Internet of things task scheduling method, Internet of things system and storage medium

Abstract

The present invention provides an Internet of Things task processing method and system thereof, comprising: an edge management node measures the total delay of a task data stream on each edge computing node according to a sampling period; determines the size relationship between the total delay and a third threshold and a fourth threshold, the third threshold being less than the fourth threshold; maintains the existing task data stream when the total delay is less than the third threshold; performs a task data stream selection operation when the total delay is greater than or equal to the third threshold and less than the fourth threshold; and performs a reset operation of the edge node when the total delay is greater than or equal to the fourth threshold. Through the above steps, the edge node is made suitable for processing data streams of different tasks, and on the other hand, the edge node is designed to support reconfiguration to meet the operation requirements of various tasks.

Description

Technical Field

[0001] The technical solution disclosed herein relates to the computer field, and specifically to a task allocation method and the Internet of Things technology field. Background Art

[0002] The Internet of Things is a network formed by the connection between various objective objects, in which various objects communicate information through various information sensors, inductors, scanners and other devices and technologies such as radio frequency identification. Edge computing, fog computing and cloud computing technologies are all used to support the development of IoT applications. However, the heterogeneity of the IoT and the need for collaborative reliability pose challenges to the development of related technologies.

[0003] Another requirement of IoT applications is low latency and low response time. Traditional cloud computing technology has its application functions deployed in the cloud. The process of obtaining IoT data from the bottom layer and handing it over to the cloud for processing is long and time-consuming. Therefore, a method is needed to deploy application algorithms at a location closer to the cloud to reduce latency losses.

[0004] Compared with traditional Internet of Things, Industrial Internet of Things has higher requirements for configurability and scalability, and the demand for low latency is further increased. Especially in the application environment of Industrial Internet of Things in processing production lines, high configurability and low latency mean improved production efficiency and product quality. Summary of the invention

[0005] Based on the traditional IoT sensor device layer and cloud computing layer, the present invention sets up a fog computing layer and an edge processing layer from top to bottom, respectively, and processes the data uploaded from the bottom layer at the edge processing layer closer to the cloud, and further receives the data processed by the edge processing layer through the fog computing layer, so that critical and time-sensitive information can be quickly processed at a lower level, reducing the response time. At the same time, such a structure also enables data to be processed closer to its source, and only relatively small amounts of data need to enter the cloud computing layer and be solidified, which can further reduce the bandwidth overhead in the communication link and the cost of data storage.

[0006] The present invention provides a task processing method, characterized in that:

[0007] Define data acquisition tasks, data standardization tasks, data filtering tasks, and data aggregation tasks;

[0008] The data acquisition task and the data standardization task are executed on multiple data acquisition nodes, the data filtering task is executed on multiple data filtering nodes, and the data aggregation task is executed on multiple data aggregation nodes, and the nodes are data processing devices with computing capabilities;

[0009] Obtaining delay times of the plurality of data acquisition nodes, and selecting a data acquisition node with the smallest delay time, wherein the delay time is the sum of the network transmission time and the calculation processing time;

[0010] The selected data acquisition node with the smallest delay time is used as the first node, and the calculation processing time of each remaining data acquisition node is compared with the delay time of the first node:

[0011] If the delay time of the first node is less than the calculation processing time of the data acquisition node for comparison, assigning the data filtering task to the first node;

[0012] If the delay time of the first node is greater than the computational processing time of the data acquisition node, the data filtering task is assigned to the data acquisition node according to a locality allocation principle, wherein the locality allocation principle refers to assigning the task to a location close to its data input source;

[0013] Repeat the above steps iteratively until all tasks in this stage are assigned or the termination exit conditions are given.

[0014] The task processing method further comprises:

[0015] Get the load of each data filtering node;

[0016] When the load is greater than the first threshold, the data filtering task on the data filtering node is allocated to an adjacent node, wherein the adjacent node is a data acquisition node whose network transmission time and computing processing time are both smaller than the data filtering node of the task to be allocated and is close to the data source in the data filtering task;

[0017] The above steps are repeated iteratively until all data filtering tasks in this stage are assigned or the termination exit conditions are given.

[0018] The task processing method further comprises:

[0019] The nodes processing the data filtering task are grouped, where adjacent nodes are grouped in the same group;

[0020] Selecting a node with the shortest delay time in each group of nodes, and allocating the data aggregation task from the data aggregation node with a load greater than a second threshold to the node with the shortest delay time;

[0021] The above steps are repeated iteratively until all data aggregation tasks in this stage are assigned or the termination exit condition is given.

[0022] The present invention also provides a task processing method, characterized in that:

[0023] The edge management node measures the total delay of the task data flow on each edge computing node according to the sampling period;

[0024] The total delay is the sum of the following delays: contention delay on the sensor device, transmission delay from the sensor to the edge computing node, processing delay of the edge computing node, processing delay of the edge management node, and transmission delay from the edge computing node to the actuator;

[0025] Determine the relationship between the total delay and the third threshold and the fourth threshold, the third threshold being less than the fourth threshold:

[0026] When the total delay is less than the third threshold, maintaining the task data flow of the existing edge node;

[0027] When the total delay is greater than or equal to the third threshold and less than a fourth threshold, performing a task data stream selection operation;

[0028] When the total delay is greater than or equal to the fourth threshold, performing a reset operation of the edge node;

[0029] The above steps are repeated iteratively until the total delay returns to the normal range.

[0030] The task processing method further comprises:

[0031] The performing task data stream selection operation includes obtaining the task priority of the task data stream and the resource consumption of processing the task data stream at the edge node;

[0032] Calculating the total priority of the task data flow, the total priority being a weighted sum of the task priority and the resource consumption of the task data flow;

[0033] The edge computing node checks the total priority of different received task data streams, and preferentially processes the task data stream with a high total priority;

[0034] If the total priorities of multiple task data streams are the same or the difference is less than a certain threshold, the task data streams are processed on a first-come, first-served basis.

[0035] The task processing method further comprises:

[0036] The reset operation of the edge node includes: the edge management node obtains the function priorities of different functions of the edge computing node for reset, and the edge management node sends a reset request to the edge computing node to be reset according to the highest ranking function priority:

[0037] When the edge management node receives the reset completion confirmation message returned by the edge computing node, it proves that the edge node reset is successful;

[0038] When the edge management node receives the reset rejection message returned by the edge computing node, the function that obtains the next function priority ranking continues to send a reset request;

[0039] When the edge management node does not receive any message returned by the edge node within the preset time, it sends a reset request to the edge computing node again according to the function indicated by the current highest function priority.

[0040] The present invention also provides an Internet of Things system, characterized in that:

[0041] It includes IoT sensor device layer, edge processing layer, fog computing layer, cloud computing layer, middleware layer, and data solidification layer;

[0042] The IoT sensor device layer is specifically a wireless sensor network, including multiple numbers and types of sensor nodes for autonomously collecting and generating data, and multiple numbers and types of actuators for sending instructions from the upper layer to the IoT devices for execution;

[0043] The edge processing layer is used to receive data uploaded by the IoT sensor device layer and perform preliminary processing on the data;

[0044] Multiple node devices in the fog computing layer receive the data initially processed by the edge processing layer and further perform computation on the data;

[0045] The cloud computing layer is used to receive the processing result data of the fog computing layer, and perform data analysis on the processing results or further process them according to the functional requirements of the user, and provide the processing results to the user in the form of FaaS;

[0046] The middleware layer is set between the cloud computing layer and the data hardening layer. By setting task queues and corresponding message queues, the devices in the cloud computing layer can access the data stored in the data hardening layer.

[0047] A database or file system is set up in the data curing layer to achieve persistent storage of IoT data.

[0048] The task processing method further comprises:

[0049] The edge processing layer further includes a plurality of edge computing nodes, and a plurality of edge management nodes not exceeding the number of edge computing nodes;

[0050] The edge computing node directly receives task data streams from sensors through the gateway, is reconfigurable, and has the function of executing multiple tasks;

[0051] The edge management node coordinates the overall network architecture of the edge computing layer and the direction of task data flow.

[0052] The task processing method further comprises:

[0053] The system is used to execute the task processing method described above.

[0054] The present invention also provides a computer program product, characterized in that:

[0055] The computer program product is stored in a non-volatile computer-readable storage medium and includes machine-executable program instructions;

[0056] When the machine executable program instructions are executed in a device, the device is caused to execute the task processing method described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is the overall system architecture diagram of the Internet of Things of the present invention.

[0058] Figure 2 Schematic diagram of an edge processing node and an edge management node of the present invention.

[0059] Figure 3 The present invention provides a flowchart of the first task allocation method.

[0060] Figure 4 This is a flow chart of a second task allocation method provided by the present invention. DETAILED DESCRIPTION

[0061] The technical solution of the present invention will be described in detail below in conjunction with the accompanying drawings. Although the preferred embodiments disclosed herein are described, it should be understood that the specific implementation of the various forms of the technical solution should not be limited by the preferred embodiments. The term "a possible implementation method" used in the specification and claims of the present invention means "at least one embodiment"; "first" "second" and the like are used to distinguish different objects themselves, and do not mean that there must be a logical before-after or progressive relationship.

[0062] Figure 1 An IoT system architecture is proposed, which is particularly suitable for industrial production environments, including IoT sensor device layer, edge processing layer, fog computing layer, cloud computing layer, middleware layer, and data solidification layer, among which:

[0063] IoT sensor device layer:

[0064] In a specific implementation, the IoT sensor device layer may be specifically an industrial wireless sensor network; the industrial wireless sensor network includes a plurality of sensor nodes of various numbers and types for autonomously collecting and generating data, and the sensor nodes are composed of devices that are difficult to perform complex data processing or have limited resources.

[0065] In an optional embodiment, the sensor device is powered by a battery or a low-voltage power supply, has low power consumption, supports the IEEE 802.15.4 standard, uses IPv6 as the network layer protocol, and selects UDP as the transport layer protocol. However, it is known to those skilled in the art that any protocol and standard suitable for use in a specific scenario falls within the disclosure scope of this architecture.

[0066] The IoT sensor device layer is also equipped with multiple numbers and types of actuators to send instructions from the upper layer to the IoT devices for execution; the instructions include sensor and device switch instructions, sensor and device switching instructions, sensor and device frequency conversion instructions, product rejection instructions, etc.

[0067] Edge processing layer:

[0068] The edge processing layer is used to receive data uploaded by the IoT sensor device layer and perform preliminary processing on the data. Considering the diversity and differentiation of sensor devices, the data collected and generated by the edge processing layer needs to be processed on the edge side in order to make quick decisions. Such quick decisions can reduce response time and improve data throughput.

[0069] In one possible implementation, the edge processing layer can handle the task of judging qualified products. The devices in the edge processing layer can quickly screen the data reported by the sensor and classify and mark abnormal data when reporting data to the upper-level computing device; for example, in a machining production line, the data reported by the sensor can be directly calculated and judged; such as product size, light reflectivity, elongation, etc., and the product can be directly judged whether it is qualified; when the product is judged to be unqualified, the data stream can be directly sent back to the IoT sensor device layer, so that the actuator therein can quickly remove the unqualified product.

[0070] In another possible implementation, the devices in the edge processing layer may perform sampling tasks on the data reported by the sensor, including basic simple limited deduplication or merging; for example, the data collection frequency of a sensor is once per second. If the data of the next collection cycle does not change, the data of the next collection cycle may be deduplicated.

[0071] The edge processing layer implements the first filtering of the data uploaded by IoT sensor devices and data verification including integrity. At the same time, sampling the data before reporting can reduce the size of the transmitted data, reduce latency and response time.

[0072] The edge processing layer includes multiple edge computing nodes and multiple edge management nodes no more than the number of edge computing nodes; the edge computing nodes directly receive task data streams from sensors through the gateway, are resettable, and have the function of performing multiple tasks. The edge management nodes coordinate the overall network architecture of the edge computing layer and the direction of the task data stream.

[0073] Fog computing layer:

[0074] Multiple node devices in the fog computing layer receive the data preliminarily processed by the edge processing layer, and further perform computational processing on the data; the computational processing includes data filtering algorithm, data conversion algorithm, and data aggregation algorithm.

[0075] The data filtering algorithm can be to filter out data that has little effect on the data analysis process or user production, or to filter out data that has not changed for a certain period of time; the data conversion algorithm can be the process of standardizing data from different sources, that is, the process of running the data standardization algorithm; the data aggregation algorithm can include grouping aggregation, metric aggregation and pipeline aggregation, etc. Metric aggregation can be, for example, to calculate the average value or correlation coefficient of a group of data, and pipeline aggregation can be further aggregation based on the results of other aggregation algorithms.

[0076] Data processed at the fog computing layer can also be quickly processed at this layer; for example, if a production line fault is found in the processed data, it can be quickly reported to on-site personnel to achieve rapid response and processing of the fault; the production line fault may include electrical or mechanical faults, etc.

[0077] The fog computing layer can further perform a second data filtering and various forms of data inspection based on the first data sampling and filtering. It should be noted that the fault alarm of the fog computing layer can be an abnormal alarm at a higher level than the error alarm level of the edge processing layer; for example, the edge processing layer can only judge whether each product is qualified or not, while the fog computing layer can further analyze the aggregated data and find out the rules, so as to find out the abnormal operation of a whole production line over a period of time.

[0078] Cloud computing layer:

[0079] The cloud computing layer is used to receive the processing result data from the fog computing layer, perform data analysis on the processing results or further process them according to the user's functional requirements, and provide the processing results to users in the form of FaaS. This function as a service approach can realize the construction of loosely coupled applications and solve the differential characteristics of IoT data.

[0080] These functions may include simple data visualization queries, such as the output of a certain assembly line over a certain period of time; they may also include predictions for the future, such as the prediction of the future output of a certain assembly line; the above examples are only illustrative and not exhaustive.

[0081] At the same time, the cloud computing layer can receive instructions conveyed by the user to the bottom layer, and send the instructions to the fog computing layer, the edge processing layer, and the Internet of Things sensor device layer in sequence, or directly send instructions to the edge processing layer if resource conditions permit; the situation where resource conditions permit may include that at least one device in the edge processing layer is in an idle state, and the idle state includes one or more indicators of processor utilization, memory utilization, and IO throughput being lower than a threshold.

[0082] Based on this design, the user instructions received by the cloud computing layer can reach the underlying IoT sensor devices with less delay; the instructions can specifically include switch instructions or control instructions for sensors, and the control instructions can switch the data collection object or data collection content of one or more specific sensors. In addition, further analysis of the acquired data is also carried out in the cloud computing layer. In this step, various artificial intelligence technologies such as neural networks can be used to realize intelligent analysis and processing of industrial data.

[0083] Middleware layer:

[0084] The middleware layer is set between the cloud computing layer and the data solidification layer. By setting task queues and corresponding message queues, the devices in the cloud computing layer can access the data stored in the data solidification layer. The access includes the control of specific data access policies. For example, after the cloud computing layer device sends a request message to obtain data or store data, the middleware layer can check the request message in advance. The check may include identity security verification, timeout detection, routing query, etc. These functions can decouple the cloud computing layer devices from the data solidification storage devices.

[0085] In some embodiments, the timeout detection in the above method can be performed before the request reaches the final destination, and the timeout request can be terminated before reaching the final destination, which not only saves the thread and bandwidth resources of the cloud computing layer device, but also avoids the waste of resources of the data solidification layer device, and improves the data processing capacity. As for the route query function, on the one hand, it includes the route query of a single data solidification device, and on the other hand, it also includes faster acquisition of data node information at different locations when the data solidification layer implements distributed storage.

[0086] Data curing layer:

[0087] The data curing layer sets up a database or file system in the data curing layer to achieve persistent storage of IoT data.

[0088] In one embodiment, the database of the data solidification layer uses the distributed database OceanBase, or HBase; in other embodiments, traditional relational databases and non-relational databases can also be configured and used individually, in multiple forms, or in combination according to actual conditions. It is known to those skilled in the art that any data storage paradigm suitable for use in a specific scenario falls within the scope of the disclosure of this architecture, and the possibility of further improvement and expansion of the data storage in this architecture is reserved.

[0089] Based on the above architectural design, by pushing resources to the edge of the network through the fog computing layer and buffering the data solidification layer and cloud computing layer through the middleware layer, the high latency, high data storage cost and security and privacy issues in traditional IoT cloud computing scenarios can be properly overcome. Although the computing power of fog computing layer devices is likely to be inferior to that of cloud computing layer devices, moving the computing closer to the data source can significantly reduce transmission delays and improve data throughput responsiveness, providing architectural support for rapid decision-making in the IoT.

[0090] In this embodiment, a first task allocation method is provided, which is as follows:

[0091] In the fog computing layer, data acquisition tasks, data standardization tasks, data filtering tasks and data aggregation tasks are defined; wherein the data acquisition tasks and data standardization tasks are run on multiple data acquisition nodes, the data filtering tasks are run on multiple data filtering nodes, and the data aggregation tasks are run on multiple data aggregation nodes; the nodes are data processing devices with computing capabilities.

[0092] Due to the limitations and heterogeneity of resources in the fog computing layer, if task priorities are adjusted only considering network conditions, multiple tasks will be assigned to a small number of processing nodes in a short period of time, which may cause an avalanche of network delays. Therefore, while considering reducing network delays, it is also necessary to consider reducing processing time.

[0093] The task allocation methods of the fog computing layer include:

[0094] S101: Calculate the delay time among multiple data acquisition nodes in the fog computing layer, and select the data acquisition node with the smallest delay time; wherein the delay time is the sum of the network transmission time and the computing processing time.

[0095] S102: The selected data acquisition node with the smallest delay time is used as the first node, and the calculation processing time of each remaining data acquisition node is compared with the delay time of the first node.

[0096] S103: If the delay time of the first node is less than the calculation processing time of the data acquisition node for comparison, the data filtering task is assigned to the first node; if the delay time of the first node is greater than the calculation processing time of the data acquisition node, the data filtering task is assigned to the data acquisition node according to the locality allocation principle; the locality allocation principle refers to assigning tasks to locations close to their data input sources. If the delay time is equal to the calculation processing time, those skilled in the art can arbitrarily select a method for assigning data filtering tasks based on experience.

[0097] S104: Iteratively repeat steps S101 to S103 until all data filtering tasks are assigned or a termination exit condition is given at this stage; in one embodiment, the termination exit condition may be a program interruption exit caused by an exception. Generally speaking, the number of data filtering nodes is less than that of data acquisition nodes, so transferring the data filtering tasks on the data filtering nodes to the data acquisition nodes for execution can effectively reduce the load of the data filtering nodes. The load calculation of the data filtering nodes can be temporarily ignored here, because the number of data filtering nodes is small, and it is easier to generate a pile of data filtering tasks, and calculating the load of the data filtering nodes itself is also a consumption of node resources.

[0098] S105: Obtain the load of each data filtering node; the load may include one or more indicators such as processor usage, memory usage, IO throughput, etc. As an improvement to the above steps, the load of the data filtering node is considered at this time. If the load of the data filtering node cannot be adjusted to a suitable range after the allocation of steps S101 to S103, when the load of the data filtering node is still too high, the data filtering task is allocated to the data acquisition node by directly obtaining the load value adjustment method. In this way, the data processing efficiency can be improved more accurately and the delay can be reduced.

[0099] S106: When the load is greater than the first threshold, the data filtering task on the data filtering node is allocated to an adjacent node; the adjacent node is a data acquisition node whose network transmission time and computing processing time are both less than the data filtering node of the aforementioned task to be allocated, and is close to the data source in the data filtering task. The threshold can be given based on experience or calculated based on various algorithms. After the data filtering task is allocated, the data aggregation task is allocated.

[0100] S107: Grouping the nodes processing the data filtering task, wherein adjacent nodes are grouped in the same group; the adjacent nodes refer to nodes directly connected via network communication.

[0101] S108: Select a node with the shortest delay time in each group of nodes, and assign the data aggregation task from the data aggregation node with a load greater than a second threshold to the node with the shortest delay time.

[0102] S109: Steps S107 to S108 are repeated iteratively until all data aggregation tasks in this stage are assigned or a termination exit condition is given.

[0103] S110: Iteratively repeat steps S101 to S109.

[0104] According to the first allocation method mentioned above, the operation time of data tasks is reduced by minimizing network and processing delays, and data tasks are allocated to locations close to their sources. This can reduce network delays by reducing communication distances and data transmission volume, and reduce task processing time by evacuating overloaded nodes. According to the task allocation method proposed in the previous article, the data processing efficiency in the Internet of Things environment can be effectively improved.

[0105] Although the allocation method is applied to fog computing in this embodiment, it can also be applied to cloud computing or edge computing when the nodes run corresponding data processing tasks.

[0106] Although the embodiments are applied in an industrial Internet of Things environment, they are also applicable to other environments involving the Internet of Things.

[0107] In this embodiment, a second task allocation method is provided, which is as follows:

[0108] In order to further improve the processing efficiency of the IoT system, based on the optimization of task allocation in the fog computing layer, we further consider the task processing problem of the edge computing layer from top to bottom. As described in the previous architecture, due to the diversity and differentiation of sensors, the amount and type of data collected are different. This heterogeneous characteristic makes it difficult for resource-constrained edge device nodes to maximize their performance.

[0109] Therefore, on the one hand, it is necessary to consider how to select edge nodes so that they are suitable for processing data streams of different tasks, and on the other hand, to design edge nodes to support reconfiguration to meet the operation requirements of various tasks. It is known to those skilled in the art that edge nodes include at least edge management nodes and edge computing nodes.

[0110] S201: The edge management node measures the total delay of the task data flow on each edge computing node according to the sampling period; the total delay is the sum of the following delays: contention delay on the sensor device, transmission delay from the sensor to the edge computing node, processing delay of the edge computing node, processing delay of the edge management node, and transmission delay from the edge computing node to the actuator. The contention delay refers to the delay caused by channel contention when multiple sensor nodes send data to the same or a few edge computing nodes.

[0111] S202: Determine the relationship between the total delay and a third threshold and a fourth threshold, wherein the third threshold is smaller than the fourth threshold.

[0112] S203: When the total delay is less than the third threshold, maintain the task data flow of the existing edge node.

[0113] S204: When the total delay is greater than or equal to the third threshold and less than a fourth threshold, perform a task data stream selection operation.

[0114] S205: When the total delay is greater than or equal to the fourth threshold, performing a reset operation on the edge node; resetting refers to reconfiguring the function of the node to a new function that is different from the previous function, and the reset process is mainly completed through software configuration, and can also be achieved through hardware physical structure changes or logical structure changes when necessary.

[0115] S206: iteratively repeat steps S201 to S205 until the total delay returns to a normal range; the normal range may be a situation where the total delay is less than a third threshold, or may be dynamically configured according to system conditions.

[0116] The step S204 further comprises:

[0117] S2041: Obtain the task priority of the task data stream and the resource consumption of processing the task data stream at the edge node; wherein different task data streams have different priorities, and the priorities are pre-defined according to the sensor acquisition equipment and the acquired data; for example, in a certain machining production line, the primary qualification condition for a certain workpiece is that its size meets certain standards, and the secondary condition is that its light reflectivity meets certain standards, then the task data stream priority of the size data is greater than the task data stream priority of the light reflectivity data.

[0118] S2042: Calculate the total priority of the task data flow; the total priority is the weighted sum of the task priority and the resource consumption of the task data flow, total priority = α×task priority + (1-α)×resource consumption priority; wherein α is a weight value, which is a constant between 0 and 1, and the influencing factor of the resource consumption priority is the resource consumption. When a task data flow consumes a large amount of processor, memory or network resources, the resource consumption priority is low, otherwise the resource consumption priority is high.

[0119] S2043: The edge computing node checks the total priority of different received task data streams, and gives priority to the task data streams with high total priority; if the total priorities of multiple task data streams are the same or the difference is less than a certain threshold, the task data streams are processed on a first-come, first-served basis.

[0120] S2044: Return to step S2041.

[0121] The step S205 further comprises:

[0122] S2051: The edge management node obtains the function priorities of different functions used by the edge computing node for resetting; among them, different function priorities can be set according to the usage frequency of different functions, the amount of transmission data corresponding to different functions, the resource consumption load corresponding to different functions, etc., and the function priority can change dynamically; for example, if the usage frequency of a certain function increases significantly over a period of time, then its priority will also increase accordingly.

[0123] S2052: The edge management node sends a reset request to the edge computing node to be reset according to the highest ranking function priority. The reset request may also carry a configuration file required for the reset.

[0124] S2053: When the edge management node receives the reset completion confirmation message returned by the edge computing node, it proves that the edge node is reset successfully.

[0125] S2054: When the edge management node receives a reset rejection message returned by the edge computing node, the function of obtaining the next function priority ranking continues to send a reset request; the situations in which the edge computing node refuses to reset may include that the edge computing node is heavily loaded, is currently in a busy state, or the existing resources cannot meet the reset requirements, etc.

[0126] S2055: When the edge management node does not receive any message returned by the edge node within the preset time, it sends a reset request to the edge computing node again according to the function indicated by the current highest function priority; in this embodiment, a limited number of attempts are made to send the reset request again. If the edge computing node still does not respond, there may be abnormalities such as network failure or node offline, and the reconfiguration process can be skipped.

[0127] S2056: Return to step S2051.

[0128] By judging the total delay, when the total delay is small, the existing task processing method can be kept unchanged, and when the total delay is large, the task data stream with a higher priority can be selected for execution. When the total delay is larger, the edge node that processes the task data stream is reset, so that the edge processing layer can give priority to high-priority tasks. This is because the process of channel contention and the process of security authentication when data enters the node have caused significant delays to it; when the delay is further amplified, it means that the adaptability of the edge node and the task data stream is further reduced. Therefore, by resetting the edge node, it can reduce the contention delay and process the task data stream more specifically, which can effectively reduce the delay waiting time and ensure the overall task data processing efficiency of the system.

[0129] The embodiment of the present application also provides a computer program product; the computer program product is stored in a non-volatile computer-readable storage medium and includes machine-executable program instructions. When the machine-executable program instructions are executed in a device, the device executes the various method processes of task allocation in the above-mentioned embodiment, and is also used to provide support for building the system architecture of the above-mentioned embodiment, and can achieve the same technical effect. The non-volatile computer-readable storage medium includes a magnetic tape, a hard disk, an optical disk, a mobile hard disk, a U disk, a read-only memory ROM, a random access memory RAM, etc.

[0130] The specific implementation methods described above are preferred specific application examples of the present invention. The above examples are examples rather than exhaustive, and do not constitute a limitation on the principles of the embodiments themselves. Without departing from the scope and spirit of the disclosed embodiments, obvious simple changes and substitutions made by those skilled in the art without creative work are all within the scope of the disclosure of the above embodiments.

Claims

1. A task processing method, characterized in that: The edge management node measures the total delay of the task data flow on each edge computing node according to the sampling period; The total delay is the sum of the following delays: contention delay on the sensor device, transmission delay from the sensor to the edge computing node, processing delay of the edge computing node, processing delay of the edge management node, and transmission delay from the edge computing node to the actuator; Determine the relationship between the total delay and the third threshold and the fourth threshold, the third threshold being less than the fourth threshold: When the total delay is less than the third threshold, maintaining the task data flow of the existing edge node; When the total delay is greater than or equal to the third threshold and less than a fourth threshold, performing a task data stream selection operation; When the total delay is greater than or equal to the fourth threshold, performing a reset operation of the edge node; The above steps are repeated iteratively until the total delay returns to the normal range.

2. The method according to claim 1, characterized in that: The contention delay refers to the delay caused by channel contention when multiple sensor nodes send data to the same or a few edge computing nodes.

3. The method according to claim 1, characterized in that: The performing task data stream selection operation includes obtaining the task priority of the task data stream and the resource consumption of processing the task data stream at the edge node; Calculating the total priority of the task data flow, the total priority being a weighted sum of the task priority and the resource consumption of the task data flow; The edge computing node checks the total priority of different received task data streams, and preferentially processes the task data stream with a high total priority; If the total priorities of multiple task data streams are the same or the difference is less than a certain threshold, the task data streams are processed on a first-come, first-served basis.

4. The method according to claim 1, characterized in that: The reset operation of the edge node includes: the edge management node obtains the function priorities of different functions of the edge computing node for reset, and the edge management node sends a reset request to the edge computing node to be reset according to the highest ranking function priority: When the edge management node receives the reset completion confirmation message returned by the edge computing node, it proves that the edge node reset is successful; When the edge management node receives the reset rejection message returned by the edge computing node, the function that obtains the next function priority ranking continues to send a reset request; When the edge management node does not receive any message returned by the edge node within the preset time, it sends a reset request to the edge computing node again according to the function indicated by the current highest function priority.

5. The method according to claim 3, characterized in that: The resource consumption of the task data flow is specifically the resource consumption priority. The influencing factor of the resource consumption priority is the resource consumption amount. When a task data flow consumes a large amount of processor, memory or network resources, the resource consumption priority is low, otherwise the resource consumption priority is high.

6. The method according to claim 4, characterized in that: Different function priorities can be set according to the usage frequency of different functions, the amount of data transmitted corresponding to different functions, and the resource consumption load corresponding to different functions. The function priorities can change dynamically.

7. An Internet of Things system, characterized in that: The Internet of Things system is used to execute the task processing method in any one of claims 1 to 6.

8. A computer program product, characterized in that: The computer program product is stored in a non-volatile computer-readable storage medium and includes machine-executable program instructions; When the machine executable program instructions are executed in a device, the device is caused to execute the method according to any one of claims 1 to 6.

Images