A method for transforming edge computing nodes of networked devices

Abstract

The present invention discloses a method for modifying edge computing nodes of networked devices. The present invention relates to the field of edge computing technology and solves the problems of not ensuring that a processing effect of equalized computing power can be achieved among multiple edge nodes and not analyzing that different computing nodes have different load characteristics in different time periods. The present invention reconfirms and allocates computing power in different time periods based on a set of combined matching nodes. According to the numerical performance in the historical period, the total computing power can be reasonably allocated according to the load difference of each node in different time lines, ensuring that each associated node can obtain adaptive computing power resources, thereby improving computing accuracy and processing efficiency, greatly improving the modification effect of edge computing nodes, and enabling them to better cope with complex and changeable data processing needs of networked devices.

Description

Technical Field

[0001] The present invention relates to the field of edge computing technology, and in particular to a method for transforming an edge computing node of a networked device. Background Art

[0002] An edge computing node is a device or system with computing, storage, and network connection capabilities located at the edge of the network, close to the data source or user end; it can be in the form of an industrial gateway, smart router, edge server, base station, etc., distributed in factories, parks, city streets and other places close to the source of data generation, rather than concentrated in traditional cloud computing data centers.

[0003] The application with publication number CN111106946B discloses a method and system for transforming edge computing nodes of networked devices. The method includes: obtaining idle computing resource information corresponding to the networked device through the preset software in the networked device, and sending the status information of the networked device to the scheduling server through the heartbeat protocol; calculating the available idle computing resources corresponding to the networked device according to the received status information of the networked device and the preset algorithm logic; dividing the work tasks to be calculated according to the available idle computing resources of the networked device and the preset task scheduling algorithm, and assigning the divided work tasks to the corresponding networked devices. The networked device edge computing node transformation method provided by the invention transforms the networked device into the edge computing node of the scheduling server, which can effectively utilize the idle computing resources of the networked device and solve the problem of insufficient computing power of the scheduling server under large and complex tasks.

[0004] Regarding the relevant transformation process of edge computing nodes, the relevant computing power of different edge computing nodes is adjusted based on the different computing tasks and different load conditions of different edge computing nodes. However, in the actual processing process, different computing nodes have different load characteristics at different times. When adjusting the computing power, it is generally supplemented by external factors, and no reasonable optimization and allocation is performed based on the load conditions between different edge computing nodes to ensure that the computing power can be evenly quantified among multiple edge nodes. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a method for transforming the edge computing nodes of networked devices, which solves the problems of not ensuring that the computing power between multiple edge nodes can be evenly quantified and not analyzing the different load characteristics of different computing nodes in different time periods.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for transforming an edge computing node of a networked device, comprising the following steps:

[0007] Step 1: Perform numerical analysis on different edge computing nodes associated with different networked devices, identify different computing amount data associated with different historical periods from past historical completion data, thereby determining different computing averages associated with different time nodes, and generate a period average curve for the corresponding edge computing node. The specific method is as follows:

[0008] S11. Define a group of historical periods, where the historical period is a preset period. For a single group of edge computing nodes, confirm different computing amount data associated with different historical periods from the historical completion data, and then average several groups of computing amount data associated with the same period moment in different historical periods to confirm the computing average associated with the corresponding moment of the corresponding historical period.

[0009] S12, based on different calculated average quantities confirmed by the corresponding edge computing node at different times, according to their time sequence, generate a period average quantity curve associated with the corresponding edge computing node, and for different edge computing nodes associated with other networked devices, use the same processing method to confirm the period average quantity curve associated with the corresponding edge computing node;

[0010] Step 2: Based on the different period average quantity curves associated with different edge computing nodes and the numerical display of different period average quantity curves, several groups of edge computing nodes are randomly combined, and the associated several groups of different period average quantity curves are confirmed by the period bus. Based on the performance status of the period bus, a set of nodes suitable for combination is selected. The specific method is as follows:

[0011] S21, randomly combining different period average curves associated with several groups of edge computing nodes, and executing several combination processes: each combination process includes all period average curves, and each combination process has different combination sequences, and the combination sequences confirmed in each combination process cannot be completely the same;

[0012] S22. Based on the confirmed groups of combined processes, a group of combinations is randomly selected for processing: the combined sequence associated in the combined process is taken as the pending sequence, the periodic average curve associated with different edge computing nodes in the pending sequence is extracted, and the calculation average associated at the same time is averaged, the characteristic average is confirmed, and based on the different characteristic averages associated at different times, a characteristic average curve belonging to the pending sequence is generated, a group of standard lines are generated in the characteristic average curve, the standard lines are parallel to the X-axis, and the characteristic average associated with the standard lines is Y1, where Y1 is a preset value, representing the standard calculation amount, and the part of the line segment exceeding Y1 in the characteristic average curve is calibrated as a load line segment;

[0013] Simultaneously confirm the load line segments associated with other pending sequences in this combined process in turn, and calibrate the total length of several load line segments confirmed in this combined process as the characteristic line length;

[0014] S23, in the same manner as step S22, the characteristic line lengths associated with other combined processes are confirmed in turn, and the minimum value is selected from the confirmed groups of characteristic line lengths, and the combined process associated with the minimum value is marked as the standard process, and the combined sequence associated with the standard process is used as the determined combined adaptation node set;

[0015] Step 3: Based on the determined set of combined adaptation nodes, reconfirm the period average curves associated with different edge computing nodes in the set. Based on the numerical performance of different time periods in the corresponding historical period, confirm and execute the allocated computing power of different edge computing nodes in the corresponding time period. The specific sub-steps are as follows:

[0016] S31, according to the determined combined adaptation node set, mark the edge computing nodes associated in the set as associated nodes, mark the period average curve associated with the associated node as an associated curve, and divide the time period associated in the associated curve into n groups of time lines, where n is a preset value;

[0017] S32, confirming the partial line segments associated with each time line from different associated curves, and performing average processing on several groups of calculated averages associated with the partial line segments to confirm the line segment feature value T i , where i represents different partial line segments, and the different line segment feature values ​​T associated with different partial line segments i Perform ratio processing to determine the ratio sequence, and then divide the total computing power parameters based on the total computing power parameters covered by several groups of associated nodes according to this ratio sequence, and distribute the divided computing power to different associated nodes associated with different parts of the line segments, and use the divided computing power as the execution computing power of the associated nodes corresponding to this timeline;

[0018] S33. For different timelines, the same method is used to confirm the execution computing power of each different associated node of each different timeline in turn, and then execute;

[0019] Step 4: Monitor the response time of different edge computing nodes in the combined matching node set at different times in real time, and allocate computing power between edge computing nodes based on the real-time monitored response time. The specific method is as follows:

[0020] Monitor the response time of different edge computing nodes at different times in real time, and calibrate the response time of real-time monitoring as XY k , where k represents different edge computing nodes, and several groups of XY monitored in real time kPerform variance processing, confirm the variance value F, and compare the confirmed variance value F with the preset value Y2, where Y2 is the preset value:

[0021] If F>Y2, the computing power is dispatched from the edge computing node with the fastest response time to the edge computing node with the slowest response time, and it is confirmed whether the variance value F confirmed at the next moment meets the evaluation conditions. If it does, it is continuously monitored. If not, it is continuously dispatched in the same way.

[0022] If F≤Y2, continue monitoring.

[0023] Preferably, in step S22, no calibration is performed on the line segments in the characteristic mean curve that do not exceed Y1.

[0024] The present invention provides a method for modifying edge computing nodes of networked devices. Compared with the prior art, it has the following beneficial effects:

[0025] By randomly combining different period average curves and comprehensively analyzing the relationship between the characteristic average curve and the standard line, the combined process with the minimum load state can be screened out, thereby obtaining the optimal combined adaptation node set. This enables better combination and allocation of edge nodes in different states, realizes efficient integration and optimal configuration of edge computing resources, effectively avoids the situation where some nodes are overloaded while other nodes are idle, and improves the overall computing resource utilization and system stability;

[0026] Based on the combination of matching node sets, the computing power of different time periods is reconfirmed and allocated. According to the numerical performance in the historical period, the total computing power can be reasonably allocated according to the load differences of each node in different time lines, ensuring that each associated node can obtain adaptive computing resources, thereby improving computing accuracy and processing efficiency, greatly improving the transformation effect of edge computing nodes, and enabling them to better cope with the complex and changeable data processing needs of networked devices;

[0027] The response time of the edge computing nodes in the group of matching nodes at different times is monitored in real time and computing power is allocated to ensure a relatively balanced response time. By comparing the variance value with the preset value to determine whether computing power needs to be dispatched, the resource allocation of edge computing nodes can be dynamically adjusted to promptly solve the problem of low processing efficiency caused by uneven response time, further improving the performance and reliability of the entire edge computing system in real-time processing tasks, enabling the system to respond to various real-time data processing scenarios more flexibly and accurately, and providing networked devices with better quality and more efficient edge computing services. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a schematic diagram of the process of the present invention;

[0029] Figure 2 A schematic diagram is determined for the load line segment of the present invention. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] First embodiment

[0032] See also Figure 1 , the present application provides a method for transforming an edge computing node of a networked device, comprising the following steps:

[0033] Step 1: Perform numerical analysis on different edge computing nodes associated with different networked devices, identify different computing amount data associated with different historical periods from past historical completion data, thereby determining different computing averages associated with different time nodes, and generate period average curves of corresponding edge computing nodes. Specifically, the edge computing nodes associated with different networked devices are all connected. In order to facilitate the rapid response of relevant data requested by each networked device, perform numerical analysis on the edge computing nodes associated with each networked device, determine the associated computing average, and thus determine the numerical curve associated with each different edge computing node, so as to facilitate subsequent load processing analysis;

[0034] Among them, the specific method of determining the periodic average curve of different edge computing nodes is:

[0035] S11. Define a group of historical cycles, where the historical cycle is a preset cycle, which is prepared by relevant operators based on experience, and is generally 24 hours, and its cycle timeline is 0h-24h. For a single group of edge computing nodes, confirm the different computing amount data associated with different historical cycles from the historical completion data, and then average the several groups of computing amount data associated with the same cycle moment in different historical cycles (each cycle is 0-24h, so there are corresponding same cycle moments) to confirm the computing mean associated with the corresponding moment of the corresponding historical cycle (in the same edge computing node, each different historical cycle has the same moment, and several same moments correspond to several groups of computing amounts. By averaging several groups of computing amounts, the computing mean associated with the corresponding moment can be determined);

[0036] S12. Based on different calculated average quantities confirmed by the corresponding edge computing node at different times, and based on their time sequence, a period average quantity curve associated with the corresponding edge computing node is generated (the horizontal coordinate axis of the curve is the timeline, and the vertical coordinate axis is the calculated quantity). For different edge computing nodes associated with other networked devices, the same processing method is used to confirm the period average quantity curve associated with the corresponding edge computing node;

[0037] Specifically, each different networked device is associated with a group of edge computing nodes, so the computing volume of the corresponding edge computing node within a period can be numerically confirmed. Because the corresponding networked devices have data habits (that is, the amount of data requested in certain time periods is large, and the amount of data requested in certain time periods is small), the average value can be confirmed at the same time in different periods, and the period average curve associated with the corresponding edge computing node can be locked to facilitate subsequent unified adjustment and allocation.

[0038] Step 2: Based on different period average quantity curves associated with different edge computing nodes and the numerical display of different period average quantity curves, several groups of edge computing nodes are randomly combined, and the associated several groups of different period average quantity curves are confirmed by the period bus. Based on the performance status of the period bus, a group of suitable node sets is selected (the group of suitable node sets includes several groups of computing nodes. In this suitable node set, when a node has insufficient computing power due to load conditions, computing power can be directly allocated from another group of nodes. Since the computing power of another group of nodes is sufficient at the corresponding time, the nodes with such related characteristics are combined to determine the group of suitable node sets), wherein the specific method of selecting the group of suitable node sets is as follows:

[0039] S21. Randomly combine different periodic average curves associated with several groups of edge computing nodes, and execute several combination processes: each combination process includes all periodic average curves, and in each combination process, there are different combination sequences. The combination sequence, the combination sequence confirmed in each combination process cannot be completely the same (that is, there cannot be two completely identical combination processes). For example: there are five edge nodes A, B, C, D and E, then there are periodic average curves corresponding to the five edge nodes, then when performing random combination, several combination processes can be executed, and the combination sequence in each combination process is different. The first combination process is proposed to be: ABC and DE, and the other combination process can be: AB and CDE, so several groups of different combination processes can be generated, and each combination process is different;

[0040] S22, Combination Figure 2, based on the confirmed groups of combined processes, a group of combinations is randomly selected for processing: the combined sequence associated in the combined process is taken as the pending sequence, the periodic average curve associated with different edge computing nodes in the pending sequence is extracted, and the calculation average associated at the same time is averaged to confirm the characteristic average, and based on the different characteristic averages associated at different times, a characteristic average curve belonging to this pending sequence is generated, and a set of standard lines are generated in the characteristic average curve, wherein the standard line is parallel to the X-axis, and the characteristic average associated with the standard line is Y1, wherein the specific value of Y1 is prepared in advance by the operator, which is the standard calculation amount (exceeding Y1 means that the corresponding computing node is in a load state, and being lower than this calculation amount means that this computing node has excess computing power), and the part of the line segment exceeding Y1 in the characteristic average curve is marked as the load line segment;

[0041] Simultaneously confirm the load line segments associated with other pending sequences in this combined process in turn, and calibrate the total length of several load line segments confirmed in this combined process as the characteristic line length;

[0042] S23, in the same manner as step S22, the characteristic line lengths associated with other combined processes are confirmed in turn, and the minimum value is selected from the confirmed groups of characteristic line lengths, and the combined process associated with the minimum value is marked as the standard process, and the combined sequence associated with the standard process is used as the determined combined adaptation node set;

[0043] Specifically, different edge computing nodes have different computing loads. In the actual operation process, different computing loads vary in level, that is, some periods are in load state, and some periods are in optimal state. In order to make better combination and allocation of edge nodes in different states, a random combination method is adopted to randomly combine several edge computing nodes to confirm several combination processes, and perform correlation analysis on each different combination, so as to select the combination process with the minimum load state after averaging processing, so as to lock the optimal combination adaptation node set from such combination process, so as to facilitate the orderly and reasonable allocation of different computing power conditions of the combination adaptation node set.

[0044] Step 3: Based on the determined set of combined adaptation nodes, reconfirm the period average curves associated with different edge computing nodes in the set, and based on the numerical performance of different time periods in the corresponding historical period, confirm and execute the allocated computing power of different edge computing nodes in the corresponding time period. The specific sub-steps for confirmation are:

[0045] S31. According to the determined combined adaptation node set, the edge computing nodes associated in the set are calibrated as associated nodes, the period average curve associated with the associated node is calibrated as an associated curve, and the time period associated in the associated curve is divided into n groups of time lines, where n is a preset value, which is determined by relevant operators based on experience, and is generally 48, that is, the time range associated with each time line is half an hour;

[0046] S32, confirming the partial line segments associated with each time line from different associated curves, and performing average processing on several groups of calculated averages associated with the partial line segments to confirm the line segment feature value T i , where i represents different partial line segments, and the different line segment feature values ​​T associated with different partial line segments i Perform ratio processing to determine the ratio sequence. Then, based on the total computing power parameters covered by several groups of associated nodes, divide the total computing power parameters equally according to this ratio sequence, and distribute the equally divided computing power to different associated nodes associated with different parts of the line segments. Use the equally divided computing power as the execution computing power of the associated nodes corresponding to this timeline. For example, if the line segment feature values ​​associated with different parts of the line segments are 4, 6, and 10, the resulting ratio sequence is 4:6:10=2:3:5. If the total computing power parameter covered is 40, the computing power allocated to each associated node in this timeline is 8, 12, and 20. Related allocation and execution are performed to achieve better computing power allocation and processing effects, thereby improving the transformation effects of several edge computing nodes.

[0047] S33. For different timelines, the same method is used to confirm the execution computing power of each different associated node of each different timeline in turn, and then execute;

[0048] Specifically, in the actual allocation process, since different associated nodes have different load conditions in different timelines, the total computing power associated with several associated nodes in the timeline is redistributed based on the different load conditions in past historical data, so that each different associated node with different load conditions can be allocated a more relevant average computing power to complete the corresponding node transformation process, thereby ensuring that multiple edge computing nodes can achieve better computing conditions and ensure computing accuracy.

[0049] Second embodiment

[0050] In the specific implementation process of this embodiment, compared with the above embodiment, this embodiment mainly implements real-time allocation of computing power associated with the process of implementing computing processing to further improve the accuracy;

[0051] Step 4: Monitor the response time of different edge computing nodes in the group matching node set at different times in real time, and allocate computing power between edge computing nodes based on the real-time monitored response time to ensure a relatively balanced response time;

[0052] The specific sub-steps for computing power allocation are:

[0053] S41. Monitor the response time of different edge computing nodes at different times in real time, and mark the response time of real-time monitoring as XY k , where k represents different edge computing nodes, and several groups of XY monitored in real time k Perform variance processing, confirm the variance value F, and compare the confirmed variance value F with the preset value Y2, where Y2 is the preset value, and its specific value is determined by the operator based on experience:

[0054] If F>Y2, the computing power is scheduled from the edge computing node with the fastest response time to the edge computing node with the slowest response time, and it is confirmed whether the variance value F confirmed at the next moment meets the evaluation conditions. If it does, it is continuously monitored. If it does not, it is continuously scheduled in the same way (in each different scheduling process, the related nodes of the edge computing node with the fastest response time and the edge computing node with the slowest response time will change. This is to achieve the effect of balanced computing power, so that multiple edge computing nodes can achieve a relatively balanced effect when processing);

[0055] If F≤Y2, continue monitoring.

[0056] Some of the data in the above formulas are dimensionless and numerically calculated. Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0057] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A method for modifying edge computing nodes of networked devices, characterized in that: The following steps are involved: Step 1: Perform numerical analysis on different edge computing nodes associated with different networked devices, identify different computing amount data associated with different historical periods from past historical completion data, thereby determining different computing averages associated with different time nodes, and generate a period average curve for the corresponding edge computing node; Step 2: Based on the different period average quantity curves associated with different edge computing nodes and the numerical display of different period average quantity curves, several groups of edge computing nodes are randomly combined, and the associated several groups of different period average quantity curves are confirmed by the period bus. Based on the performance status of the period bus, a set of nodes suitable for combination is selected. The specific method is as follows: S21, randomly combining different period average curves associated with several groups of edge computing nodes, and executing several combination processes: each combination process includes all period average curves, and each combination process has different combination sequences, and the combination sequences confirmed in each combination process cannot be completely the same; S22. Based on the confirmed groups of combined processes, a group of combinations is randomly selected for processing: the combined sequence associated in the combined process is taken as the pending sequence, the periodic average curve associated with different edge computing nodes in the pending sequence is extracted, and the calculation average associated at the same time is averaged, the characteristic average is confirmed, and based on the different characteristic averages associated at different times, a characteristic average curve belonging to the pending sequence is generated, a group of standard lines are generated in the characteristic average curve, the standard lines are parallel to the X-axis, and the characteristic average associated with the standard lines is Y1, where Y1 is a preset value, representing the standard calculation amount, and the part of the line segment exceeding Y1 in the characteristic average curve is calibrated as a load line segment; Simultaneously confirm the load line segments associated with other pending sequences in this combined process in turn, and calibrate the total length of several load line segments confirmed in this combined process as the characteristic line length; S23, in the same manner as step S22, the characteristic line lengths associated with other combined processes are confirmed in turn, and the minimum value is selected from the confirmed groups of characteristic line lengths, and the combined process associated with the minimum value is marked as the standard process, and the combined sequence associated with the standard process is used as the determined combined adaptation node set; Step 3: Based on the determined set of combined adaptation nodes, reconfirm the period average curves associated with different edge computing nodes in the set. Based on the numerical performance of different time periods in the corresponding historical period, confirm and execute the allocated computing power of different edge computing nodes in the corresponding time period. The specific sub-steps are as follows: S31, according to the determined combined adaptation node set, mark the edge computing nodes associated in the set as associated nodes, mark the period average curve associated with the associated node as an associated curve, and divide the time period associated in the associated curve into n groups of time lines, where n is a preset value; S32, confirming the partial line segments associated with each time line from different associated curves, and performing average processing on several groups of calculated averages associated with the partial line segments to confirm the line segment feature value T i , where i represents different partial line segments, and the different line segment feature values ​​T associated with different partial line segments i Perform ratio processing to determine the ratio sequence. Then, based on the total computing power parameters covered by several groups of associated nodes, divide the total computing power parameters equally according to this ratio sequence, and distribute the equally divided computing power to different associated nodes associated with different parts of the line segments. The equally divided computing power is used as the execution computing power of the associated nodes corresponding to this timeline.

2. A method for modifying an edge computing node of a networked device according to claim 1, characterized in that: In step 1, the specific method of determining the periodic average quantity curves of different edge computing nodes is: S11. Define a group of historical periods, where the historical period is a preset period. For a single group of edge computing nodes, confirm different computing amount data associated with different historical periods from the historical completion data, and then average several groups of computing amount data associated with the same period moment in different historical periods to confirm the computing average associated with the corresponding moment of the corresponding historical period. S12. Based on the different calculated averages confirmed by the corresponding edge computing nodes at different times, and based on their time sequence, a periodic average curve associated with the corresponding edge computing nodes is generated. For different edge computing nodes associated with other networked devices, the same processing method is used to confirm the periodic average curve associated with the corresponding edge computing nodes.

3. A method for modifying an edge computing node of a networked device according to claim 1, characterized in that: In the step S22, no calibration is performed on the line segments in the characteristic mean curve that do not exceed Y1.

4. A method for modifying an edge computing node of a networked device according to claim 1, characterized in that: The step three also includes: S33. For different timelines, the same method is used to confirm the execution computing power of each different associated node of each different timeline in turn and execute them.

5. A method for modifying an edge computing node of a networked device according to claim 1, characterized in that: Also includes: Step 4: Monitor the response time of different edge computing nodes in the combined matching node set at different times in real time, and allocate computing power between edge computing nodes based on the real-time monitored response time.

6. A method for modifying an edge computing node of a networked device according to claim 5, characterized in that: In step 4, the specific method of allocating computing power between edge computing nodes is: Monitor the response time of different edge computing nodes at different times in real time, and calibrate the response time of real-time monitoring as XY k , where k represents different edge computing nodes, and several groups of XY monitored in real time k Perform variance processing, confirm the variance value F, and compare the confirmed variance value F with the preset value Y2, where Y2 is the preset value: If F>Y2, the computing power is dispatched from the edge computing node with the fastest response time to the edge computing node with the slowest response time, and it is confirmed whether the variance value F confirmed at the next moment meets the evaluation conditions. If it does, it is continuously monitored. If not, it is continuously dispatched in the same way. If F≤Y2, continue monitoring.

Images