Task unloading method based on multiple edge computing nodes

Abstract

The invention discloses a task unloading method based on multiple edge computing nodes. The method comprises the following steps: constructing an edge computing network consisting of a client layer, an edge layer and a central layer; a user initiates a task unloading request through the client layer and uploads the task unloading request to the central layer for auditing, and after the auditing is passed, the user uploads task data information of a corresponding task to the central layer, obtains an edge server set covering the user and screens the edge server set to obtain a corresponding optimal edge server; uploading the task to an optimal edge server, dividing the task into a plurality of sub-tasks which are not associated with one another, and constructing a preliminary unloading strategy; a mapping relation is established according to the preliminary unloading strategy, response time needed for executing the corresponding subtasks is obtained according to the mapping relation, whether the edge computing nodes can process the corresponding subtasks or not is judged according to the obtained response time, and if yes, corresponding task results are obtained; according to the method, the task can be effectively unloaded, so that the task processing efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of edge computing, and in particular to a task offloading method based on multiple edge computing nodes. Background Art

[0002] With the development of cloud computing and the Internet of Things, a large number of data processing tasks need to be performed at the edge of the network to reduce data transmission delays and bandwidth consumption; however, the processing power and storage resources of edge computing nodes are limited and cannot handle all tasks; therefore, how to effectively offload tasks to suitable edge computing nodes has become an important research issue. For this purpose, we provide a task offloading method based on multiple edge computing nodes. Summary of the invention

[0003] The object of the present invention is to provide a task offloading method based on multiple edge computing nodes.

[0004] The purpose of the present invention can be achieved by the following technical solution: A task offloading method based on multiple edge computing nodes, comprising the following steps:

[0005] Step 1: Build an edge computing network consisting of a client layer, an edge layer, and a central layer;

[0006] Step 2: The user initiates a task offloading request through the client layer and uploads it to the central layer for review. If the review is passed, the user uploads the task data information of the corresponding task to the central layer, and obtains the edge server set covering the user, screens it, and obtains the corresponding optimal edge server;

[0007] Step 3: Upload the task to the optimal edge server, divide it into several unrelated subtasks, and build a preliminary offloading strategy;

[0008] Step 4: Establish a mapping relationship based on the preliminary offloading strategy, and obtain the response time required to execute the corresponding subtask based on the mapping relationship. Determine whether the edge computing node can process the corresponding subtask based on the obtained response time. If so, obtain the corresponding task result.

[0009] Furthermore, the client layer is used for users to initiate task offloading requests and upload corresponding tasks and task data information;

[0010] The edge layer includes edge servers and edge server nodes; the edge servers are used to monitor and coordinate the edge computing nodes, and the edge computing nodes and edge servers can be used to execute the tasks uploaded by the client layer;

[0011] The central layer is used to store and share data information, solve tasks beyond the computing capacity of edge servers, record the IP addresses, parameter indicators, workload conditions, etc. of all edge servers within the edge layer, and store them in the edge database;

[0012] Further, the process of obtaining the set of edge servers covering the user includes:

[0013] The user initiates a task offloading request through the client layer and sends it to the central layer. The task offloading request includes location information and task type. The central layer reviews the corresponding task offloading request. If the review is passed, it obtains the corresponding set of edge servers covering the user according to the location information in the task request.

[0014] Further, the process of screening the edge servers in the set of edge servers to obtain the optimal edge server includes:

[0015] When receiving the request approval notice, upload the corresponding task data information to the central layer, and read the basic information of all edge servers in the set of edge servers from the edge database,

[0016] Set the capacity index and capability index of the corresponding task according to the task data information uploaded by the user, and screen all edge servers in the set of edge servers according to it to obtain the corresponding first list of edge servers;

[0017] The central layer obtains the workload conditions of all edge servers in the corresponding first list of edge services according to the obtained basic information, marks the working status of the edge servers with high load as the busy status; and removes them from the first list of edge servers to obtain the corresponding second list of edge servers. The edge servers with high load refer to the edge servers with the CPU usage rate higher than K, where K is determined according to actual requirements; at the same time, the central layer will periodically obtain the basic information of the edge servers, record and save it, and update the working status of the corresponding edge servers;

[0018] Obtain the edge computing overhead K required for the corresponding task in the second list of edge servers s , set the overhead threshold, set the overhead threshold K1, and compare the obtained edge computing overhead K s with the overhead threshold K1,

[0019] If K s < K1, it indicates that the corresponding edge server s can process the corresponding task, then retain the corresponding edge server s and the temporary communication channel;

[0020] If K sIf it is ≥ K1, it indicates that the corresponding edge server cannot process the corresponding task. Then, the corresponding edge server s is removed from the second edge server list, and the corresponding temporary communication channel is cancelled. At the same time, a failure notice is fed back to the client layer.

[0021] Summarize the remaining edge servers to obtain a list of available edge servers, and sort all the edge servers in the available edge server list according to the edge computing overhead and the basic information of the corresponding edge servers to obtain the optimal edge server. The edge server with the highest ranking is used as the optimal edge server.

[0022] Furthermore, when receiving the failure notice, the user will offload the corresponding task to the central layer through the client layer, and the central layer will execute the corresponding task to obtain the corresponding task result.

[0023] Furthermore, uploading the corresponding task to the optimal edge server, and the process that the optimal edge server divides it into several uncorrelated subtasks and constructs a preliminary offloading strategy includes:

[0024] Read the task data information, divide it, obtain several uncorrelated subtasks, obtain the sub-data information corresponding to the subtasks according to the task data information of the corresponding task, and at the same time obtain the performance data of all edge computing nodes connected to the optimal edge server; and formulate the corresponding initial offloading strategy according to the corresponding sub-data information and performance data.

[0025] Furthermore, establish a mapping relationship according to the preliminary offloading strategy, obtain the response time required to execute the corresponding subtask according to the mapping relationship, and judge whether the edge computing node can process the corresponding subtask according to the obtained response time. If it can, the process of obtaining the corresponding task result includes:

[0026] The optimal edge server establishes a corresponding mapping relationship between the corresponding subtask and the edge computing node according to the initial offloading strategy to obtain several offloading groups;

[0027] Number the obtained offloading groups, denoted as i, where i = 1, 2, ……, m;

[0028] Obtain the sub-data information of the corresponding subtask in the offloading group i and the performance data of the edge computing node, and obtain the calculation time required for the corresponding edge computing node to calculate the corresponding subtask according to it In the formula, SL i is the data volume of the subtask i in the offloading group; NL i is the computing power of the edge computing node in the offloading group i, and f1 is the CPU clock frequency of the corresponding edge computing node;

[0029] Obtain the communication time required for the corresponding edge computing node to calculate the corresponding subtask according to the sub-data information of the subtask and the performance data of the edge computing node;

[0030] Among them,

[0031] In the formula, B is the transmission bandwidth of the sub-communication channel, P is the transmission power of the edge server, D is the distance between the edge server and the edge computing node, and N0 is Gaussian white noise;

[0032] Calculate the response time T required to offload the corresponding subtask to the edge computing node according to the mapping relationship 响 , and its corresponding data formula is T 响 = T 计 + T 通 ;

[0033] Set the response threshold T1, and compare the response time T 响 of each obtained offloading group with T1;

[0034] If T 响 ≤ T1, the edge server offloads the corresponding subtask to the corresponding edge computing node according to the mapping relationship of the offloading group. The edge computing node executes the corresponding subtask, obtains the corresponding subtask result, and feeds it back to the edge server;

[0035] If T 响 > T1, cancel the mapping relationship of the corresponding node. Then, the edge server locally executes the corresponding subtask and obtains the corresponding subtask result;

[0036] When the edge server summarizes the obtained subtask results to obtain the corresponding task result, and returns it to the client layer through the temporary communication channel for the user to view.

[0037] The beneficial effects of the present invention are as follows: By using the task offloading method based on multiple edge computing nodes, we can achieve the following beneficial effects: By offloading tasks to the edge server, the delay caused by ultra-long-distance transmission or slow calculation of mobile devices can be reduced; The tasks uploaded can be executed by the edge layer, thereby reducing the computing burden of mobile devices and improving the overall computing efficiency; Improve network bandwidth utilization: By dividing tasks into several independent subtasks and constructing a preliminary offloading strategy, network bandwidth can be better utilized and transmission efficiency can be improved. Brief Description of the Drawings

[0038] Figure 1 is the schematic diagram of the present invention. Detailed Embodiment

[0039] Such asFigure 1 As shown, a task offloading method based on multiple edge computing nodes includes the following steps:

[0040] Step 1: Build an edge computing network consisting of a client layer, an edge layer, and a central layer;

[0041] Step 2: The user initiates a task offloading request through the client layer and uploads it to the central layer for review. If the review is passed, the user uploads the task data information of the corresponding task to the central layer, and obtains the edge server set covering the user, screens it, and obtains the corresponding optimal edge server;

[0042] Step 3: Upload the task to the optimal edge server, divide it into several unrelated subtasks, and build a preliminary offloading strategy;

[0043] Step 4: Establish a mapping relationship based on the preliminary offloading strategy, and obtain the response time required to execute the corresponding subtask based on the mapping relationship. Determine whether the edge computing node can process the corresponding subtask based on the obtained response time. If so, obtain the corresponding task result.

[0044] It should be further explained that, in the specific implementation process, the client layer is used for users to initiate task offloading requests and upload corresponding tasks and task data information after the request is approved; the client layer includes various mobile devices used by users, including smart phones, computers, tablets and other devices that provide services to users.

[0045] The edge layer includes edge servers and edge server nodes; the edge servers are used to monitor and coordinate the edge computing nodes, and the edge computing nodes and edge servers can be used to execute tasks uploaded by the client layer to reduce delays caused by ultra-long-distance transmission or slow computing of mobile devices;

[0046] The central layer is used to store and share data information and solve tasks that exceed the computing capacity of the edge servers. It also records the IP addresses, parameter indicators, and workload conditions of all edge servers in the edge layer and stores them in the edge database.

[0047] It should be further explained that, in a specific implementation process, the edge server is connected to the edge computing node via a sub-communication channel, and one edge server is connected to several edge computing nodes.

[0048] It should be further explained that, in the specific implementation process, the user initiates a task offloading request through the client layer and uploads the corresponding task and task data information. The process of obtaining the edge server set covering the user includes:

[0049] User y initiates a task offloading request through the client layer and sends it to the central layer. The task offloading request includes location information and task type, where the task type includes data tasks and computing tasks. Here, y is the user number and y is an integer greater than zero.

[0050] Read the task type in the task offloading request. If the task type is a data task, return a request failure notice to the client layer. If it is a computing task, obtain all edge servers covering this user based on the location information in the task offloading request. If the number of obtained edge servers is zero, return a request failure notice to the user through the corresponding client layer. If the number of obtained edge servers is not zero, return a request passed notice to the user, summarize the corresponding edge servers, and obtain the corresponding edge server set.

[0051] It should be further noted that in the specific implementation process, the edge servers covering the user refer to the edge servers whose distance between the client layer used by user y and the edge servers is less than the coverage range of the edge servers.

[0052] It should be further noted that in the specific implementation process, the process of screening the edge servers in the edge server set to obtain the optimal edge server includes:

[0053] When the customer service layer receives the request passed notice, upload the corresponding task data information to the central layer through the client layer. The task data information includes computing requirements and data size. Obtain the IP addresses of all edge servers in the edge server set and read the basic information of all edge servers in the edge server set from the edge database based on them. The basic information includes computing power, current load condition, remaining storage capacity, and network bandwidth and other information, where computing power, storage capacity, and bandwidth are all one of the parameter indicators.

[0054] Set the capacity index and ability index for the corresponding task according to the task data information uploaded by the user, and screen all edge servers in the edge server set based on them. Extract the edge servers in the edge server set whose computing power and remaining storage capacity meet the ability index and capacity index respectively, and obtain the corresponding first edge server list.

[0055] The central layer obtains the workload conditions of all edge servers in the corresponding first edge service list, marks the working status of the edge servers with high load as the busy status, and removes them from the first edge server list to obtain the corresponding second edge server list. It should be further noted that the edge servers with high load refer to the edge servers with a CPU usage rate higher than K, where K is determined according to actual requirements. At the same time, the central layer will periodically obtain the basic information of the edge servers, record and save it, and update the working status of the corresponding edge servers.

[0056] Number the edge servers in the second edge server list as s, where s = 1, 2,..., n, n > 0 and n is an integer.

[0057] Build a temporary communication channel between the corresponding edge server s and the client layer used by the user y to facilitate the real-time progress of data transmission and interaction processes, making the processing of corresponding tasks more efficient and timely.

[0058] Obtain the transmission bandwidth W of the corresponding temporary communication channel and calculate the wireless transmission rate X between the corresponding mobile device and the edge server based on it. (y，s) ;

[0059] Where

[0060] In the formula, P y is the transmission power of the mobile device used by the user y, and the transmission power is one of the attribute parameters of the mobile device.

[0061] h (y，s) is the channel gain between the mobile device of the user y and the edge server.

[0062] σ is the noise power.

[0063] Obtain the transmission energy consumption E of the corresponding task in the constructed temporary communication channel based on the obtained wireless transmission rate. (y，s) and the transmission delay T (y，s) ;

[0064] Where

[0065]

[0066] In the formula, Z(y, s) is the data volume of the corresponding task, that is, the data size.

[0067] Calculate the corresponding edge computing overhead K based on the obtained transmission energy consumption and transmission delay. s ;

[0068] K s= λ1 × E (y,s) + λ2 × T (y,s) ;

[0069] Where λ1 and λ2 are weight parameters determined according to actual requirements;

[0070] Set an overhead threshold K1, and compare the obtained edge computing overhead K s with the overhead threshold K1.

[0071] If K s < K1, it indicates that the corresponding edge server s can handle the corresponding task, and then retain the corresponding edge server s and the temporary communication channel;

[0072] If K s ≥ K1, it indicates that the corresponding edge server cannot handle the corresponding task, then remove the corresponding edge server s from the second edge server list, cancel the corresponding temporary communication channel, and at the same time feedback an execution failure notice to the client layer; when the user receives the execution failure notice, the user will unload the corresponding task to the central layer through the client layer, and the central layer will execute the corresponding task, obtain the corresponding task result, and return it to the client layer;

[0073] Summarize the retained edge servers to obtain a corresponding list of available edge servers, and perform a priority sorting on all edge servers in the available edge server list according to the basic information of all edge servers and the corresponding edge computing overhead in the obtained available edge server list, and use the edge server with the highest sorting as the optimal edge server;

[0074] It should be further noted that in the specific implementation process, uploading the task data information to the optimal edge server, and the process that the optimal edge server divides it into several uncorrelated subtasks and constructs a preliminary offloading strategy includes:

[0075] Read the task data information, divide it, obtain several uncorrelated subtasks, and obtain the corresponding sub-data information according to the task data information of the corresponding task. The sub-data information includes the amount of computation, the amount of data transmission, and the computing requirements;

[0076] Obtain the performance data of all edge computing nodes connected to it according to the obtained optimal edge server. The performance data includes computing power, load situation, cpu clock frequency, etc.;

[0077] The optimal edge server formulates a corresponding initial offloading strategy according to the corresponding sub-data information and performance data;

[0078] It should be further noted that in the specific implementation process, a mapping relationship is established according to the preliminary offloading strategy, and the response time required to execute the corresponding subtasks is obtained according to the mapping relationship. Judging whether the edge computing node can process the corresponding subtasks according to the obtained response time. If it can, the process of obtaining the corresponding task results includes:

[0079] The optimal edge server establishes a corresponding mapping relationship between the corresponding subtasks and the edge computing nodes according to the initialization offloading strategy, and obtains a number of offloading groups;

[0080] Number the obtained offloading groups, denoted as i, where i = 1, 2,..., m;

[0081] Obtain the sub-data information of the corresponding subtasks in the offloading group i and the performance data of the edge computing node, and obtain the computing time required for the corresponding edge computing node to compute the corresponding subtasks according to them In the formula, SL i is the data volume of subtask i in the offloading group; NL i is the computing power of the edge computing node in the offloading group i, and f1 is the CPU clock frequency of the corresponding edge computing node;

[0082] Obtain the communication time required for the corresponding edge computing node to compute the corresponding subtasks according to the sub-data information of the subtasks and the performance data of the edge computing node;

[0083] Among them,

[0084] In the formula, B is the transmission bandwidth of the sub-communication channel, P is the transmission power of the edge server, D is the distance between the edge server and the edge computing node, and N0 is the Gaussian white noise;

[0085] Calculate the response time T required to offload the corresponding subtasks to the edge computing node according to the mapping relationship 响 , and its corresponding data formula is T 响 = T 计 + T 通 ;

[0086] Set the response threshold T1, and compare the response time T 响 of each obtained offloading group with T1;

[0087] If T 响 ≤ T1, the edge server offloads the corresponding subtasks to the corresponding edge computing nodes according to the mapping relationship of the offloading group. The edge computing node executes the corresponding subtasks, obtains the corresponding subtask results, and feeds them back to the edge server;

[0088] If T 响If it is greater than T1, the mapping relationship of the corresponding node is cancelled. Furthermore, the edge server locally executes the corresponding subtask and obtains the result of the corresponding subtask;

[0089] When the edge server aggregates the obtained subtask results to obtain the corresponding task result, and returns it to the client layer through the temporary communication channel for the user to view.

[0090] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A task offloading method based on multiple edge computing nodes, characterized in that, It includes the following steps: Step 1: Construct an edge computing network composed of a client layer, an edge layer, and a central layer; Step 2: The user initiates a task offloading request through the client layer and uploads it to the central layer for review. If the review is passed, the user uploads the task data information of the corresponding task to the central layer, obtains the set of edge servers covering the user, filters it, and obtains the corresponding optimal edge server; Step 3: Upload the task to the optimal edge server, divide it into several uncorrelated subtasks, and construct a preliminary offloading strategy; Step 4: Establish a mapping relationship according to the preliminary offloading strategy, obtain the response time required to execute the corresponding subtask according to the mapping relationship, and judge whether the edge computing node can process the corresponding subtask according to the obtained response time. If it can, obtain the corresponding task result.

2. The task offloading method based on multiple edge computing nodes according to claim 1, wherein The client layer is used for the user to initiate a task offloading request and upload the corresponding task and task data information; The edge layer includes edge servers and edge server nodes; the edge servers are used to monitor and coordinate the management of edge computing nodes, and both the edge computing nodes and the edge servers can be used to execute the tasks uploaded by the client layer; The central layer is used to store and share data information, and is also used to execute the tasks uploaded by the client layer. An edge database is set in the central layer to record the IP addresses, parameter indicators, and workload conditions of all edge servers in the edge layer.

3. The task offloading method based on multiple edge computing nodes according to claim 2, wherein, The process of obtaining the set of edge servers covering the user includes: The user initiates a task offloading request through the client layer and sends it to the central layer. The task offloading request includes location information and task type. The central layer reviews the corresponding task offloading request. If the review is passed, the set of edge servers covering the user is obtained according to the location information in the task offloading request.

4. A task offloading method based on multiple edge computing nodes according to claim 3, characterized in that, The process of filtering the edge servers in the set of edge servers to obtain the optimal edge server includes: After receiving the request passed notification, upload the corresponding task data information to the central layer, and read the basic information of all edge servers in the set of edge servers from the edge database; Set the capacity index and ability index of the corresponding task according to the task data information uploaded by the user, and filter all edge servers in the set of edge servers according to it to obtain the corresponding first list of edge servers; Obtain the workload conditions of all edge servers in the corresponding first list of edge servers, mark the working status of the edge servers with high load as the busy status; and remove it from the first list of edge servers to obtain the corresponding second list of edge servers; Obtain the edge computing overhead K required to execute the corresponding task within the list of the second edge servers s , screen the edge servers again to obtain a list of available edge servers, sort all the edge servers in the list of available edge servers according to the edge computing overhead and the basic information of the corresponding edge servers, and select the edge server with the highest ranking as the optimal edge server.

5. The task offloading method based on multiple edge computing nodes according to claim 4, wherein The process of filtering the edge servers again includes: Set an overhead threshold K1 and compare the obtained edge computing overhead K s with the overhead threshold K1; If K s < K1, the corresponding edge server is retained; If K s ≥ K1, the corresponding edge server is removed from the second edge server list, and a failure notice is fed back to the client layer; Summarize the remaining edge servers to obtain the corresponding list of available edge servers.

6. The task offloading method based on multiple edge computing nodes according to claim 4, wherein The process of uploading the task data information to the optimal edge server, where the optimal edge server divides it into several uncorrelated subtasks and constructs a preliminary offloading strategy includes: Read the task data information, divide it, obtain several uncorrelated subtasks, obtain the corresponding sub-data information of the subtasks according to the task data information of the corresponding tasks, and at the same time obtain the performance data of all edge computing nodes connected to the optimal edge server; and formulate corresponding initial offloading strategies according to the corresponding sub-data information and performance data.

7. A task offloading method based on multiple edge computing nodes according to claim 6, characterized in that, Establish the corresponding mapping relationship between the subtasks and the edge computing nodes according to the initialization offloading strategy, and then obtain the response time T required for the edge computing nodes to execute the corresponding subtasks according to the mapping relationship. 响 , judge whether the edge computing nodes can process the corresponding subtasks according to the obtained response time. If so, obtain the corresponding task results.

8. A task offloading method based on multiple edge computing nodes according to claim 7, characterized in that, Judge whether the edge computing node can process the corresponding subtask according to the obtained response time. If it can, the process of obtaining the corresponding task result includes: Set a response threshold T1 and compare the obtained response time T of each offloading group 响 with T1; If T 响 ≤ T1, the edge server unloads the corresponding subtask to the corresponding edge computing node according to the mapping relationship of the offloading group. The edge computing node executes the corresponding subtask, obtains the corresponding subtask result, and feeds it back to the edge server; If T 响 > T1, the mapping relationship of the corresponding node is cancelled, and then the edge server locally executes the corresponding subtask and obtains the corresponding subtask result; Summarize the obtained subtask results to obtain the corresponding task result, and return it to the client layer through the temporary communication channel for the user to view.

Images