an edge computing gateway, which relates to the technical field of edge computing and industrial Internet of Things, and comprises the following steps of: respectively recording a communication bandwidth occupancy rate, a buffer area residual rate and a scheduling thread occupancy rate of the edge computing gateway in a preset fixed time period; and constructing a resource use original data matrix covering all time points in the fixed time period. According to the method, by periodically monitoring the resource use state and fusing the high-priority task scheduling performance, the scheduling resource abnormal occupancy index is dynamically generated, and intelligent sensing and scheduling optimization of the edge computing gateway on the resource pressure are achieved. When the abnormal index is increased, the system automatically triggers buffer area redistribution and low-optimal task data compression, data writing and scheduling real-time performance of key tasks are guaranteed preferentially, the problems of task starvation and data loss are effectively avoided, and the stability and the response capability of the system in a high-pressure environment are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of edge computing and industrial Internet of Things, and particularly relates to a time series data management method of an edge computing gateway. BACKGROUND

[0002] The time series data management of the edge computing gateway refers to a whole process management method of localized and efficient storage, organization, processing and scheduling of time-stamped continuous data (i.e. time series data) collected from various sensors, control devices or terminal systems at the gateway side in the edge computing architecture. The method not only supports large-scale, high-frequency data writing and low-latency data reading, but also needs to combine the edge computing capability to perform real-time preprocessing, cleaning, compression and aggregation on the data, improve the data utilization efficiency, and dynamically determine when to upload the data to the cloud or the central system according to the network state, business strategy or computing load, so as to realize the data autonomy, bandwidth saving and response acceleration of the edge end, and is widely applied to industrial sites, intelligent manufacturing, energy monitoring, traffic perception and other scenes.

[0003] The prior art has the following disadvantages: the existing edge computing gateway has the problems of rigid scheduling strategy and lack of dynamic perception of resource allocation mechanism in the scene of multi-task concurrency and priority queue mixed scheduling, which specifically manifests as: when a low-priority task (such as periodic data uploading and background data cleaning) continuously occupies core scheduling resources (including but not limited to communication bandwidth, buffer space and scheduling threads) in a certain time period, it may cause that a high-priority critical time series task (such as safety alarm information and device abnormal event) cannot obtain timely scheduling permission, thereby causing scheduling delay or even "task starvation" phenomenon. The problem is difficult to be exposed in the conventional system test and function verification stage, because the test environment is often sufficient in bandwidth resources and low in task density, and it is difficult to simulate the working conditions of multi-source data burst concurrency or bandwidth limitation in actual operation. Once the above scheduling resources are continuously occupied, the critical event data cannot be uploaded to the platform in the first time, thereby causing the failure of the early warning mechanism, the lag of the safety control response, and the serious production interruption, device damage and even personal injury and other irreversible consequences.

[0004] The above information disclosed in the background section is only intended to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute the prior art known to those of ordinary skill in the art. SUMMARY

[0005] The purpose of the present application is to provide a time series data management method of an edge computing gateway to solve the problems in the background.

[0006] In order to achieve the above purpose, the present application provides the following technical scheme: a time series data management method of an edge computing gateway, comprising the following steps:

[0007] In a preset fixed time period, the communication bandwidth occupancy rate, the buffer remaining rate and the scheduling thread occupancy rate of the edge computing gateway are recorded respectively to construct a resource usage raw data matrix covering all time points in the fixed time period;

[0008] Based on the resource usage raw data matrix, the usage fluctuation slope and the resource occupancy density of the communication bandwidth occupancy rate, the buffer remaining rate and the scheduling thread occupancy rate in the fixed time period are calculated respectively, and a weighting factor is configured according to the task priority level corresponding to each resource index to fuse the resource index to generate a single resource anomaly representation value;

[0009] In the fixed time period, the execution state information of all tasks set to high priority is extracted, the number of tasks whose waiting time exceeds a preset response threshold is identified, and the cumulative blocking time is counted to generate a high-priority task blocking coefficient;

[0010] The resource anomaly representation value and the high-priority task blocking coefficient are input into a nonlinear function for mapping and fusion to obtain a scheduling resource anomaly occupancy index of the current scheduling period as an anomaly perception index of the dynamic scheduling system;

[0011] When the scheduling resource anomaly occupancy index exceeds a set threshold, a buffer resource reallocation mechanism is triggered to dynamically adjust the buffer quota proportion corresponding to tasks of different priorities to ensure the continuity of data writing and uploading operations of critical tasks; at the same time, compression processing is performed on data segments of low-priority tasks, and only statistical summary information is retained to reduce the buffer resource pressure.

[0012] Preferably, in a preset fixed time period, the communication bandwidth occupancy rate, the buffer remaining rate and the scheduling thread occupancy rate of the edge computing gateway are recorded respectively to construct a resource usage raw data matrix as follows:

[0013] For the communication behavior of the edge computing gateway during operation, a fixed sampling time interval is set, the real-time data transmission rate of the communication channel is continuously sampled, and the communication bandwidth occupancy rate sequence is calculated by the ratio of the sampling point to the theoretical maximum bandwidth;

[0014] At time points synchronized with communication sampling, the buffer usage state allocated to each task type in the edge computing gateway is periodically read, the ratio of the buffer remaining capacity to the total capacity at each sampling time is calculated, and a buffer remaining rate sequence is formed;

[0015] The occupation ratio between the active number of the synchronization statistical edge computing gateway internal task scheduling thread and the total thread pool capacity is obtained, a scheduling thread occupation rate sequence is obtained, and the communication bandwidth occupation rate sequence, the buffer remaining rate sequence and the scheduling thread occupation rate sequence are time-aligned, and a resource use original data matrix is constructed according to a unified timestamp.

[0016] Preferably, based on the resource use original data matrix, the specific steps of fusing the resource indicators to generate a single resource anomaly representation value are as follows:

[0017] The time sequence occupation data corresponding to the communication bandwidth occupation rate, the buffer remaining rate and the scheduling thread occupation rate in the resource use original data matrix are processed, a continuous change model is constructed in a fixed time period by using curve fitting, and the derivative mean is calculated by taking the derivative of the fitting curve at each sampling point, to obtain the use fluctuation slope of each resource type;

[0018] Based on the use fluctuation slope calculation, the time sequence occupation data in the same period is further used to calculate the cumulative time of the communication bandwidth occupation rate, the buffer remaining rate and the scheduling thread occupation rate in the area close to the resource saturation threshold, and the cumulative time is ratioed with the total sampling time to calculate the resource occupation density;

[0019] According to the preset task priority level system of the edge computing gateway, the corresponding weight factors are configured for the communication bandwidth occupation rate, the buffer remaining rate and the scheduling thread occupation rate, and the use fluctuation slope and the resource occupation density are normalized according to the weighted fusion rule to finally synthesize a single resource anomaly representation value for quantitatively describing the severity of the overall resource anomaly occupation in a fixed time period.

[0020] Preferably, the specific steps of configuring the weight factors according to the preset task priority level system of the edge computing gateway and fusing to generate the resource anomaly representation value are as follows:

[0021] According to the pre-defined task priority level system of the edge computing gateway, the sensitivity of the communication bandwidth occupation rate, the buffer remaining rate and the scheduling thread occupation rate resources to various high-priority tasks is quantitatively evaluated, the weight factor value of each type of resource is set according to the sensitivity, and a mapping function relationship between the task priority level and the resource use indicator is established;

[0022] The use fluctuation slope and the resource occupation density data calculated in a fixed time period are uniformly normalized.

[0023] According to the one-to-one correspondence relationship between the preset weight factor and the normalized communication bandwidth occupation rate, the buffer remaining rate and the scheduling thread occupation rate index, the normalized resource occupation density and the normalized resource fluctuation slope of each type of resource are sequentially calculated to obtain a weighted value, and finally a single resource abnormality representation value is generated.

[0024] Preferably, the specific steps of extracting the execution state information of the high-priority task in a fixed time period and generating the high-priority task blocking coefficient are as follows:

[0025] In the fixed time period set by the edge computing gateway, based on the task scheduling queue log and the task execution state record, all tasks marked as high-priority are filtered, the waiting time data from entering the scheduling queue to starting execution is extracted, and compared with the preset response time threshold, a set of high-priority tasks with waiting time exceeding the response time threshold is identified, and a high-priority task response timeout list is established.

[0026] For the identified high-priority task response timeout list, the scheduling waiting interval of each task is traversed, the actual waiting time is counted and accumulated, the total duration of all high-priority tasks being blocked in the scheduling link in the fixed time period is calculated, and the number of blocked tasks is recorded at the same time.

[0027] The total duration of high-priority tasks being blocked is divided by the total number of high-priority tasks in the current period, and further combined with the average waiting tolerance time of unit task to generate the high-priority task blocking coefficient.

[0028] Preferably, the specific steps of mapping and fusing the resource abnormality representation value and the high-priority task blocking coefficient in a nonlinear function to generate the scheduling resource abnormality occupation index are as follows:

[0029] After the fixed time period ends, the scheduling system of the edge computing gateway obtains the resource abnormality representation value calculated and the high-priority task blocking coefficient obtained based on the high-priority task scheduling state analysis, and combines the two to form a joint index input vector.

[0030] The scheduling system calls the pre-trained nonlinear function model, and takes the resource abnormality representation value and the high-priority task blocking coefficient in the joint index input vector as function arguments for calculation.

[0031] The fusion result output by the nonlinear function model is defined as the scheduling resource abnormality occupation index of the current scheduling period, and the scheduling resource abnormality index is registered as a dynamic abnormality perception index in the scheduling control module of the edge computing gateway.

[0032] Preferably, when the scheduling resource abnormal occupation index exceeds a set threshold value, a buffer resource reallocation mechanism is triggered to dynamically adjust the buffer quota proportion corresponding to different priority tasks, and the specific steps of performing compression processing on the data segments of the low-priority task are as follows:

[0033] Based on the generated scheduling resource abnormal occupation index and the set reference threshold value, the buffer pressure index of the current period is calculated, and the calculation expression is as follows:

[0034]

[0035] In the formula, is the buffer pressure index, is the scheduling resource abnormal occupation index, and θ ref is the scheduling resource abnormal occupation index reference threshold value.

[0036] After the buffer pressure index is calculated, the target buffer quota proportion corresponding to the high-priority task and the low-priority task is calculated according to the buffer pressure index , and the calculation expression is as follows:

[0037]

[0038] In the formula, β high is the target buffer quota proportion of the high-priority task, β base is the basic quota proportion of the high-priority task, δ is the quota adjustment amplitude coefficient, tanh is the hyperbolic tangent function, β low is the buffer quota proportion of the low-priority task.

[0039] Based on the obtained quota proportion, the write space that can be occupied by the high-priority task and the write space that can be reserved by the low-priority task are determined in combination with the available buffer capacity, and the calculation expression is as follows:

[0040]

[0041] In the formula, C high is the buffer capacity of the high-priority task, C avail is the current available buffer capacity, η is the compression reservation factor, C low is the buffer capacity of the low-priority task after compression.

[0042] In the above technical solution, the technical effects and advantages provided by the present application are as follows:

[0043] The application realizes efficient linkage between resource pressure perception and task scheduling behavior by periodically monitoring and modeling the use state of core resources such as communication bandwidth, buffer and scheduling thread in the edge computing gateway, and dynamically generating a scheduling resource abnormal occupation index combined with the scheduling execution of high-priority tasks. The method uses a nonlinear mapping mechanism to fuse the resource abnormal representation value and the task blocking coefficient, improving the response sensitivity and judgment accuracy of the system to sudden resource shortage and task blocking state. When the abnormal index rises, the system can trigger buffer resource reallocation and data compression strategies in real time, release resource bottlenecks, and guarantee the write continuity and scheduling real-time of high-priority tasks, thereby effectively preventing the occurrence of problems such as task starvation, data loss and early warning failure. Through this mechanism, the scheduling stability, task reliability and resource scheduling intelligence of the edge computing system under high concurrency and complex working conditions are significantly enhanced, and the system has good engineering adaptability and practical application promotion value. BRIEF DESCRIPTION OF DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0045] Figure 1 A method flow chart of a timing data management method of an edge computing gateway. DETAILED DESCRIPTION

[0046] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations may, however, be implemented in many different forms and should not be construed as limited to the examples set forth herein; rather, these example implementations are provided so that this disclosure will be thorough and complete, and will fully convey the gist of each example to those skilled in the art.

[0047] The present application provides a timing data management method of an edge computing gateway as shown in Figure 1 The timing data management method of the edge computing gateway comprises the following steps:

[0048] In a preset fixed time period, the communication bandwidth occupation rate, the buffer remaining rate and the scheduling thread occupation rate of the edge computing gateway are recorded respectively, a resource use raw data matrix covering all time points in the fixed time period is constructed, and the resource use raw data matrix is used to represent the timing occupation state of various types of core scheduling resources in the period;

[0049] In a preset fixed time period, the communication bandwidth occupancy rate, buffer remaining rate and scheduling thread occupancy rate of the edge computing gateway are recorded respectively, and a resource usage original data matrix is constructed. The core role is to provide a complete, detailed and quantifiable resource dynamic portrait for the edge computing system, and to provide data basis support for subsequent scheduling strategy optimization and abnormal identification. As a multi-task concurrent execution computing node, the edge computing gateway needs to simultaneously cope with multiple operations such as uploading tasks, processing tasks and alarm response, and these tasks will compete for limited resources such as communication, cache and computing threads during execution. By continuously recording the state of the three types of key resources in a fixed period, a high time resolution resource usage time series can be formed. After time alignment, the original data matrix formed by these sequences can reflect the actual occupancy fluctuation trajectory of the resources in the period. For example, the change trend of the communication bandwidth occupancy rate can reveal whether the burst data upload squeezes the normal upload channel, the buffer remaining rate can warn whether the cache space is close to exhaustion, and the scheduling thread occupancy rate can reflect whether the system processing capacity is in a saturated state. The original data matrix not only realizes the quantitative expression of the current resource state, but also provides an indispensable data basis for subsequent construction of resource abnormality representation value, analysis of task blocking root cause, and triggering of dynamic scheduling control mechanism. It enables the system to have intelligent decision-making ability based on objective resource state perception, and enhances the scheduling robustness and real-time guarantee capability of multi-task collaborative processing in the edge computing environment.

[0050] In a preset fixed time period, the communication bandwidth occupancy rate, buffer remaining rate and scheduling thread occupancy rate of the edge computing gateway are recorded respectively, and a resource usage original data matrix is constructed. The core role is to provide a complete, detailed and quantifiable resource dynamic portrait for the edge computing system, and to provide data basis support for subsequent scheduling strategy optimization and abnormal identification. As a multi-task concurrent execution computing node, the edge computing gateway needs to simultaneously cope with multiple operations such as uploading tasks, processing tasks and alarm response, and these tasks will compete for limited resources such as communication, cache and computing threads during execution. By continuously recording the state of the three types of key resources in a fixed period, a high time resolution resource usage time series can be formed. After time alignment, the original data matrix formed by these sequences can reflect the actual occupancy fluctuation trajectory of the resources in the period. For example, the change trend of the communication bandwidth occupancy rate can reveal whether the burst data upload squeezes the normal upload channel, the buffer remaining rate can warn whether the cache space is close to exhaustion, and the scheduling thread occupancy rate can reflect whether the system processing capacity is in a saturated state. The original data matrix not only realizes the quantitative expression of the current resource state, but also provides an indispensable data basis for subsequent construction of resource abnormality representation value, analysis of task blocking root cause, and triggering of dynamic scheduling control mechanism. It enables the system to have intelligent decision-making ability based on objective resource state perception, and enhances the scheduling robustness and real-time guarantee capability of multi-task collaborative processing in the edge computing environment.

[0051] For the communication behavior of the edge computing gateway during operation, a fixed sampling time interval is set, the real-time data transmission rate of the communication channel is continuously sampled, and the communication bandwidth occupancy rate sequence is calculated by the ratio of the sampling point to the theoretical maximum bandwidth, to reflect the dynamic change process of the communication resources in the fixed time period;

[0052] In the running process of the edge computing gateway, in order to accurately capture the dynamic state of the communication resources, first, a fixed sampling time interval (such as once per second) is set, and the real-time data transmission rate of each communication interface of the gateway is continuously collected. The actual transmission rate at each time is calculated by the ratio of the theoretical maximum bandwidth corresponding to the interface, thereby forming a time series of communication bandwidth occupancy rate. This sequence can accurately reflect the load change trend of the communication link in the period, and can provide data basis for subsequent identification of resource abnormal states such as burst bandwidth congestion and bandwidth saturation, and is one of the key inputs in the scheduling resource perception mechanism.

[0053] At the time point synchronized with the communication sampling, the buffer usage state allocated to each task type in the edge computing gateway is periodically read, the ratio of the remaining capacity to the total capacity of the buffer at each sampling time is calculated, a buffer remaining rate sequence is formed, and a historical capacity change trend is combined for smoothing processing to improve the accuracy of abnormal change identification;

[0054] To realize real-time monitoring of memory resource usage, at the time point synchronized with the communication data sampling, the edge computing gateway periodically acquires buffer usage information corresponding to various tasks (such as uploading, processing, cleaning, etc.), calculates the ratio between the remaining capacity and the total capacity at each sampling point, and generates a buffer remaining rate sequence. This sequence reflects the dynamic availability of cache resources within the scheduling period. After processing with a smoothing algorithm such as moving average, it can identify the trend of gradual depletion or instantaneous sharp decline of buffer resources, which is of great warning significance for preventing high-priority tasks from being unable to write data due to cache depletion.

[0055] The active number of task scheduling threads in the edge computing gateway and the total thread pool capacity are synchronously counted to obtain a scheduling thread occupancy rate sequence. The communication bandwidth occupancy rate sequence, the buffer remaining rate sequence, and the scheduling thread occupancy rate sequence are time-aligned to construct a resource usage raw data matrix according to a unified timestamp, thereby forming a data infrastructure that fully reflects the time-series occupancy status of the three key resources: communication bandwidth, buffer, and scheduling thread within the period.

[0056] To evaluate the availability of processing capacity, the system counts the ratio between the number of currently active scheduling threads and the total capacity of the thread pool at the same time as the sampling of the previous two types of data, forming a time series of scheduling thread occupancy rates. This sequence reflects the thread resource pressure of the gateway when scheduling tasks. By aligning with the communication bandwidth occupancy rate sequence and the buffer remaining rate sequence with a unified timestamp, a resource usage raw data matrix is constructed. This matrix, as a complete time-series resource occupancy mapping structure, supports subsequent resource fusion modeling, anomaly detection, and scheduling control decisions, and is the basic and core component of the dynamic resource regulation logic.

[0057] Based on the resource usage raw data matrix, the usage fluctuation slope and resource occupancy density of the communication bandwidth occupancy rate, the buffer remaining rate, and the scheduling thread occupancy rate within a fixed time period are calculated, and weighting factors are configured according to the task priority level corresponding to each resource indicator to fuse the resource indicators to generate a single resource anomaly representation value, which is used to quantify the overall resource anomaly occupancy degree within the period;

[0058] The core role of this step is to convert the multi-dimensional, heterogeneous resource usage state information of the edge computing gateway within a fixed time period into a unified, quantifiable comprehensive abnormality index, thereby providing the scheduling system with accurate, real-time resource abnormality perception capability. By calculating the usage fluctuation slope and resource occupation density of the communication bandwidth occupancy rate, buffer remaining rate, and scheduling thread occupancy rate, three key resource indicators, not only can the dynamic fluctuation trend in the resource usage process be captured, but also the persistence of high pressure state can be reflected. This two-dimensional analysis mode breaks through the traditional static judgment method of single occupancy rate or average value, and can more sensitively identify the risk of resource instantaneous burst or continuous congestion. At the same time, in the fusion stage, by introducing the weighting factor set by the task priority level system, the model fully considers the actual influence weight of various resources in the system on the success rate of high-priority task scheduling. The performance of resource-sensitive tasks in the system will therefore be highlighted in the fusion process, improving the response accuracy of the overall scheduling model to the key task guarantee capability. Finally, the single resource abnormality representation value generated by the weighted normalization operation is used as the core quantitative index for the scheduling system to perceive the resource state within the entire period, which can be directly used to determine whether resource reallocation, task preemption scheduling, or buffer adjustment control strategies are needed, effectively improving the stability and real-time performance of the edge computing system in complex multi-task concurrent scenarios. This step realizes the logical closed loop between multi-source resource state perception and scheduling control response, and is the basic link for the system to realize intelligent and fine-grained resource scheduling management.

[0059] Based on the resource usage original data matrix, the specific steps of fusing the resource indicators to generate a single resource abnormality representation value are as follows:

[0060] The time series occupancy data corresponding to the communication bandwidth occupancy rate, buffer remaining rate, and scheduling thread occupancy rate in the resource usage original data matrix are processed. The continuous change model is constructed by using curve fitting in a fixed time period, and the derivative mean is calculated by taking the derivative of the fitting curve at each sampling point, to obtain the usage fluctuation slope of each resource type, which is used to reflect the usage change rate and fluctuation intensity of the communication bandwidth, buffer, and scheduling thread resources within the period, providing a quantitative basis for evaluating the dynamic stability of the system resources.

[0061] In the resource usage raw data matrix collected by the edge computing gateway, the communication bandwidth occupancy rate, the buffer remaining rate and the scheduling thread occupancy rate are presented in the form of time series. To obtain the usage fluctuation trend of each type of resource in a fixed time period, first, the time series data of each type of resource is processed by curve fitting, such as using polynomial regression or spline interpolation, to construct a continuous change curve of each resource with time. Then, the derivative of the curve is calculated to extract the instantaneous change rate at each time point, and the mean value of the derivative in the period is calculated as the usage fluctuation slope of the resource. The fluctuation slope reflects the change rate and fluctuation intensity of resource occupancy in the period, which can be used to identify whether there is a sudden and high-frequency dynamic fluctuation of the resource, and to provide a basis for the system to judge the scheduling stability.

[0062] Based on the usage fluctuation slope calculation, the time series occupancy data in the same period is used to calculate the cumulative time of the communication bandwidth occupancy rate, the buffer remaining rate and the scheduling thread occupancy rate in the area close to the resource saturation threshold, and the cumulative time is compared with the total sampling time to calculate the resource occupancy density, which is used to describe the persistence of the communication resource, the cache resource and the scheduling thread resource in the high-pressure running state in the period, and to supplement the time stability feature of the resource abnormal state;

[0063] After obtaining the usage fluctuation slope, the cumulative duration of the communication bandwidth occupancy rate, the buffer remaining rate and the scheduling thread occupancy rate in the near-saturation state is calculated based on the resource usage time series in the same time period. The specific implementation is to set a reasonable resource high-load threshold (such as communication occupancy rate higher than 85%, buffer remaining rate lower than 15%, thread occupancy rate higher than 90%), determine whether it enters the high-pressure state at each sampling point, and accumulate the duration in this state, and finally compare it with the total cycle time to obtain the resource occupancy density index. This index reflects whether the resource is in a high-stress state for a long time in the period, and quantifies the system running pressure, which is an important feature for auxiliary evaluation of resource overload risk and scheduling bottleneck.

[0064] According to the pre-set task priority level system of the edge computing gateway, the communication bandwidth occupancy rate, the buffer remaining rate and the scheduling thread occupancy rate are respectively configured with corresponding weight factors, and the usage fluctuation slope and the resource occupancy density are normalized according to the weighted fusion rule to finally synthesize a single resource abnormality representation value, which is used to quantitatively describe the severity of the overall resource abnormality occupancy in a fixed time period, and is an important judgment index for the subsequent execution of resource reallocation, task priority preemption and other control strategies of the scheduling system.

[0065] After the calculation of the fluctuation slope and the resource occupation density is completed, in order to enhance the accuracy of the abnormal state identification, the task priority needs to be introduced to adjust the difference in resource sensitivity. According to the setting of the edge computing gateway for different task priority levels, the corresponding weighting factors are configured for the communication bandwidth occupation rate, the buffer remaining rate and the scheduling thread occupation rate, for example, the key task may depend more on communication and thread resources, and then the weighting value is higher. After the fluctuation slope and the occupation density of each resource index are normalized, the single resource abnormal representation value is fused according to the weight rule. The representation value comprehensively considers the resource change rate and the high pressure persistence, and quantitatively reflects the difference in resource dependence of the task importance, which is the key input parameter for the scheduling system to judge whether the resource is abnormally occupied in the current period, and plays a decisive role in triggering the dynamic control strategy.

[0066] The specific steps of configuring the weight factor and generating the resource abnormal representation value according to the task priority level system preset by the edge computing gateway are as follows:

[0067] According to the task priority level system predefined in the edge computing gateway, the sensitivity of each type of high priority task to the communication bandwidth occupation rate, the buffer remaining rate and the scheduling thread occupation rate is quantitatively evaluated, and the weight factor value of each type of resource is set according to the sensitivity. The higher the weight factor value, the more critical the resource to the scheduling and timeliness of high priority tasks and the task completion degree, and a mapping function relationship between the task priority level and the resource usage index is established to provide task demand oriented parameters for the subsequent fusion model;

[0068] In the edge computing gateway system, the dependence of different types of tasks on resources is significantly different, for example, the security alarm information uploading task is extremely sensitive to communication bandwidth and scheduling thread resources, while the periodic log cleaning task mainly depends on buffer resources. In order to accurately reflect this difference, first of all, a task priority level system needs to be established, which divides all system tasks into multiple priority levels according to business urgency and data timeliness. Then, according to the resource calling behavior of high priority tasks in the historical running process, the dependence intensity of each type of task on the communication bandwidth occupation rate, the buffer remaining rate and the scheduling thread occupation rate is determined through statistical analysis, and the corresponding weight factor is configured for each type of resource. The role of this step is to enable the subsequent resource abnormality determination to reflect the actual urgency demand in task scheduling, realize the transition from resource oriented modeling to business demand oriented modeling, and lay the foundation for finally generating a fusion index with scheduling decision significance.

[0069] The "predefined task priority hierarchy" refers to a set of hierarchical rules set by system designers or operation personnel in advance in the edge computing gateway system according to the importance, urgency and timeliness of tasks in the business process, which is used for hierarchical classification and scheduling ordering of all tasks running in the system. The system usually contains several priority levels, such as "high priority", "medium priority", "low priority" or a more detailed five-level classification system, and each level of task corresponds to different scheduling guarantee strength. Its core role is to provide decision basis for the scheduling system, so that in the case of resource shortage or concurrent task conflict, the system can prioritize the scheduling resource allocation of high priority tasks to ensure the real-time processing capability of critical business (such as device alarm, state synchronization, fault upload, etc.), thereby improving the robustness and response reliability of the overall system.

[0070] In determining the sensitivity of high-priority tasks to communication bandwidth occupancy, buffer remaining rate, and scheduling thread occupancy, a dual method based on running data analysis and business logic correlation can be used for quantitative evaluation. First, from the perspective of running data, the completion time of tasks under different resource load conditions can be quantified through historical scheduling logs and task performance statistics records. For example, when the communication bandwidth utilization rate exceeds 80%, the completion delay of a certain type of task increases significantly, indicating that it is sensitive to communication resources. Similarly, by analyzing the correlation between task execution failure rate and buffer occupancy or thread blocking, a sensitivity model between resources and tasks can be established.

[0071] Second, modeling from the perspective of business logic, the dependence of tasks on specific resources is determined by the functional target. For example, event alarm tasks usually require millisecond-level real-time performance and are extremely sensitive to scheduling thread occupancy. Periodic log archiving tasks are more dependent on the continuous writing ability of the buffer and have a higher dependence on buffer remaining rate. Through the above two types of analysis, a sensitivity coefficient between 0 and 1 can be assigned to each type of resource, which is then mapped to a weighted factor based on the task level structure for use in the subsequent weighted fusion model of resource abnormality representation, ensuring that the impact of critical resource state changes on critical tasks is accurately quantified and responded to.

[0072] The usage fluctuation slope and resource occupancy density data obtained by calculating the communication bandwidth occupancy, buffer remaining rate and scheduling thread occupancy in a fixed time period are uniformly normalized. The minimum-maximum linear transformation method is used to map all original data to the same standard numerical interval, eliminating the differences in numerical scale and dimension of each resource indicator, ensuring that each type of indicator has an equivalent basis for comparison in the weighted fusion process, and avoiding the undue dominant effect of a certain type of resource due to the large original magnitude in the calculation;

[0073] Since the numerical range, dimension and distribution characteristics of the communication bandwidth occupancy rate, buffer remaining rate and scheduling thread occupancy rate in the two dimensions of usage fluctuation slope and resource occupancy density are significantly different, directly performing weighted fusion will lead to the result being biased towards the indicators with larger numerical absolute values, affecting the fairness of the model. Therefore, the fluctuation slope and occupancy density of each resource type need to be normalized. The specific method is to use the minimum-maximum value linear transformation method to map the minimum value of each indicator in the current period to 0 and the maximum value to 1, and the remaining values are mapped to the interval in a linear proportion. Through the normalization operation, different resource indicators are unified to the same standardized numerical scale, making the weighted fusion process consistent and comparable. The core role of this step is to eliminate the interference brought by the numerical characteristics of the original indicators, so that the fusion result truly reflects the actual degree of indicator change and the abnormal characteristics that need to be focused on by the scheduling system.

[0074] According to the one-to-one correspondence between the preset weight factor and the normalized communication bandwidth occupancy rate, buffer remaining rate and scheduling thread occupancy rate, the normalized usage fluctuation slope and the normalized resource occupancy density of each type of resource are sequentially calculated by the weighted average fusion algorithm to generate a single resource abnormality representation value. The resource abnormality representation value is used to quantitatively represent the severity of overall resource abnormality occupancy in the current fixed time period, and serves as an important basis for the edge computing gateway scheduling system to decide whether to execute resource reallocation, task preemption or buffer expansion and other scheduling control strategies in real-time scenarios.

[0075] After completing the normalization of the resource indicators, the system establishes a one-to-one correspondence between the weight factor set in the first step and the indicators processed in the second step. Then, according to the weighted average fusion method, the normalized fluctuation slope and the normalized resource occupancy density of the communication bandwidth occupancy rate, the buffer remaining rate and the scheduling thread occupancy rate are weighted and superimposed to finally fuse into a single resource abnormality representation value. This abnormality representation value reflects the overall abnormality occupancy degree of the system core resources in the current scheduling period in the form of a continuous numerical value. The key role of this step is to compress multi-dimensional resource state data into an indicator that can be called and referenced in real time, enabling the scheduling system to quickly determine whether there is a resource bottleneck or scheduling obstacle, and accordingly trigger resource reallocation, task priority adjustment and other dynamic control strategies, effectively improving the real-time response capability and task guarantee level of the edge computing system.

[0076] In a fixed time period, the execution state information of all tasks set to high priority is extracted, the number of tasks with waiting time exceeding the preset response threshold is identified, and the cumulative blocking time is counted to generate a high-priority task blocking coefficient, which represents the degree of scheduling impact on high-priority tasks due to resource abnormality occupancy.

[0077] The role of this step is to dynamically evaluate the system's ability to protect critical tasks under resource shortage or concurrent conflict conditions by conducting a fine-grained quantitative analysis of the scheduling execution status of high-priority tasks in the edge computing gateway within a specific time period, thereby providing real-time and reliable performance feedback basis for resource scheduling strategy optimization. In the edge computing environment, critical tasks such as device alarm upload and fault event reporting are usually set to high priority, and their scheduling timeliness directly relates to the safety and response capability of the system. However, in a multi-task concurrent and resource-intensive scenario, even if the priority is set, if the resource allocation mechanism cannot respond in time, it may still cause high-priority tasks to wait for a long time, affecting the efficiency of event processing. Through this step, the system first identifies which high-priority tasks have actual waiting time exceeding the pre-set response threshold within the current fixed period, then calculates the total blocking time, and generates a high-priority task blocking coefficient combined with the number of blocked tasks. This coefficient, as a continuous and quantifiable scheduling performance indicator, can objectively reflect the scheduling delay of critical tasks in the real operating environment and reveal the impact of resource abnormal occupation on business core tasks. Through this indicator, the scheduling system can determine whether to adjust the current resource allocation strategy, whether to trigger task preemption, whether to expand buffer resources, and other dynamic control measures, thereby realizing the evolution of task scheduling from "rule-driven" to "feedback-driven", and improving the stability, response speed and business continuity of the edge computing system under complex working conditions.

[0078] The specific steps of extracting the execution status information of high-priority tasks within a fixed time period and generating a high-priority task blocking coefficient are as follows:

[0079] Within the fixed time period set by the edge computing gateway, based on the task scheduling queue log and task execution status record, all tasks marked as high priority are filtered, the waiting time data from entering the scheduling queue to starting execution is extracted, and compared with the pre-set response time threshold to identify the set of high-priority tasks whose waiting time exceeds the response time threshold, and a high-priority task response timeout list is established.

[0080] Within the fixed time period set by the edge computing gateway, the system first extracts scheduling logs and task execution status data from the task scheduling management module, and filters the set of tasks marked as high priority. Then, for each high-priority task, the time stamp of entering the scheduling queue and the time stamp of actually starting execution are obtained, the scheduling waiting time of the task is calculated, and compared with the pre-set response time threshold in the system. All tasks with waiting time exceeding the response threshold are classified into the high-priority task response timeout list.

[0081] The role of this step is to identify abnormal tasks with response delay due to untimely resource scheduling from all high-priority tasks, and to provide target samples for subsequent statistical scheduling system to ensure the ability of key tasks.

[0082] For the identified high-priority task response timeout list, the scheduling waiting interval of each task is traversed, the actual waiting time is counted and accumulated, the total time of all high-priority tasks being blocked in the scheduling link in a fixed time period is calculated, and the number of blocked tasks is recorded at the same time, which is used to reflect the resource response ability of the scheduling system to high-priority tasks;

[0083] After obtaining the high-priority task response timeout list, the system enters the statistical link, and the accumulation operation of the scheduling waiting time of each task in the list is performed to obtain the total blocking time experienced by all high-priority tasks with response timeout in the current time period. At the same time, the system counts the number of tasks with response timeout, which is used to reflect the unsatisfactory condition of the scheduling system to high-priority tasks in this period.

[0084] The role of this step is to evaluate the severity of the scheduling bottleneck from the two dimensions of time and task quantity, reveal whether there is a problem of untimely processing of high-priority tasks in the current resource scheduling strategy of the system, and provide necessary data for generating scheduling performance indicators reflecting the task guarantee ability.

[0085] The total time of high-priority tasks being blocked is divided by the total number of high-priority tasks in the current period, and further combined with the average waiting tolerance time of unit task to generate the high-priority task blocking coefficient. This coefficient is a dynamic quantitative indicator of the scheduling performance of the edge computing gateway, which is used to measure the actual performance of the current resource scheduling strategy in the high-pressure condition to the key task guarantee ability, and as a decision reference basis for whether to trigger resource reallocation or priority strategy adjustment.

[0086] After the blocking time and the number of blocked tasks are counted, the system calculates the high-priority task blocking coefficient based on the two variables. The specific way is to divide the total time of high-priority tasks being blocked by the number of all high-priority tasks in the period to obtain the average blocking time of unit task, and further perform ratio processing on this value and the system-defined average response tolerance time of unit task, and finally generate the high-priority task blocking coefficient. This coefficient is a quantifiable scheduling performance indicator, which can dynamically reflect the degree of delay in processing high-priority tasks by the current scheduling mechanism, and is an important basis for judging whether to adjust the scheduling strategy, perform resource reallocation, or trigger key task preemption logic.

[0087] This step converts the scheduling performance of high-priority tasks into a numerical model that can be used for decision feedback, which helps to improve the agility and task response guarantee ability of the scheduling system.

[0088] The resource abnormality characterization value and the high-priority task blocking coefficient are input into a nonlinear function for mapping and fusion to obtain a scheduling resource abnormality occupation index of the current scheduling period, which is used as an abnormality perception index of the dynamic scheduling system to determine whether resource reallocation or task preemption scheduling strategy needs to be performed in real time to ensure the scheduling real-time performance of high-priority tasks in a resource shortage situation.

[0089] The step is to realize the real-time perception and early warning control capability of the edge computing gateway to scheduling risks in a multi-task concurrent and resource competition environment by constructing an abnormality index system with strong fusion and sensitive response. Specifically, the scheduling resource abnormality occupation index is generated by inputting the resource abnormality characterization value and the high-priority task blocking coefficient into a nonlinear function model for mapping and fusion. This index not only reflects the overall pressure state of core resources such as communication bandwidth, buffer and scheduling thread in the current scheduling period, but also considers the interference degree of high-priority tasks in the actual scheduling process. Compared with the traditional single-index monitoring method, the fusion index uses a nonlinear function to model the complex coupling relationship between input variables, has enhanced response characteristics to extreme states, can amplify scheduling risk signals in critical situations such as resource sudden shortage and task intensive blocking, and improves the identification sensitivity of the system to scheduling abnormal trends. By taking the index as an abnormality perception index of the dynamic scheduling system, the scheduling engine can automatically trigger a series of control strategies such as resource reallocation, task preemption, buffer adjustment when detecting that the abnormal value exceeds the preset threshold, thereby effectively avoiding that high-priority tasks cannot be processed in time due to resource saturation. This mechanism realizes a logical closed loop from "resource state monitoring" to "scheduling behavior driving", enables the system to have intelligent scheduling capabilities such as autonomous judgment and rapid response, and significantly improves the real-time performance of task execution, the robustness of the scheduling system and the reliability of the overall service of edge computing.

[0090] The specific steps of inputting the resource abnormality characterization value and the high-priority task blocking coefficient into a nonlinear function for mapping and fusion to generate a scheduling resource abnormality occupation index are as follows:

[0091] After the fixed time period ends, the scheduling system of the edge computing gateway obtains the calculated resource abnormality characterization value and the high-priority task blocking coefficient based on the analysis of the scheduling state of high-priority tasks, and combines the two to form a joint index input vector as an input variable to express the composite characteristics of the resource usage state and the key task scheduling performance of the system in the current period, providing multi-dimensional basic data for subsequent nonlinear fusion mapping;

[0092] After the fixed time period ends, the edge computing gateway scheduling system first extracts key input variables from the two sub-modules: one is the resource anomaly representation value calculated by the fusion of communication bandwidth occupancy, buffer remaining rate and scheduling thread occupancy, and the other is the high-priority task blocking coefficient obtained based on high-priority task execution state analysis. The system combines these two independent indicators into a two-dimensional vector form of joint input structure, and constructs input samples based on timestamp alignment. The role of this step is to integrate and express the two key dimensions of information in the scheduling system, "resource state" and "task performance", to form a multi-dimensional feature vector that can reflect the resource load pressure and task execution abnormality in the current scheduling period, providing data support for subsequent nonlinear model processing and improving the integrity and expression ability of system modeling.

[0093] The scheduling system calls the pre-trained nonlinear function model, and takes the resource anomaly representation value and the high-priority task blocking coefficient in the joint index input vector as the function independent variables for calculation. The nonlinear function model uses an activation function structure with adjustable parameters, such as the S-shaped function, exponential growth function or logistic regression function, to model the nonlinear coupling relationship between input variables and amplify the response to critical states such as resource sudden abnormality or continuous blocking of key tasks, thereby improving the scheduling system's ability to sensitively identify abnormal situations.

[0094] After the input vector is constructed, the system inputs it into the pre-trained nonlinear function model. The model can use the activation function commonly used in neural networks, such as the S-shaped function, exponential function or ReLU function, by introducing adjustable weight parameters and nonlinear factors to model the complex coupling relationship between the resource anomaly representation value and the high-priority task blocking coefficient. The function calculation process not only considers the nonlinear amplification effect between variables, but also responds to the deterioration trend of scheduling performance in extreme states, so that the output value more sensitively reflects the abnormal intensity when the system is in a sudden resource occupation or key task intensive blocking state. The role of this step is to build a "discrimination function" of scheduling performance anomaly through fine modeling, significantly improving the system's ability to identify critical states and providing a dynamic decision basis for scheduling control strategies.

[0095] The pre-trained nonlinear function model refers to a mathematical function structure constructed and optimized based on a large amount of historical running data, task scheduling records and resource usage states through machine learning or fitting algorithms before the deployment of the edge computing gateway, which is used to capture the complex nonlinear relationship between resource state indicators and scheduling performance. Unlike traditional static rules, this model not only represents the interaction effect between resource abnormal representation values and high-priority task blocking coefficients, but also produces an enhanced response to scheduling bottlenecks in critical states. For example, during the model training process, the system identifies the nonlinear coupling trend between resource occupation surge and task waiting time increase, and realizes sensitive amplification of abnormal points through function parameter adjustment. Its role is to enable the scheduling system of the edge computing gateway to have a highly flexible abnormality identification capability. When the input features are in a complex intersection interval, the model can output an abnormality index with higher discrimination and judgment ability, significantly improving the system's perception accuracy and response speed to abnormal scheduling trends, and providing reliable data support for intelligent and real-time resource management strategies.

[0096] The fusion result output by the nonlinear function model is defined as the scheduling resource abnormal occupation index of the current scheduling period, and the scheduling resource abnormal index is registered as a dynamic abnormality perception indicator in the scheduling control module of the edge computing gateway, which is used to judge whether the system is in a resource scheduling performance abnormal state in real time. When the index exceeds the set threshold, the scheduling system can automatically trigger the resource reallocation mechanism, task priority improvement strategy or buffer dynamic adjustment logic, realize the transformation from fixed rule driving to dynamic resource scheduling control mechanism based on intelligent perception feedback, and effectively improve the real-time performance of key task scheduling and the stability of the overall system operation.

[0097] After the nonlinear function model is calculated, the system defines its output value as the scheduling resource abnormal occupation index, and registers the index in the scheduling engine as a dynamic abnormality perception indicator. The scheduling system continuously monitors the numerical change of the index and compares it with the set abnormal threshold in real time. Once the index exceeds the threshold, the built-in resource reallocation mechanism, high-priority task promotion strategy or buffer dynamic expansion logic is triggered immediately to ensure the continuity of key task scheduling and the integrity of data transmission. The role of this step is to compress the complex resource and task abnormal state into an operable and controllable scheduling indicator, enabling the scheduling system to have a closed-loop capability from "state perception" to "adaptive response", truly realizing the transformation from traditional static strategy to intelligent perception driven scheduling mechanism, and comprehensively improving the operation intelligence and stability of the edge computing system.

[0098] When the abnormal scheduling resource occupation index exceeds the set threshold, the buffer resource redistribution mechanism is triggered, the buffer quota proportion corresponding to different priority tasks is dynamically adjusted, the buffer space occupied by the low-priority task is released, and the buffer capacity of the high-priority task is temporarily expanded, so as to ensure the continuity of data writing and uploading operation of the key task; at the same time, the data segments of the low-priority task are compressed, only the statistical summary information is retained, the buffer resource pressure is reduced, the queuing block or data loss of the high-priority task due to insufficient buffer resources is prevented, and the scheduling efficiency and response reliability of the system are improved;

[0099] The role of this step is to provide a buffer dynamic regulation mechanism with real-time response ability and resource perception ability under the complex running environment of the edge computing gateway facing high scheduling resource tension, so as to guarantee the data processing continuity and system stability of the high-priority key task in the burst situation. Specifically, when the abnormal scheduling resource occupation index exceeds the set threshold of the system, the system automatically judges that the current resource scheduling state has appeared potential bottleneck, which may cause the key task to fail to write or upload data in time, so as to trigger the buffer resource redistribution mechanism. According to the scheduling pressure intensity, the buffer quota of high and low priority tasks is dynamically adjusted, the buffer space occupied by the low-priority task is released, and the cache writing capacity of the high-priority task is temporarily expanded, so as to ensure that the key data of the high-priority task will not be interrupted or blocked in writing during resource tension. At the same time, in order to prevent the overflow of the whole system memory, the compression strategy is implemented for the low-priority task, only the key statistical summary information is retained, so as to reduce the resource consumption on the premise of retaining the business integrity as much as possible. The mechanism not only improves the response ability of the system to the key task, but also enhances the flexibility of resource scheduling and the hierarchical fault tolerance ability of task processing, effectively avoids the serious consequences such as data loss, uploading delay or alarm failure caused by unbalanced resource scheduling, and finally realizes the high reliability, high real-time performance and business continuity in the concurrent processing of multiple tasks in the edge computing environment.

[0100] When the abnormal scheduling resource occupation index exceeds the set threshold, the buffer resource redistribution mechanism is triggered, the buffer quota proportion corresponding to different priority tasks is dynamically adjusted, and the data segments of the low-priority task are compressed. The specific steps are as follows:

[0101] Based on the generated abnormal scheduling resource occupation index and the set reference threshold, the buffer pressure index of the current period is calculated, and the calculation expression is as follows:

[0102]

[0103] In the formula, is a buffer pressure index, indicating the buffer resource pressure level caused by the overall system resource scheduling abnormality in the current scheduling period, as a direct driving parameter for the subsequent buffer resource dynamic reallocation mechanism, the larger the value, the greater the resource scheduling pressure on the buffer, is a scheduling resource abnormality occupation index, which comprehensively reflects the fusion index of the state of the system key resources (communication bandwidth occupation rate, buffer remaining rate, scheduling thread occupation rate) and the degree of high priority task blocking in the current scheduling period, the larger the value, the more serious the scheduling abnormality, and the stronger the intervention requirement for system resource reallocation, θ ref is a scheduling resource abnormality occupation index reference threshold value;

[0104] The implementation of this step realizes the effective conversion of the scheduling awareness index to the buffer control signal.

[0105] After the completion of the buffer pressure index calculation, according to the buffer pressure index calculation of the target quota proportion of the buffer corresponding to the high priority and low priority task, the calculation expression is as follows:

[0106]

[0107] , wherein, β high is the target buffer quota proportion of the high priority task, indicating the buffer resource proportion that the system should allocate to the high priority task in the current period, which is one of the final output parameters of this step and directly affects the data write protection level of the high priority task, β base is the basic quota proportion of the high priority task, which is the buffer allocation reference value preset by the system in the normal scheduling state, usually taking a value between 0.5 and 0.7, indicating the default protection degree of the system to the high priority task in the absence of abnormality, δ is the quota adjustment amplitude coefficient, indicating the maximum dynamic increase amplitude of the buffer proportion of the high priority task allowed when the resource abnormal state is detected; the larger the value, the stronger the tilt ability of the system under pressure, usually taking a value range of 0.1-0.4, tanh is the hyperbolic tangent function, β low is the buffer quota proportion of the low priority task, indicating the buffer space proportion reserved by the system for the low priority task in the current period; its value is determined inversely by the high priority task quota, ensuring that the total quota sum is 1;

[0108] The hyperbolic tangent function (Tanh function) is a common S-shaped (Sigmoid) nonlinear function, whose function value in the entire domain has an output range of (-1, 1) and is symmetric about the origin. When the input approaches 0, the output of tanh approaches 0; when the input approaches positive infinity, the output approaches 1; when the input approaches negative infinity, the output approaches -1. This smooth, nonlinear growth characteristic makes it used in many machine learning and dynamic scheduling systems to amplify the trend of changes in input variables while suppressing the sharp fluctuations caused by extreme values. The introduction of the buffer pressure index mainly serves to normalize the mapping of the buffer pressure index (which can be an arbitrarily large positive value) to a finite interval (0 to 1) and maintain a sensitive response to abnormal pressure changes. In this way, when the system senses a sudden increase in buffer pressure, the buffer quota proportion of high-priority tasks can be quickly increased; when the pressure is small or stable, the scheduling system can still maintain a stable state to prevent frequent resource fluctuations caused by small changes, thereby achieving the three goals of smooth adjustment, nonlinear amplification, and boundary control of the scheduling strategy.

[0109] Through this step, the system can achieve dynamic tilt scheduling and allocation of resources between different priority tasks without interrupting the original task logic.

[0110] Based on the obtained quota proportion, in combination with the available buffer capacity, the write space that can be occupied by high-priority tasks and the write space that can be reserved for low-priority tasks are determined, and the calculation expression is as follows:

[0111]

[0112] In the formula, C high is the high-priority task buffer capacity, indicating the buffer capacity dynamically allocated to high-priority tasks in the current scheduling period, with units of bytes (Byte) or megabytes (MB), which directly affects the write continuity and upload reliability of high-priority tasks in a resource-constrained environment, C avail is the current available buffer capacity, indicating the physical or logical buffer space that can still be allocated in the system for allocating write areas for different priority tasks in the scheduling period, which is usually updated in real time by the memory management subsystem or cache control module, η is the compression reservation factor, indicating the application intensity of the compression mechanism, used to control the compression reservation proportion of low-priority task data, with a value range of [0, 1], η = 1 represents no compression full reservation, and η = 0.2 represents only 20% of the summary information is reserved. This parameter is the core control variable of the embedded compression strategy of the system and can be dynamically configured in combination with the task type and data characteristics, C lowis the low-priority task compressed buffer capacity, representing the effective write buffer capacity actually allocated to the low-priority task in the current period; in the resource shortage state, the capacity is usually limited to the minimum reserved space after compression, for storing task statistical abstracts or part of simplified data, to reduce the system load.

[0113] This step realizes the write continuity guarantee of the high-priority task, and significantly reduces the resource occupation of the buffer under the premise of not completely discarding the low-priority data, effectively avoids the task queuing blockage and the occurrence of the critical data loss phenomenon, and further improves the real-time performance and stability of the scheduling system.

[0114] The application realizes efficient linkage between resource pressure perception and task scheduling behavior by periodically monitoring and modeling the use state of core resources such as communication bandwidth, buffer and scheduling thread in the edge computing gateway, and dynamically generating a scheduling resource abnormal occupation index combined with the scheduling execution situation of the high-priority task. The method uses a nonlinear mapping mechanism to fuse the resource abnormal representation value and the task blocking coefficient, improves the response sensitivity and judgment accuracy of the system to the sudden resource shortage and task blocking state. When the abnormal index rises, the system can trigger buffer resource reallocation and data compression strategy in real time, release the resource bottleneck, guarantee the write continuity and scheduling real-time performance of the high-priority task, thereby effectively preventing the occurrence of problems such as task starvation, data loss and early warning failure. Through the mechanism, the scheduling stability, task reliability and resource scheduling intelligence of the edge computing system under high concurrency and complex working conditions are significantly enhanced, and the application has good engineering adaptability and practical application promotion value.

[0115] The above formulas are dimensionless values calculated, the formulas are obtained by software simulation of a large amount of data to obtain a formula of the latest real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0116] The above only describes some exemplary embodiments of the application by way of illustration, and it is needless to say that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the application. Therefore, the above figures and descriptions are illustrative in nature and should not be understood as limiting the scope of protection of the claims of the application.

[0117] It should be noted that, in the present document, relational terms are used to convey a relationship of one entity or action to another entity or action. For example, without necessarily implying any actual relationship or order between entities or actions, the terms "first," "second," "top" and "bottom" are used to name different entities and actions, and are used to distinguish one element from another. Furthermore, the terms "comprise," "include," and "have," and variations thereof, do not preclude the presence or addition of one or more other elements or steps, unless the context clearly indicates otherwise. The term "including" as used herein, means "comprising" in the sense of the transition word "comprising," and not by way of limitation. The use of "including" and "including" and "have" and variations thereof herein, does not imply that there are no additional elements or steps. The term "coupled" as used herein, means the joining of two members together such that the members together and / or one or both members can function together.

[0118] It should be understood that the sequence of the above processes is not intended to mean the execution order of the processes, and the execution order of the processes should be determined according to their functions and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0119] Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0120] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0121] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments according to actual needs.

[0122] In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit.

[0123] The above merely provides the specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0124] The foregoing merely describes certain exemplary embodiments of this application by way of illustration. Obviously, modifications and alterations can be made by those skilled in the art without departing from the spirit and scope of the application. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of the claims of the present application.

Claims

1. A time series data management method for an edge computing gateway, characterized in that: The following steps are involved: During a preset fixed time period, the communication bandwidth occupancy rate, buffer remaining rate, and scheduling thread occupancy rate of the edge computing gateway are recorded respectively, and a resource usage raw data matrix covering all time points within the fixed time period is constructed; Based on the original resource usage data matrix, the usage fluctuation slope and resource occupancy density of communication bandwidth occupancy, buffer remaining rate, and scheduling thread occupancy within a fixed time period are calculated respectively. Weighting factors are configured according to the task priority level corresponding to each resource indicator, and the resource indicators are integrated to generate a single resource anomaly representation value. In a fixed time period, the execution status information of all tasks set as high priority is extracted, the number of tasks whose waiting time exceeds the preset response threshold is identified, and their cumulative blocking time is calculated to generate the blocking coefficient of high priority tasks; The resource anomaly representation value and the high-priority task obstruction coefficient are input into a nonlinear function for mapping and fusion, and the scheduling resource anomaly occupancy index of the current scheduling cycle is obtained as the anomaly perception indicator of the dynamic scheduling system. When the abnormal occupancy index of scheduling resources exceeds the set threshold, the buffer resource reallocation mechanism is triggered to dynamically adjust the buffer quota ratio corresponding to tasks of different priorities to ensure the continuity of data writing and uploading operations of critical tasks; at the same time, the data fragments of low-priority tasks are compressed, retaining only their statistical summary information to reduce the pressure on buffer resources.

2. The time series data management method of an edge computing gateway according to claim 1, characterized in that: Within a preset fixed time period, the communication bandwidth occupancy rate, buffer remaining rate, and scheduling thread occupancy rate of the edge computing gateway are recorded respectively. The steps to construct the resource usage raw data matrix are as follows: Based on the communication behavior of the edge computing gateway during operation, a fixed sampling time interval is set to continuously sample the real-time data transmission rate of the communication channel. The communication bandwidth occupancy rate sequence is calculated by calculating the ratio of the sampling point to the theoretical maximum bandwidth. At the time point synchronized with the communication sampling, the usage status of the buffer allocated to each task type in the edge computing gateway is periodically read, and the ratio of the remaining capacity of the buffer to the total capacity at each sampling moment is calculated to form a buffer remaining rate sequence; The ratio of the number of active task scheduling threads in the edge computing gateway to the total thread pool capacity is synchronously counted to obtain the scheduling thread occupancy sequence. The communication bandwidth occupancy sequence, buffer remaining rate sequence and scheduling thread occupancy sequence are time-aligned to construct the resource usage original data matrix according to the unified timestamp.

3. The time series data management method of an edge computing gateway according to claim 1, characterized in that: The specific steps for fusing resource indicators to generate a single resource anomaly representation value based on the original resource usage data matrix are as follows: The time series occupancy data corresponding to the communication bandwidth occupancy rate, buffer remaining rate, and scheduling thread occupancy rate in the original resource usage data matrix are processed. A continuous change model is constructed using curve fitting within a fixed time period. The usage fluctuation slope of each resource type is obtained by taking the derivative of the fitting curve at each sampling point and calculating the mean of the derivatives. Based on the calculation of the fluctuation slope, we continue to use the time series occupancy data in the same period to count the cumulative time of the communication bandwidth occupancy rate, buffer remaining rate, and scheduling thread occupancy rate in the area close to the resource saturation threshold. We then calculate the resource occupancy density by ratioing the cumulative time with the total sampling time. According to the task priority system preset by the edge computing gateway, corresponding weight factors are configured for the communication bandwidth occupancy rate, buffer remaining rate and scheduling thread occupancy rate respectively, and the fluctuation slope and resource occupancy density are normalized according to the weighted fusion rule. Finally, a single resource anomaly representation value is synthesized to quantitatively describe the severity of the overall resource anomaly occupancy within a fixed time period.

4. The time series data management method of an edge computing gateway according to claim 3, characterized in that: The specific steps for configuring weight factors and integrating them to generate resource anomaly characterization values ​​based on the task priority system preset by the edge computing gateway are as follows: Based on the predefined task priority system within the edge computing gateway, a quantitative assessment is conducted on the sensitivity of various high-priority tasks to the communication bandwidth occupancy rate, buffer remaining rate, and scheduling thread occupancy rate resources. The weight factor value of each type of resource is set according to the sensitivity, and a mapping function relationship between task priority and resource utilization indicators is established; The usage fluctuation slope and resource occupancy density data of the communication bandwidth occupancy rate, buffer remaining rate, and scheduling thread occupancy rate calculated within a fixed time period are uniformly normalized. A one-to-one correspondence is established between the preset weight factor and the normalized communication bandwidth occupancy rate, buffer remaining rate and scheduling thread occupancy rate indicators, and the normalized usage fluctuation slope and normalized resource occupancy density of each type of resource are calculated in turn with weighted values ​​to finally generate a single resource anomaly characterization value.

5. The time series data management method of an edge computing gateway according to claim 1, characterized in that: The specific steps of extracting the execution status information of the high-priority task within a fixed time period and generating the high-priority task obstruction coefficient are as follows: Within a fixed time period set by the edge computing gateway, based on the task scheduling queue log and task execution status records, all tasks marked as high priority are screened out, and the waiting time data from entering the scheduling queue to starting execution is extracted. This data is compared with the preset response time threshold, and a set of high-priority tasks whose waiting time exceeds the response time threshold is identified. A list of high-priority task response timeouts is then established. For the identified high-priority task response timeout list, traverse the scheduling waiting interval of each task, count and accumulate its actual waiting time, calculate the total time that all high-priority tasks are blocked in the scheduling link within a fixed time period, and record the number of blocked tasks at the same time; The total duration of high-priority tasks being blocked is divided by the total number of high-priority tasks in the current cycle, and further combined with the average waiting tolerance time of the unit task to generate the high-priority task blocking coefficient.

6. The time series data management method of an edge computing gateway according to claim 1, characterized in that: The specific steps for mapping and fusing the resource anomaly representation value and the high-priority task obstruction coefficient into a nonlinear function to generate the scheduling resource anomaly occupancy index are as follows: After the fixed time period ends, the scheduling system of the edge computing gateway obtains the calculated resource anomaly characterization value and the high-priority task obstruction coefficient obtained based on the high-priority task scheduling status analysis, and combines the two to form a joint indicator input vector; The scheduling system calls a pre-trained nonlinear function model and calculates the resource anomaly representation value and the high-priority task obstruction coefficient in the joint indicator input vector as function independent variables; The fusion result output by the nonlinear function model is defined as the scheduling resource anomaly occupancy index of the current scheduling period, and the scheduling resource anomaly index is registered in the scheduling control module of the edge computing gateway as a dynamic anomaly perception indicator.

7. The time series data management method of an edge computing gateway according to claim 1, characterized in that: When the abnormal scheduling resource occupancy index exceeds the set threshold, the buffer resource reallocation mechanism is triggered to dynamically adjust the buffer quota ratio corresponding to tasks of different priorities. At the same time, the data fragments of low-priority tasks are compressed. The specific steps are as follows: Based on the generated abnormal scheduling resource occupancy index and the set reference threshold, the buffer pressure index of the current cycle is calculated. The calculation expression is as follows: Where, is the buffer pressure index, is the abnormal occupancy index of scheduling resources, θ ref It is the reference threshold of the abnormal scheduling resource occupancy index; Completing the buffer pressure index After calculation, according to the buffer pressure index Calculate the target quota ratio of the buffer corresponding to high-priority and low-priority tasks. The calculation expression is as follows: Where, β high is the target buffer quota ratio for high-priority tasks, β base is the basic quota ratio of high-priority tasks, δ is the quota adjustment coefficient, tanh is the hyperbolic tangent function, β low is the buffer quota ratio of low-priority tasks; Based on the obtained quota ratio and the available buffer capacity, determine the write space that high-priority tasks can occupy and the write space that low-priority tasks can retain. The calculation expression is as follows: Where C high is the high priority task buffer capacity, C avail is the current available buffer capacity, η is the compression retention factor, C low It is the buffer capacity after compression of low priority tasks.

Images