A data link selection method for cloud-edge collaborative data transmission

Abstract

The application relates to the technical field of digital information transmission, and discloses a data link selection method for cloud-edge cooperative data transmission, which comprises the following steps: obtaining the dominant degree of each sensor terminal according to the data packet and the calculation duration of the sensor terminal; obtaining the estimated calculation duration by combining the dominant degree according to the difference relationship of the data packet in the target calculation task and the historical calculation task; obtaining the event degree according to the number of the reference calculation task of the calculation task and the fluctuation of the calculation duration; obtaining the calculation load degree according to the resource occupancy rate of the calculation task and the estimated calculation duration; obtaining the event intensive degree according to the event degree of the recent calculation task; obtaining the selected calculation unit according to the event degree and the event intensive degree; and transmitting the to-be-calculated task to the corresponding selected calculation unit and performing link transmission. The to-be-calculated task is processed by the selected calculation unit, so that the efficiency of the edge calculation unit in processing the calculation task is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital information transmission, and in particular to a data link selection method for cloud-edge collaborative data transmission. BACKGROUND

[0002] With the rapid development of clean energy and ecological environmental protection industry, the industrial control system puts forward higher requirements on the intelligentization of data acquisition, the long-distance of transmission, the high security of safety and the flexibility of system deployment. At present, industrial information is mainly collected through information physical system, so as to realize the perception of industrial process. The information physical system is mainly composed of sensor terminal, edge computing unit and long-distance communication base station. The edge computing unit has high integration and low cost computing capability, can directly process the field data, and upload the processed data to the long-distance communication base station, so as to realize intelligent perception.

[0003] In the prior art, in the process from the sensor terminal to the edge computing unit, the traditional single link or static scheduling mechanism is often used. However, with the rapid increase of the number of sensors in the current clean energy and ecological environmental protection industry and the increase of data processing complexity, the single link and static scheduling mechanism will cause the shortage of computing resources of the edge computing unit, and cannot meet the requirements of real-time and high reliability. SUMMARY

[0004] The present application provides a data link selection method for cloud-edge collaborative data transmission, which solves the problem of insufficient real-time of existing edge computing unit processing computing task. The technical scheme adopted is as follows:

[0005] The present application provides a data link selection method for cloud-edge collaborative data transmission, which solves the problem of insufficient real-time of existing edge computing unit processing computing task. The technical scheme adopted is as follows:

[0006] The CPU occupancy rate and the calculated time length of each computing task in each edge computing unit are obtained, a plurality of historical computing tasks and their calculation time length and CPU occupancy rate are obtained, a to-be-calculated task is obtained, and the computing task contains data packets of a plurality of sensor terminals;

[0007] Any one computing task in any one edge computing unit is recorded as a target computing task; according to the change relationship between the data packets of the sensor terminal and the increase amount of the calculation time length in the historical computing task, the reference computing task of the target computing task is obtained, and the dominant degree of each sensor terminal in the sensor terminal corresponding to the data packet in the target computing task is obtained; according to the difference relationship between the data packets in the target computing task and the historical computing task, the estimated calculation time length of the target computing task is obtained in combination with the dominant degree;

[0008] According to the reference computing task quantity and the fluctuation of the computing duration of the computing task, an event degree of each computing task is obtained; the resource occupancy rate of each computing task in the edge computing unit is obtained, and the computing load degree of each edge computing unit is obtained in combination with the estimated computing duration of the computing task; the event intensive degree is obtained according to the event degree of the recent computing task; the selected computing unit of the to-be-computed task is obtained according to the event degree and the event intensive degree of the to-be-computed task;

[0009] The to-be-computed task is transmitted to the corresponding selected computing unit and link transmission is performed.

[0010] Further, the reference computing task of the target computing task and the dominant degree of each sensor terminal of the sensor terminal corresponding to all data packets in the target computing task are obtained according to the change relationship between the data packet and the increase amount of the computing duration of the sensor terminal in the historical computing task, and the specific method comprises:

[0011] The computing task of the sensor terminal corresponding to the data packet in the target computing task in the historical computing task is recorded as the reference computing task of the target computing task;

[0012] Any two reference computing tasks of the target computing task are recorded as a pair of reference task sets, the reference computing task with the longest computing duration in the pair of reference task sets is recorded as the first reference computing task, and the reference computing task with the shortest computing duration is recorded as the second reference computing task;

[0013] For any one sensor terminal with the same data packet source in the pair of reference task sets, the data increment degree of the sensor terminal is obtained according to the size difference of the data packet of the sensor terminal in the first reference computing task and the second reference computing task;

[0014] The difference value obtained by subtracting the computing duration of the second reference computing task from the computing duration of the first reference computing task is recorded as the duration increment degree of the pair of reference task sets;

[0015] For any one sensor terminal in any pair of reference task sets of the target computing task, the ratio of the data increment degree of the sensor terminal to the maximum value of the data increment degrees of all sensor terminals except the sensor terminal in the pair of reference task sets is obtained, and the linear rectification result of the difference value between the ratio and 1 is recorded as the dominant judgment weight of the pair of reference task sets at the sensor terminal;

[0016] The dominant degree of the sensor terminal in the sensor terminal corresponding to all data packets in the target computing task is obtained according to the dominant judgment weight of all reference task sets of the target sensor at the sensor terminal and the duration increment degree of the reference task set.

[0017] Further, the dominant degree of each sensor terminal corresponding to all data packets in the target computing task is obtained according to the dominant judgment weight of each reference task set in the sensor terminal and the time length increment degree of the reference task set, and the specific obtaining method is as follows:

[0018] The weight normalization result of the dominant judgment weight of each reference task set in each sensor terminal is weighted and summed as the weight of the time length increment degree of the corresponding reference task set, to obtain the dominant degree of each sensor terminal corresponding to all data packets in the target computing task.

[0019] Further, the estimated computing time length of the target computing task is obtained according to the difference relationship of data packets in the target computing task and historical computing tasks, in combination with the dominant degree, and the specific method includes:

[0020] The computing method of the time length similarity of the target computing task and the first reference computing task is as follows:

[0021]

[0022] In the formula, is the time length similarity of the target computing task and the first reference computing task; is the number of sensor terminals corresponding to all data packets in the target computing task; is the dominant degree of the first sensor terminal in the sensor terminals corresponding to all data packets in the target computing task; is the size of the data packet of the first sensor terminal in the sensor terminals corresponding to all data packets in the target computing task; is the size of the data packet of the first sensor terminal in the sensor terminals corresponding to all data packets in the first reference computing task of the target computing task; is an exponential function with a natural constant as the base; is a linear normalization function;

[0023] The estimated computing time length of the target computing task is obtained according to the time length similarity of the target computing task and the reference computing task and the computing time length.

[0024] Further, the estimated computing time length of the target computing task is obtained according to the time length similarity of the target computing task and the reference computing task and the computing time length, and the specific method includes:

[0025] The weight normalized result of the similarity of the target computing task and the time length of each reference computing task of the target computing task is weighted and summed as a weight of the computing time length of the corresponding reference computing task, to obtain an estimated computing time length of the target computing task.

[0026] Further, the event degree of each computing task is obtained according to the number of reference computing tasks of the computing task and the fluctuation of the computing time length, and the specific obtaining method is:

[0027]

[0028] In the formula, is the event degree of the computing task; is the variance of the computing time length of all reference computing tasks of the computing task; is the number of reference computing tasks of the computing task; is the maximum value of the number of reference computing tasks of all computing tasks in the historical computing tasks; is a linear normalization function.

[0029] Further, the computing load degree of each edge computing unit is obtained by combining the estimated computing time length of the computing task and the resource occupancy rate of each computing task in the edge computing unit, and the specific method includes:

[0030] The CPU occupancy rate of each computing task running in each edge computing unit is obtained as the resource occupancy rate of each computing task;

[0031] For any one computing task in the input cache queue, the mean value of the CPU occupancy rate of all reference computing tasks of the computing task is recorded as the resource occupancy rate of the computing task;

[0032] For any one edge computing unit, the computing tasks running in the edge computing unit and the computing tasks in the input cache queue are recorded as the load tasks of the edge computing unit;

[0033] For any one load task of the edge computing unit, the difference between the estimated computing time length and the calculated time length of the load task is obtained, and the product of the difference and the resource occupancy rate of the load task is recorded as the estimated occupancy index of the load task;

[0034] The sum of the estimated occupancy indexes of all load tasks of the edge computing unit is recorded as the computing load degree of the edge computing unit.

[0035] Further, the event intensive degree is obtained according to the event degree of the recent computing task, and the specific obtaining method is:

[0036] For any one computing task in the near one hour, the product of the linear normalization result of the event degree of the computing task and the event degree of the computing task is recorded as the event performance degree of the computing task;

[0037] The sum of the event performance degrees of all computing tasks in the near one hour is recorded as the event intensive degree.

[0038] Further, the method for obtaining the selected computing unit of the to-be-computed task according to the event degree and the event intensive degree of the to-be-computed task comprises the following specific steps:

[0039] All edge computing units are sorted in descending order according to the corresponding computing load degree, and a load descending sequence is obtained;

[0040] The difference obtained by subtracting the event degree of the to-be-computed task from 1 is recorded as the proportional adjustment base of the to-be-computed task;

[0041] The difference obtained by subtracting the linear normalization result of the event intensive degree from 1 is recorded as the proportional adjustment degree of the to-be-computed task;

[0042] The product of the proportional adjustment base and the proportional adjustment degree of the to-be-computed task is recorded as the proportional adjustment amplitude of the to-be-computed task;

[0043] The sum of the event degree and the proportional adjustment amplitude of the to-be-computed task is recorded as the preferred sorting proportion of the to-be-computed task;

[0044] The selected computing unit of the to-be-computed task is obtained according to the preferred sorting proportion of the to-be-computed task.

[0045] Further, the method for obtaining the selected computing unit of the to-be-computed task according to the preferred sorting proportion of the to-be-computed task comprises the following specific steps:

[0046] The upper limit result of the product of the number of edge computing units in the load descending sequence and the preferred sorting proportion of the to-be-computed task is recorded as the computing allocation serial number of the to-be-computed task;

[0047] The edge computing unit in the load descending sequence is recorded as the selected computing unit of the to-be-computed task; wherein, is the computing allocation serial number of the to-be-computed task.

[0048] ​The beneficial effects of the present application are: in the process of linking the sensor terminal to the edge computing unit, the single link and the static scheduling mechanism can cause the computing resource of the edge computing unit to be insufficient, so it is necessary to dynamically select the edge computing unit for the to-be-computed task, when selecting the edge computing unit, firstly, the operation state of each edge computing unit needs to be acquired, the present application obtains the estimated computation time of the target computing task through the difference relationship of the data packets in the target computing task and the historical computing task, and combines the leading degree, and then obtains the computation load degree of each edge computing unit in combination with the resource occupancy rate of the computing task; when selecting the edge computing unit, if there are more computing tasks triggered by recent events, since the business priority of the computing task triggered by the event is higher, the computing resource should be reserved for the computing task triggered by the event, the present application obtains the selection computing unit of the to-be-computed task through the event degree and the event intensity of the to-be-computed task, and selects the data link of the to-be-computed task. Thus, the present application selects the computing unit to process the to-be-computed task, and improves the efficiency of the edge computing unit in processing the computing task. BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only show some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0050] Figure 1 A data link selection method flow chart for cloud edge collaborative data transmission provided by an embodiment of the present application. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0052] Please refer to Figure 1 which shows a data link selection method flow chart for cloud edge collaborative data transmission provided by an embodiment of the present application, the method comprises the following steps:

[0053] Step S001, acquire the CPU occupancy rate and the computed time length of each computing task in each edge computing unit; acquire a plurality of historical computing tasks and their computation time length and CPU occupancy rate; acquire a to-be-computed task.

[0054] It should be noted that in the sensing process of the information physical system, the sensor terminal first needs to collect corresponding data to obtain a plurality of data packets, and then transmit the data packets to the edge gateway, the upstream of which is linked to a plurality of edge computing units, and the edge gateway needs to transmit the data packets to the edge computing unit with relatively rich computing resources, so as to realize the selection of data link and improve the data processing timeliness.

[0055] It should be further pointed out that in the information physical system of the hydropower station, in order to ensure safe operation of equipment, optimize power generation efficiency, realize real-time monitoring and intelligent control, the running state of equipment, the water structure and liquid level, and the environmental safety need to be sensed.

[0056] Specifically, the sensor terminal in the information physical system of the hydropower station includes but is not limited to:

[0057] vibration sensors, temperature sensors and pressure sensors for obtaining the running state of equipment;

[0058] liquid level sensors, flow sensors and pressure head sensors for obtaining the water structure and liquid level;

[0059] seepage sensors, temperature rise sensors and smoke sensors for obtaining environmental safety.

[0060] It should be noted that in the running process of the information physical system, the sensor terminal generates raw data through periodic or event-driven driving.

[0061] Specifically, for any sensor terminal, the data collected by the built-in acquisition controller of the sensor terminal is preliminarily packaged, compressed and encrypted to obtain a plurality of data packets, and then the data packets are transmitted to the edge gateway; wherein the packaging, compression and encryption processing are prior art, and the specific method is not introduced here.

[0062] It should be noted that the edge gateway serves as an intelligent scheduling node, responsible for receiving data packets of a plurality of sensor terminals, and according to the type of computing task, data aggregation is performed to obtain a to-be-computed task, which is then transmitted to the input buffer queue of the edge computing unit, waiting for the analysis and processing of the computing module of the subsequent edge computing unit.

[0063] Specifically, the CPU occupancy rate and the calculated duration of each computing task running in each edge computing unit are obtained by using the edge gateway, and the computing tasks in the input buffer queue of each edge computing unit are obtained.

[0064] The to-be-computed task is obtained by using the edge gateway; the to-be-computed task contains data packets of a plurality of sensor terminals;

[0065] Obtain a plurality of historical computing tasks and their computing time length and CPU occupancy rate by using the edge gateway.

[0066] In step S002, any one computing task in any one edge computing unit is recorded as a target computing task; according to the change relationship between the data packet of the sensor terminal and the increase amount of the computing time length in the historical computing task, a reference computing task of the target computing task is obtained, and the dominant degree of each sensor terminal in the sensor terminal corresponding to the data packet in the target computing task is obtained; according to the difference relationship between the data packet in the target computing task and the historical computing task, the estimated computing time length of the target computing task is obtained in combination with the dominant degree.

[0067] It should be noted that when the edge gateway selects the edge computing unit for the to-be-computed task, the computing state of each edge computing unit being performed needs to be judged, so as to ensure the timeliness of data processing when the computing task is allocated, and at the same time, the resource utilization is maximized.

[0068] It should be further noted that when the computing state of each edge computing unit being performed is judged, the computing time length required by each computing task needs to be predicted. The influence of data of various sensor terminals on the processing flow is significantly different when participating in edge computing. Since the edge computing unit usually performs a light-weight computing task with clear structure and low complexity relative to the core processing unit, the main influencing factor of the task execution time length is the data amount collected by the sensor terminal. The data of some sensor terminals has a much higher influence on the overall computing time length than other auxiliary sensors because the data amount required for computing is large, so it is necessary to judge the dominant sensor in the sensor terminal involved in each task type.

[0069] Specifically, any one computing task in any one edge computing unit is recorded as a target computing task, wherein the computing task in the edge computing unit includes a computing task being run and a computing task in the input buffer queue, and the computing task is recorded as the target computing task. The same computing task as the data packet in the target computing task in the historical computing task is recorded as the reference computing task of the target computing task.

[0070] Any two reference computing tasks of the target computing task are recorded as a pair of reference task sets, the reference computing task with the longest computing time length in the pair of reference task sets is recorded as the first reference computing task, and the reference computing task with the shortest computing time length is recorded as the second reference computing task; any one sensor terminal with the same data packet source in the pair of reference task sets is obtained, and the data increment degree of the sensor terminal is calculated as follows:

[0071]

[0072] wherein, is the data increment degree of the sensor terminal; is the size of the data packet of the sensor terminal in the first reference computing task; is the size of the data packet of the sensor terminal in the second reference computing task; is a linear rectifier function.

[0073] The difference between the computation duration of the first reference computing task and the computation duration of the second reference computing task is denoted as the duration increment degree of the pair of reference task sets.

[0074] It is to be noted that, if is larger, it means that the data amount of the sensor terminal in the first reference computing task is more than that in the second reference computing task; and the larger the duration increment degree of the pair of reference task sets is, the longer the computation duration of the first reference computing task is relative to the second reference computing task.

[0075] It is to be further noted that, in a certain pair of reference task sets of the target computing task, if the data increment degree of a certain sensor terminal is larger while the data increment degrees of other sensor terminals are smaller, the size of the duration increment degree of the pair of reference task sets reflects the dominance of the sensor terminal on the computation duration.

[0076] Further, for any pair of reference task sets of the target computing task, the dominance judgment weight of the pair of reference task sets on the first sensor terminal is calculated as:

[0077]

[0078] wherein, is the dominance judgment weight of the pair of reference task sets on the first sensor terminal; is the data increment degree of the pair of reference task sets on the first sensor terminal; is the maximum value of the data increment degrees of the pair of reference task sets on all sensor terminals except the first sensor terminal; is a linear rectifier function.

[0079] It is to be noted that, is larger, it means that the data increment of the pair of reference task sets on the first sensor terminal is larger than that on other sensor terminals, and it can better reflect the dominance of the first sensor terminal on the computation duration.

[0080] Furthermore, the first sensor terminal corresponding to all data packets within the target computation task... The method for calculating the dominance of each sensor terminal is as follows:

[0081]

[0082] In the formula, For the target calculation task, the sensor terminal corresponding to all data packets is the first one. The degree of dominance of individual sensor terminals; Calculate the number of reference task sets for the target task; For the target sensor's first For the reference task set in the 1st The dominant judgment weight of each sensor terminal; For the target sensor's first The degree of duration increment for the reference task set; The weight normalization function is used to normalize the set of all reference tasks of the target sensor in the th order. The dominant judgment weight of each sensor terminal.

[0083] It should be noted that, if The larger the value, the higher the value of the target sensor. For the reference task set in the 1st The more individual sensor terminals can demonstrate whether computation time is affected by the first... The impact of individual sensor terminals, if simultaneously The larger the value, the greater the change in computation time. The more likely a sensor terminal is to be the dominant sensor in terms of computation time.

[0084] It should be noted that after obtaining the dominance level of each sensor terminal, if the size of the data packets of the target computing task and its reference computing task are more similar on the sensor terminal with a higher dominance level, the corresponding computing time will also be more similar.

[0085] Specifically, the target computation task and its first The method for calculating the duration similarity of the reference computing tasks is as follows:

[0086]

[0087] In the formula, The target computation task and its first Duration similarity of the reference computing tasks; To calculate the number of sensor terminals corresponding to all data packets in the target task; For the target calculation task, the sensor terminal corresponding to all data packets is the first one. The degree of dominance of individual sensor terminals; For the target calculation task, the sensor terminal corresponding to all data packets is the first one. The size of the data packet for each sensor terminal; The first task of calculating the objective is... The sensor terminal corresponding to all data packets within the reference computing task. The size of the data packet for each sensor terminal; It is an exponential function with the natural constant as its base; It is a linear normalization function, and the normalization object is the degree of dominance of the sensor terminal corresponding to all data packets in the target computing task.

[0088] It should be further explained that by using the similarity in duration to the reference computing task as a weight, the computing time of the target computing task can be estimated.

[0089] Specifically, the estimated computation time for the target computation task is calculated as follows:

[0090]

[0091] In the formula, Calculate the estimated computation time for the target task; The number of reference computational tasks for the target computational task; The target computation task and its first Duration similarity of the reference computing tasks; The first task of calculating the objective is... The computation time of each reference computing task; This is the weight normalization function, which normalizes the time similarity between the target computation task and all its reference computation tasks.

[0092] Step S003: Based on the number of reference computing tasks and the fluctuation of computing time, obtain the event severity of each computing task; obtain the resource utilization rate of each computing task in the edge computing unit, and combine it with the estimated computing time of the computing task to obtain the computing load of each edge computing unit; obtain the event density based on the event severity of recent computing tasks; and obtain the selected computing unit for the task to be computed based on the event severity and event density of the task to be computed.

[0093] It should be noted that the triggering mode of the computing task includes event triggering and periodic triggering. The periodic triggering refers to that the sensor terminal automatically collects and uploads data according to a preset fixed time interval, and has the characteristics of stable data generation and controllable frequency. The event triggering refers to that data is actively collected and uploaded when a specific condition is met or an abnormal state occurs, and has the characteristics of strong burst, irregular data and high business priority. When the edge gateway selects the edge computing unit for the to-be-computed task, the type of the computing task needs to be considered.

[0094] It should be further noted that the periodic triggering computing task usually has a short and stable computing time, and the average computing task duration fluctuation is small in long-term observation, which is embodied as low variance, and the task triggering frequency is high. The event-triggered computing task has the characteristics of long average computing duration and relatively complex task processing process, accompanied by significant computing duration fluctuation, and the triggering frequency is relatively low.

[0095] For any computing task, the calculation method of the event degree of the computing task is as follows:

[0096]

[0097] In the formula, is the event degree of the computing task; is the variance of the computing duration of all reference computing tasks of the computing task; is the number of reference computing tasks of the computing task; is the maximum value of the number of reference computing tasks of all computing tasks in the historical computing task; is a linear normalization function, and the normalization object is the variance of the computing duration of all reference computing tasks of each computing task.

[0098] It should be noted that, The greater the value is, the less frequent the reference computing task of the computing task occurs, and the more likely it is an event-triggered computing task; The greater the value is, the greater the computing duration fluctuation of the reference computing task of the computing task is, and the more likely it is an event-triggered computing task.

[0099] It should be noted that when the edge gateway uploads the to-be-computed task to the edge computing unit, the current load of each edge computing unit needs to be evaluated, so as to select the edge computing unit for uploading the to-be-computed task according to the current load.

[0100] Specifically, the CPU occupancy rate of each computing task running in each edge computing unit is obtained as the resource occupancy rate of each computing task;

[0101] For any computation task in the input buffer queue, the average CPU utilization of all reference computation tasks of that computation task is denoted as the resource utilization of that computation task.

[0102] For any edge computing unit, the computing tasks currently running in the edge computing unit and the computing tasks in the input cache queue are denoted as the load tasks of the edge computing unit.

[0103] For any load task in the edge computing unit, the difference between the estimated computation time and the calculated time is obtained. The product of this difference and the resource utilization rate of the load task is recorded as the estimated utilization index of the load task.

[0104] The sum of the estimated occupancy indices of all load tasks in the edge computing unit is denoted as the computing load level of the edge computing unit.

[0105] As an example, the first The calculation method for the computing load of each edge computing unit is as follows:

[0106]

[0107] In the formula, For the first The computational load of each edge computing unit; For the first The number of load tasks per edge computing unit; For the first The first edge computing unit Resource utilization of each load task; For the first The first edge computing unit Estimated computation time for each load task; For the first The first edge computing unit The calculated time for each load task is shown. It should be noted that the calculated time for computation tasks in the input buffer queue is 0.

[0108] It should be noted that the greater the computational load of the edge computing unit, the longer the edge computing unit needs to run under high load. If the task to be computed is assigned to the edge computing unit, the processing time of the task will be increased.

[0109] It needs to be further explained that the event triggered task is often concentrated in the burst scene, because the event triggered task is usually caused by a certain type of burst physical phenomenon or system state anomaly, such as equipment failure, environmental change, water level anomaly or safety alarm, when these events occur, multiple related or adjacent sensors will almost simultaneously meet the trigger condition, thereby generating more event triggered tasks in the near future, so the event intensity is calculated.

[0110] Specifically, for any one computing task in the past one hour, the product of the linear normalization result of the event degree of the computing task and the event degree of the computing task is recorded as the event performance degree of the computing task.

[0111] The sum of the event performance degrees of all computing tasks in the past one hour is recorded as the event intensity.

[0112] As an example, the calculation method of the event intensity is as follows:

[0113]

[0114] In the formula, is the event intensity; is the number of all computing tasks in the past one hour; is the event degree of the i-th computing task in the past one hour; is the event degree of the i-th computing task in the past one hour; is a linear normalization function, and the object of normalization is the event degree of all computing tasks in the past one hour.

[0115] It needs to be explained that since the event intensity is mainly concerned when measuring the event intensity, the greater the event degree of the computing task, the greater the weight of the computing task in the event intensity. The greater the event degree of the i-th computing task in the past one hour, the greater the weight of the i-th computing task in the recent event intensity.

[0116] It needs to be explained that when selecting the edge computing unit for the computing task, the computing load degree of the edge computing unit needs to be allocated.

[0117] Specifically, all edge computing units are sorted in descending order according to the corresponding computing load degree, and a load descending sequence is obtained;

[0118] The difference obtained by subtracting the event degree of the computing task from 1 is recorded as the proportion adjustment base of the computing task;

[0119] The difference obtained by subtracting the linear normalization result of the event intensity from 1 is recorded as the proportion adjustment degree of the computing task;

[0120] ​The product of the adjustment base and the adjustment degree is denoted as the adjustment range of the proportion of the to-be-computed task.

[0121] The sum of the event degree and the adjustment range of the proportion of the to-be-computed task is denoted as the preferred sorting proportion of the to-be-computed task.

[0122] As an example, the preferred sorting proportion of the to-be-computed task is calculated as follows:

[0123]

[0124] In the formula, is the preferred sorting proportion of the to-be-computed task; is the event degree of the to-be-computed task; is the event intensity; is a linear normalization function, and the normalization object is all event intensities in the historical computing process of each edge computing unit.

[0125] It is to be noted that, The greater the value is, the more likely the to-be-computed task is an event-triggered computing task, and since the business priority of the event-triggered computing task is high, a later edge computing unit in the load descending sequence should be selected, thereby improving the processing efficiency of the to-be-computed task; when the value is smaller, The smaller the value is, the more likely the to-be-computed task is a periodic computing task, and if at the same time The greater the value is, the less likely there are a large number of intensive event-triggered computing tasks in the recent period, and in this case, a later edge computing unit in the load descending sequence should be selected; when the value is smaller, The smaller the value is, The smaller the value is, the more likely the to-be-computed task is a periodic computing task, but a large number of intensive event-triggered computing tasks have occurred in the recent period, and sufficient computing resources should be reserved for the event-triggered computing tasks, so a later edge computing unit in the load descending sequence should be selected for the event-triggered computing task.

[0126] Further, the upward rounding result of the product of the number of edge computing units in the load descending sequence and the preferred sorting proportion of the to-be-computed task is denoted as the computing allocation serial number of the to-be-computed task.

[0127] The edge computing unit in the load descending sequence is denoted as the selected computing unit of the to-be-computed task; wherein, is the computing allocation serial number of the to-be-computed task.

[0128] Step S004: The to-be-computed task is transmitted to the corresponding selected computing unit and link transmission is performed.

[0129] ​It should be noted that after obtaining the selected computing unit of the to-be-computed task, the data packets in the to-be-computed task need to be transmitted to the selected computing unit for operation processing.

[0130] Specifically, the data packets in the to-be-computed task are transmitted to the input buffer queue of the selected computing unit, and then the subsequent edge computing unit calls the computing module for analysis and processing, and then the processing result is transmitted to the remote communication base station, so as to realize remote monitoring, data storage and global scheduling control.

[0131] It should be noted that by selecting the edge computing unit in the link process of the sensor terminal, the edge gateway, the edge computing unit and the remote communication base station, the processing efficiency of the computing task is improved.

[0132] The embodiment adopts a model to present an inverse proportional relationship and normalization processing, as the input of the model, the implementer can set the inverse proportional function and the normalization function according to the actual situation.

[0133] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A data link selection method for cloud-edge collaborative data transmission, characterized in that, The method comprises the following steps: Obtaining the CPU occupancy rate and the calculated duration of each computing task in each edge computing unit; obtaining a plurality of historical computing tasks and the computing duration and CPU occupancy rate thereof; obtaining a to-be-calculated task; the computing task comprises a plurality of data packets of sensor terminals; Regarding any one computing task in any one edge computing unit as a target computing task; obtaining the reference computing task of the target computing task and the dominant degree of each sensor terminal corresponding to all data packets in the target computing task according to the change relationship between the data packets of the sensor terminals and the increase amount of the computing duration in the historical computing tasks; obtaining the estimated computing duration of the target computing task according to the difference relationship between the data packets in the target computing task and the historical computing tasks and in combination with the dominant degree; Obtaining the event degree of each computing task according to the number of the reference computing tasks of the computing task and the fluctuation of the computing duration; obtaining the resource occupancy rate of each computing task in the edge computing unit and the computing load degree of each edge computing unit in combination with the estimated computing duration of the computing task; obtaining the event intensive degree according to the event degree of the recent computing task; obtaining the selected computing unit of the to-be-calculated task according to the event degree and the event intensive degree of the to-be-calculated task; Transmitting the to-be-calculated task to the corresponding selected computing unit and performing link transmission. 2.The method for data link selection for cloud-edge collaborative data transmission according to claim 1, wherein, The specific method for obtaining the reference computing task of the target computing task and the dominant degree of each sensor terminal corresponding to all data packets in the target computing task according to the change relationship between the data packets of the sensor terminals and the increase amount of the computing duration in the historical computing tasks comprises the following steps: Regarding the computing task of the same sensor terminal corresponding to the data packet in the target computing task in the historical computing task as the reference computing task of the target computing task; Regarding any two reference computing tasks of the target computing task as a pair of reference task sets, regarding the reference computing task with the longest computing duration in the pair of reference task sets as the first reference computing task and the reference computing task with the shortest computing duration as the second reference computing task; For any one sensor terminal with the same data packet source in the pair of reference task sets, obtaining the data increment degree of the sensor terminal according to the difference between the sizes of the data packets of the sensor terminal in the first reference computing task and the second reference computing task; Regarding the difference between the computing duration of the first reference computing task and the computing duration of the second reference computing task as the duration increment degree of the pair of reference task sets; For any one sensor terminal in any pair of reference task sets of the target computing task, obtaining the ratio of the data increment degree of the sensor terminal to the maximum value of the data increment degrees of all sensor terminals other than the sensor terminal in the pair of reference task sets, regarding the linear rectification result of the difference between the ratio and 1 as the dominant judgment weight of the pair of reference task sets in the sensor terminal. According to the dominant decision weight of all reference task sets of the target sensor in the sensor terminal and the time length increment degree of the reference task sets, a dominant degree of the sensor terminal in the sensor terminal corresponding to all data packets in the target computing task is obtained.

3. The method of claim 2, wherein, The dominant degree of the sensor terminal in the sensor terminal corresponding to all data packets in the target computing task is obtained according to the dominant decision weight of all reference task sets of the target sensor in the sensor terminal and the time length increment degree of the reference task sets, and the specific obtaining method is as follows: The weight normalization result of the dominant decision weight of each reference task set in each sensor terminal is weighted and summed as the weight of the time length increment degree of the corresponding reference task set, so as to obtain the dominant degree of each sensor terminal in the sensor terminal corresponding to all data packets in the target computing task.

4. The method of claim 1, wherein, The estimated computing time length of the target computing task is obtained according to the difference relationship of the data packets in the target computing task and the historical computing task, and in combination with the dominant degree, and the specific method includes: The target computing task and the time length similarity of its first The computing manner of the target computing task and the time length similarity of its first In the formula, The target computation task and its first Duration similarity of the reference computing tasks; To calculate the number of sensor terminals corresponding to all data packets in the target task; For the target calculation task, the sensor terminal corresponding to all data packets is the first one. The degree of dominance of individual sensor terminals; For the target calculation task, the sensor terminal corresponding to all data packets is the first one. The size of the data packet for each sensor terminal; The first task of calculating the objective is... The sensor terminal corresponding to all data packets within the reference computing task. The size of the data packet for each sensor terminal; It is an exponential function with the natural constant as its base; It is a linear normalization function; The estimated computing time length of the target computing task is obtained according to the time length similarity and the computing time length of the target computing task and its reference computing task.

5. The method of claim 4, wherein, The estimated computing time length of the target computing task is obtained according to the time length similarity and the computing time length of the target computing task and its reference computing task. The weight normalization result of the time length similarity of the target computing task and each reference computing task is weighted and summed as the weight of the computing time length of the corresponding reference computing task, so as to obtain the estimated computing time length of the target computing task.

6. The method of claim 1, wherein, The event degree of each computing task is obtained according to the number of reference computing tasks of the computing task and the fluctuation of the computing time length, and the specific obtaining method is as follows: wherein is the event degree of the computing task; is the variance of the computing duration of all reference computing tasks of the computing task; is the number of reference computing tasks of the computing task; is the maximum of the number of reference computing tasks of all computing tasks in the historical computing tasks; is a linear normalization function.

7. The method of claim 1, wherein, The computing load degree of each edge computing unit is obtained by combining the estimated computing time length of the computing task with the resource occupation rate of each computing task in the edge computing unit, and the specific method includes: The CPU occupation rate of each computing task running in each edge computing unit is obtained as the resource occupation rate of each computing task. For any one computing task in the input cache queue, the mean value of the CPU occupation rates of all reference computing tasks of the computing task is recorded as the resource occupation rate of the computing task. For any one edge computing unit, the computing tasks running in the edge computing unit and the computing tasks in the input cache queue are recorded as the load tasks of the edge computing unit. For any one load task of the edge computing unit, the product of the difference value obtained by subtracting the calculated time length from the estimated computing time length of the load task and the resource occupation rate of the load task is recorded as the estimated occupation index of the load task. The sum value of the estimated occupation indexes of all load tasks of the edge computing unit is recorded as the computing load degree of the edge computing unit.

8. The method of claim 1, wherein, The event intensive degree is obtained according to the event degree of the recent computing task, and the specific obtaining method is as follows: For any one computing task in the past one hour, the product of the linear normalization result of the event degree of the computing task and the event degree of the computing task is recorded as the event performance degree of the computing task; The sum of the event performance degrees of all computing tasks in the past one hour is recorded as the event intensive degree.

9. The method of claim 1, wherein, The specific method for obtaining the selected computing unit of the to-be-computed task according to the event degree and the event intensive degree of the to-be-computed task comprises: All edge computing units are sorted in descending order of corresponding computing load degrees to obtain a load descending sequence; The difference obtained by subtracting the event degree of the to-be-computed task from 1 is recorded as the proportional adjustment base of the to-be-computed task; The difference obtained by subtracting the linear normalization result of the event intensive degree from 1 is recorded as the proportional adjustment degree of the to-be-computed task; The product of the proportional adjustment base and the proportional adjustment degree of the to-be-computed task is recorded as the proportional adjustment amplitude of the to-be-computed task; The sum of the event degree and the proportional adjustment amplitude of the to-be-computed task is recorded as the preferred sorting proportion of the to-be-computed task; The selected computing unit of the to-be-computed task is obtained according to the preferred sorting proportion of the to-be-computed task.

10. The method of claim 9, wherein, The specific method for obtaining the selected computing unit of the to-be-computed task according to the preferred sorting proportion of the to-be-computed task comprises: The upward rounding result of the product of the number of edge computing units in the load descending sequence and the preferred sorting proportion of the to-be-computed task is recorded as the computing allocation serial number of the to-be-computed task. The first in the load descending sequence Each edge computing unit serves as a selected computing unit for the task to be computed; among them... Assign ordinal numbers to the computation tasks to be performed.

Images