Scheduling decision determination method of fog computing system and fog computing system

Abstract

The invention provides a scheduling decision determination method of a fog computing system and the fog computing system. The method comprises the following steps: step S102, acquiring resource demand information of an active task in each time slice in the running process of the fog computing system; step S104, inputting the resource demand information into a resource demand information prediction model, and predicting to obtain the resource demand information of the active task in the current time slice; step S106, acquiring a scheduling decision of a previous time slice, and inputting the resource demand information of the current time slice, the node resource information of the fog computing system and the optimal scheduling decision of the previous time slice into a pre-trained performance evaluation model to obtain system performance measurement information; and step S108, optimizing the optimal scheduling decision of the previous time slice based on the system performance measurement information to obtain the optimal scheduling decision of the current time slice. According to the method, the optimal scheduling decision can be obtained in a short time, and the computing performance of the fog computing system is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fog computing, and in particular to a scheduling decision determination method for a fog computing system and a fog computing system. Background Art

[0002] Fog computing is a distributed computing framework that extends cloud service resources to the edge of the network. Fog computing systems face the dual challenges of node heterogeneity and device diversity, and require appropriate task scheduling methods to ensure their good performance. Therefore, in order to improve the computing performance of fog computing systems during operation, how to design fog computing task scheduling decisions with good performance is an urgent problem to be solved. Summary of the invention

[0003] In view of this, the purpose of the present invention is to provide a scheduling decision determination method for a fog computing system and a fog computing system, which can obtain the optimal scheduling decision for the current time slice in a shorter time, thereby improving the computing performance of the fog computing system.

[0004] In order to achieve the above purpose, the technical solution adopted by the embodiment of the present invention is as follows:

[0005] In a first aspect, an embodiment of the present invention provides a method for determining a scheduling decision of a fog computing system, comprising:

[0006] Step S102, obtaining resource demand information of active tasks in each time slice during the operation of the fog computing system;

[0007] Step S104, inputting the resource demand information into a resource demand information prediction model to predict the resource demand information of the active task in the current time slice;

[0008] Step S106, obtaining the scheduling decision of the previous time slice, inputting the resource demand information of the current time slice, the node resource information of the fog computing system and the optimal scheduling decision of the previous time slice into a pre-trained performance evaluation model to obtain system performance measurement information; wherein the performance evaluation model is trained based on a sample set marked with system performance measurement information, and the sample set is constructed based on the resource demand information, the node resource information and the scheduling decision of each time slice;

[0009] Step S108, optimizing the optimal scheduling decision of the previous time slice based on the system performance measurement information to obtain the optimal scheduling decision of the current time slice;

[0010] The above steps S104 to S108 are repeatedly executed until the operation of the fog computing system ends.

[0011] Furthermore, an embodiment of the present invention provides a first possible implementation of the first aspect, wherein the step of inputting the resource demand information of the current time slice, the node resource information of the fog computing system, and the optimal scheduling decision of the previous time slice into a pre-trained performance evaluation model to obtain system performance measurement information includes:

[0012] A splicing matrix is ​​constructed based on the resource demand information of the current time slice, the node resource information of the fog computing system and the optimal scheduling decision of the previous time slice, and the splicing matrix is ​​input into a pre-trained performance evaluation model to obtain system performance measurement information within the implementation period of the optimal scheduling decision of the previous time slice.

[0013] Furthermore, the embodiment of the present invention provides a second possible implementation of the first aspect, which further includes:

[0014] Obtaining scheduling evaluation data of the fog computing system; wherein the scheduling evaluation data includes resource demand information, node resource information, scheduling decision and system performance measurement information of each time slice in which the scheduling decision is implemented;

[0015] A splicing matrix is ​​constructed based on the resource demand information, node resource information and scheduling decision of each time slice to form a sample set, and the sample set marked with the system performance measurement information is input into a neural network model for training to obtain the performance evaluation model.

[0016] Furthermore, the embodiment of the present invention provides a third possible implementation of the first aspect, wherein the step of obtaining the scheduling evaluation data of the fog computing system includes:

[0017] Obtaining historical scheduling evaluation data of the fog computing system;

[0018] or,

[0019] The computing resource requirement information of the task is extracted from the container tracking data set, and fog computing simulation is performed on the computing resource requirement information based on the random scheduler. The resource requirement information, node resource information, scheduling decision and system performance measurement information of each time slice in which the scheduling decision is implemented are recorded to obtain the scheduling evaluation data of the fog computing system.

[0020] Further, an embodiment of the present invention provides a fourth possible implementation manner of the first aspect, wherein the step of optimizing the optimal scheduling decision of the previous time slice based on the system performance metric information to obtain the optimal scheduling decision of the current time slice includes:

[0021] The system performance measurement information is used as a loss value, and multiple rounds of iterative optimization are performed on the optimal scheduling decision of the previous time slice based on the quasi-Newton algorithm and the loss value, until a new scheduling decision obtained by iteration meets the preset requirements;

[0022] The new scheduling decision that meets the preset requirements is used as the optimal scheduling decision for the current time slice.

[0023] Furthermore, an embodiment of the present invention provides a fifth possible implementation of the first aspect, wherein the preset requirement includes the number of iterations reaching a first preset number, or the number of times the new scheduling decision obtained by iteration is the same as the optimal scheduling decision of the previous time slice reaches a second preset number.

[0024] Furthermore, an embodiment of the present invention provides a sixth possible implementation of the first aspect, wherein the step of inputting the resource demand information into a resource demand information prediction model to predict the resource demand information of the active task in the current time slice includes:

[0025] Obtain historical resource demand information, input the historical resource demand information into the trained resource demand information prediction model, predict the resource demand information of the active task in the current time slice, and re-train the resource demand information prediction model based on the resource demand information of the current time slice.

[0026] Further, an embodiment of the present invention provides a seventh possible implementation of the first aspect, wherein the resource demand information prediction model is a hole convolutional neural network model;

[0027] The performance evaluation model is a fully connected neural network model.

[0028] Furthermore, an embodiment of the present invention provides an eighth possible implementation of the first aspect, wherein the system performance measurement information includes average response time and average energy consumption.

[0029] In a second aspect, an embodiment of the present invention further provides a fog computing system, including: a processor and a storage device;

[0030] The storage device stores a computer program, and when the computer program is executed by the processor, the method according to any one of the first aspects is executed.

[0031] The embodiment of the present invention provides a method for determining a scheduling decision of a fog computing system and a fog computing system, the method comprising: step S102, obtaining resource demand information of active tasks in each time slice during the operation of the fog computing system; step S104, inputting the resource demand information into a resource demand information prediction model, and predicting the resource demand information of the active tasks in the current time slice; step S106, obtaining the scheduling decision of the previous time slice, inputting the resource demand information of the current time slice, the node resource information of the fog computing system and the optimal scheduling decision of the previous time slice into a pre-trained performance evaluation model, and obtaining system performance measurement information; wherein the performance evaluation model is trained based on a sample set marked with system performance measurement information, and the sample set is constructed based on the resource demand information, the node resource information and the scheduling decision of each time slice; step S108, optimizing the optimal scheduling decision of the previous time slice based on the system performance measurement information, and obtaining the optimal scheduling decision of the current time slice; and repeating the above steps S104 to S108 until the operation of the fog computing system ends. The present invention predicts the computing resource demand information of the active task in the current time slice during the operation of the fog computing system, and inputs the predicted resource demand information, node resource information and the optimal scheduling decision of the previous time slice into the performance evaluation model to obtain system performance measurement information. The impact of the computing resource demand information and the scheduling decision on the fog computing system can be determined, and the evaluation of the scheduling decision is realized. By optimizing the optimal scheduling decision of the previous time slice based on the system performance measurement information to obtain the optimal scheduling decision of the current time slice, the optimal scheduling decision of the current time slice can be obtained in a shorter time, thereby improving the computing performance of the fog computing system.

[0032] Other features and advantages of the embodiments of the present invention will be described in the following description, or some features and advantages can be inferred or determined without doubt from the description, or can be learned by implementing the above-mentioned techniques of the embodiments of the present invention.

[0033] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0035] Figure 1 A flow chart of a method for determining a scheduling decision of a fog computing system provided by an embodiment of the present invention is shown;

[0036] Figure 2 A structural diagram of a fog computing system scheduling method based on a dilated convolutional neural network and a quasi-Newton algorithm provided by an embodiment of the present invention is shown;

[0037] Figure 3 A structural schematic diagram of a prediction model and a deep proxy model based on a dilated convolutional neural network provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. The embodiments of the present invention are described in detail below.

[0039] This embodiment provides a method for determining scheduling decisions for a fog computing system. The method can be applied to electronic devices such as computers. Figure 1 The flowchart of the method for determining the scheduling decision of the fog computing system shown in the figure mainly includes the following steps:

[0040] Step S102, obtaining resource demand information of active tasks in each time slice during the operation of the fog computing system;

[0041] For each task, when it starts to execute, it becomes an active task. After it becomes an active task, its basic computing resource requirements are recorded in each time slice.

[0042] The fog computing system to which the above method is applicable may include nodes with the same or different computing resources. Tasks with the same or different computing resource requirements are submitted to a certain node and can be migrated between nodes. The total computing amount of the task is invisible to the scheduler. The task has a basic computing resource requirement in each time slice, and the calculation can only start when the basic computing resource requirement is met; the task has an upper limit on available resources in each time slice, and the computing resources actually used by the task in each time slice cannot exceed the upper limit on available resources.

[0043] The basic resource requirements of a task in each time slice are visible to the scheduler only when the calculation of the time slice is in progress. After a task is submitted to the fog computing system, it will first wait in the waiting queue submitted to the node. At the beginning of each time slice, the fog computing system checks the tasks waiting at each node in the order of submission. If the node corresponding to the queue where the task is located can still meet the basic computing resource requirements of the task after meeting the basic computing resource requirements of the active tasks on it, and the total number of active tasks in the system does not reach the maximum number of active tasks allowed by the system, then the task becomes an active task.

[0044] Step S104, inputting the resource demand information into the resource demand information prediction model to predict the resource demand information of the active task in the current time slice;

[0045] Before the current time slice calculation starts and after the active task is updated, the resource demand information obtained before the current time slice is input into the resource demand information prediction model, and the resource demand information of the active task in the current time slice is predicted based on the resource demand information prediction model.

[0046] The resource demand information prediction model is constructed as needed. Each computing resource considered by each active task when making a scheduling decision needs to have a corresponding resource demand information prediction model.

[0047] Step S106, obtaining the scheduling decision of the previous time slice, inputting the resource demand information of the current time slice, the node resource information of the fog computing system and the optimal scheduling decision of the previous time slice into the pre-trained performance evaluation model to obtain system performance measurement information;

[0048] The performance evaluation model is trained based on a sample set marked with system performance measurement information, and the sample set is constructed based on resource demand information, node resource information and scheduling decisions of each time slice;

[0049] The above-mentioned performance evaluation model can be a deep neural network model. The deep neural network can learn complex features, which can give the performance evaluation model in this method the ability to learn the complex features of the impact of scheduling decisions on the fog computing system, so that the performance evaluation model can make a more accurate evaluation of the scheduling decision.

[0050] Step S108, optimizing the optimal scheduling decision of the previous time slice based on the system performance measurement information to obtain the optimal scheduling decision of the current time slice;

[0051] Repeat the above steps S104 to S108. At the beginning of each time slice, the optimal scheduling decision for the current time slice is calculated based on the above steps S104 to S108 until the fog computing system ends.

[0052] The scheduling decision determination method for the above-mentioned fog computing system provided in this embodiment predicts the computing resource demand information of the current time slice of the active task during the operation of the fog computing system, and inputs the predicted resource demand information, node resource information and the optimal scheduling decision of the previous time slice into the performance evaluation model to obtain system performance measurement information. It can determine the impact of the computing resource demand information and the scheduling decision on the fog computing system, realize the evaluation of the scheduling decision, and optimize the optimal scheduling decision of the previous time slice based on the system performance measurement information to obtain the optimal scheduling decision of the current time slice. The optimal scheduling decision of the current time slice can be obtained in a shorter time, thereby improving the computing performance of the fog computing system.

[0053] In one embodiment, this embodiment provides a specific implementation method of inputting resource demand information of the current time slice, node resource information of the fog computing system, and the optimal scheduling decision of the previous time slice into a pre-trained performance evaluation model to obtain system performance measurement information:

[0054] A splicing matrix is ​​constructed based on the resource demand information of the current time slice, the node resource information of the fog computing system and the optimal scheduling decision of the previous time slice. The splicing matrix is ​​input into the pre-trained performance evaluation model to obtain the system performance measurement information during the implementation period of the optimal scheduling decision of the previous time slice.

[0055] In a specific implementation, the resource demand information based on the current time slice and the node resource information of the fog computing system are normalized to form a matrix, the optimal scheduling decision information of the previous time slice is organized into a matrix (the optimal scheduling decision of the previous time slice uses unique hot encoding and does not need to be normalized), and the matrices are spliced ​​to form a spliced ​​matrix.

[0056] The computing resource ownership data of all nodes in the fog computing system are normalized and organized into a matrix. Each row of the matrix corresponds to a node and each column corresponds to a computing resource. The value of the "m"th row and "n"th column describes the number of the "n"th computing resource owned by the "m"th node. The computing resource demand data of the active task in this time slice obtained by the above prediction is normalized and organized into a matrix. Each row of the matrix corresponds to an active task and each column corresponds to a computing resource. The value of the "m"th row and "n"th column describes the demand of the "m"th active task for the "n"th computing resource. Quantity; organize the new scheduling decision information obtained from the last optimization of this time slice into a matrix. If the scheduling decision has not been optimized in this time slice, the current active task allocation information before the migration operation of this time slice is used as the initial scheduling decision and organized into a matrix. Each row of the matrix corresponds to an active task, and each column corresponds to a node. Each row of the matrix uses a unique hot encoding, where the value of the "m"th row and the "n"th column is 1, indicating that the "m"th active task is assigned to the "n"th node, and if it is 0, it means that the "m"th active task is not assigned to the "n"th node. Then the above three matrices are spliced ​​and input into the performance evaluation model. The output of the performance evaluation model is the system performance measurement information for the scheduling decision implemented in the time slice.

[0057] In one embodiment, the performance evaluation model is a fully connected neural network model.

[0058] The performance evaluation model is:

[0059] y=f a (x; W) (1)

[0060] Among them, f a is the performance evaluation model, x is the input, y is the output, and W is the weight. The input x consists of three dimensions: |H|×|R|, |A| max ×|R|、|H|×|A| max The matrices are spliced ​​together, which respectively describe the node resource information, the basic computing resource demand information of active tasks, and the scheduling decision information. Among them, |H| represents the number of nodes in the system, |R| represents the number of computing resource types considered when making scheduling decisions, and |A| max Indicates the maximum number of active tasks supported by the system. After the input x enters the proxy model, it will be flattened into a one-dimensional vector for subsequent processing. The output y is the time slice performance metric of the system. The performance evaluation model uses a fully connected neural network with 3 hidden layers. The first and second hidden layers use the softplus activation function, the third hidden layer uses the tanhshrink activation function, and the output layer uses the sigmoid activation function. The formal expression is as follows:

[0061] x1=softplus(x;W1) (2)

[0062] x2=softplus(x1;W2) (3)

[0063] x3=tanhshrink(x2;W3) (4)

[0064] y=sigmoid(x3;W4) (5)

[0065] Among them, x represents the input of the entire model, y represents the output of the entire model, and x i represents the output of the i-th hidden layer in the performance evaluation model, W i represents the weight of the i-th hidden layer.

[0066] The specific construction structure of the above fully connected neural network can be:

[0067] Including dimensions of |H|×|R|+|A| max ×|R|+|H|×|A| max The input layer of, where |H| represents the number of nodes in the system, |R| represents the number of types of computing resources considered, and |A| max Indicates the maximum number of active tasks supported by the system;

[0068] A fully connected layer with 128 neurons and softplus activation function;

[0069] A fully connected layer with 128 neurons and softplus activation function;

[0070] A fully connected layer with 64 neurons and tanhshrink activation function;

[0071] Contains 2 neurons and a fully connected output layer with sigmoid activation function.

[0072] In one embodiment, the method provided in this embodiment further includes:

[0073] Obtaining scheduling evaluation data of the fog computing system; wherein the scheduling evaluation data includes resource demand information of each time slice, node resource information, scheduling decision, and system performance measurement information of the time slice in which each scheduling decision is implemented;

[0074] Based on the resource demand information of each time slice, node resource information and scheduling decision, a splicing matrix is ​​constructed to form a sample set. The sample set marked with system performance measurement information is input into the neural network model for training to obtain a performance evaluation model.

[0075] For the data corresponding to a time slice in the scheduling evaluation data of the fog computing system, the node resource information data describing the node computing resource ownership information in the data corresponding to the time slice is normalized and organized into a matrix, where each row of the matrix corresponds to a node and each column corresponds to a computing resource, wherein the value of the "m"th row and the "n"th column describes the number of the "n"th computing resource owned by the "m"th node; the resource requirement information data describing the computing resource requirement information of the active task in the data corresponding to the time slice is normalized and organized into a matrix, where each row of the matrix corresponds to an active task and each column corresponds to a The data describing the scheduling decision information in the data corresponding to the time slice is organized into a matrix, each row of the matrix corresponds to an active task, each column corresponds to a node, and each row of the matrix is ​​encoded using one-hot encoding, wherein if the value of the mth row and the nth column is 1, it means that the mth active task is assigned to the nth node after the scheduling decision is executed, and if it is 0, it means that the mth active task is not assigned to the nth node after the scheduling decision is executed. The above three matrices are concatenated as the input of the performance evaluation model, and the values ​​describing the system time slice performance measurement information in the data corresponding to the time slice are normalized, and the values ​​of the normalized performance measurement information are used as the output value of the neural network model, and the neural network model is trained. The trained neural network model is recorded as the performance evaluation model, so that the performance evaluation model can capture the complex characteristics of the impact of scheduling decisions on the fog computing system and realize the evaluation of scheduling decisions.

[0076] In one embodiment, this embodiment provides a specific implementation method for obtaining scheduling evaluation data of a fog computing system:

[0077] Obtain historical scheduling evaluation data of the fog computing system;

[0078] or,

[0079] The computing resource requirement information of the task is extracted from the container tracking dataset, and fog computing simulation is performed on the computing resource requirement information based on the random scheduler. The resource requirement information, node resource information, scheduling decision and system performance measurement information of each time slice are recorded, and the scheduling evaluation data of the fog computing system is obtained.

[0080] The scheduling evaluation data of the fog computing system includes multiple pieces of information, each of which corresponds to a time slice. The scheduling evaluation data includes the resource requirement information of the fog computing system in each time slice (that is, the basic resource requirement information of the active task in the time slice), node resource information (including node computing resource ownership information), scheduling decision (active task allocation information after the scheduling decision made in the time slice is executed) and system performance measurement information of the time slice in which the scheduling decision is implemented.

[0081] If the fog computing system does not have the above-mentioned scheduling evaluation data, the public container tracking data set can be used to extract the computing resource requirement information of the task, and the random scheduler can be used to perform fog computing simulation. During the simulation process, the node computing resource ownership information of each time slice, the basic computing resource requirement information of the active task in the time slice, the scheduling decision information made in the time slice, and the system time slice performance measurement information of the scheduling decision implementation are recorded, so as to obtain the scheduling evaluation data about the fog computing system for training the performance evaluation model (also called the proxy model).

[0082] In one embodiment, this embodiment provides a specific implementation method for optimizing the optimal scheduling decision of the previous time slice based on system performance measurement information to obtain the optimal scheduling decision of the current time slice:

[0083] The system performance measurement information is used as the loss value, and multiple rounds of iterative optimization are performed on the optimal scheduling decision of the previous time slice based on the quasi-Newton algorithm and the loss value until the new scheduling decision obtained by the iteration meets the preset requirements; the new scheduling decision that meets the preset requirements is used as the optimal scheduling decision for the current time slice.

[0084] The system performance measurement information output by the performance evaluation model is used as the loss value, and the performance evaluation model is used as the loss function. The L-BFGS quasi-Newton algorithm is used to optimize the input of the performance evaluation model. The optimized matrix describing the active task allocation information (i.e., scheduling decision information) is taken out. During iterative optimization, this part of the matrix is ​​spliced ​​with the task resource demand information and the node resource information into a new matrix and input into the performance evaluation model again to obtain the loss value again, and then the scheduling decision information of the performance evaluation model is input again. This iteration is repeated until the new scheduling decision obtained by iteration meets the preset requirements. For each row of the above matrix, the largest value is set to 1, and the rest of the values ​​are set to 0 to obtain a matrix containing the new scheduling decision information, where the value of the "m"th row and the "n"th column is 1, indicating that the "m"th active task is assigned to the "n"th node, and if it is 0, it means that the "m"th active task is not assigned to the "n"th node. The L-BFGS quasi-Newton algorithm can quickly find a relatively good scheduling decision.

[0085] Determine whether the new scheduling decision obtained in each round of optimization meets the preset requirements. If so, the new scheduling decision is used as the optimal scheduling decision for the current time slice (that is, the final scheduling decision). If not, it is iterated again based on the quasi-Newton algorithm and the loss value until the new scheduling decision meets the preset requirements.

[0086] The above-mentioned preset requirements include that the number of iterations reaches a first preset number (the value range can be 180 to 220 times, preferably 200 times), or that the number of times the new scheduling decision obtained by iteration is the same as the optimal scheduling decision of the previous time slice reaches a second preset number (the value range can be 25 to 35 times, preferably 30 times).

[0087] If the new scheduling decision information matrix generated by the optimization is the same as the scheduling decision information matrix before the optimization for a specified number of consecutive times, the allocation described in the new scheduling decision matrix that is different from the current active task allocation will be used as the optimal scheduling decision for the current time slice.

[0088] In one embodiment, the resource demand information prediction model is a dilated convolutional neural network model.

[0089] The hole convolutional neural network model is:

[0090] y=f p (x; W) (6)

[0091] Among them, f p is a dilated convolutional neural network model, with x as input, y as output, and W as weight. The input x is a one-dimensional vector describing the historical demand value of a computing resource for the corresponding active task, and the output y is the predicted value describing the demand value of the computing resource for the active task in this time slice. The dilated convolutional neural network includes 2 dilated convolutional layers using tanh activation function, 1 pooling layer, and 1 linear layer, which are formally expressed as follows:

[0092] x1=tanh(conv1d(x;W1)) (7)

[0093] x2=tanh(conv1d(x1;W2)) (8)

[0094] x3=maxpool1d(x3) (9)

[0095] y=W linear x3 (10)

[0096] Among them, x represents the input of the entire model, y represents the output of the entire model, and x i represents the output of the i-th layer in the proxy model except the input layer, W i represents the weight of the i-th atrous convolutional layer, W linear Represents the weight of the linear layer, conv1d represents the dilated convolution operation on a one-dimensional vector, and maxpool1d represents the pooling operation on a one-dimensional vector.

[0097] The specific construction structure of the above-mentioned hole convolutional neural network model can be:

[0098] The input layer is a one-dimensional sequence with a size of the specified historical data using a window size;

[0099] The input channel is 1, the output channel is 8, the convolution kernel size is 2, the dilation rate is 2, and the one-dimensional convolution layer with tanh activation function is used;

[0100] The input channels are 8, the output channels are 8, the convolution kernel size is 2, the dilation rate is 4, and the one-dimensional convolution layer with tanh activation function is used;

[0101] Pooling layer with pooling kernel size of 2;

[0102] A fully connected linear output layer with 1 neuron and no activation function.

[0103] The advantage of applying the atrous convolutional neural network to time series prediction is that it can make relatively excellent prediction results, requires fewer parameters, has fast inference speed, and helps to achieve lightweight prediction models. For scheduling methods, they are usually sensitive to the speed of making scheduling decisions, so the application of the atrous convolutional neural network in this method can effectively play its advantages.

[0104] In one embodiment, this embodiment provides a specific implementation method of inputting resource demand information into a resource demand information prediction model to predict resource demand information of active tasks in the current time slice:

[0105] When the resource demand information prediction model completes the set number of trainings, the historical resource demand information is obtained, the historical resource demand information is input into the trained resource demand information prediction model, the resource demand information of the active task in the current time slice is predicted, and the resource demand information prediction model is retrained based on the resource demand information of the current time slice.

[0106] When the resource demand information prediction model has not completed the set number of trainings, or the recorded historical data cannot complete the training of the atrous convolutional neural network model, that is, the atrous convolutional neural network model cannot make predictions, the value of the basic computing resource demand of the last time slice of the active task is used as the resource demand information of the current time slice. If the atrous convolutional neural network model can make predictions, the value of the basic computing resource demand of the last time slice of the active task is used as the resource demand information of the current time slice, and the basic computing resource demand information of the historical time slices and the resource demand information of the current time slice are input into the resource demand information prediction model for a model training.

[0107] For each active task, one or more atrous convolutional neural network models are maintained, and their number is equal to the types of computing resources to be considered in making scheduling decisions. Check the recorded historical basic computing resource demand information. If the recorded historical data does not meet the prediction of the atrous convolutional neural network model, the value of the basic computing resource demand of the active task in the previous time slice is used as the predicted value of the basic computing resource demand of the active task in this time slice.

[0108] If the recorded historical data meets the prediction requirements of the atrous convolutional neural network model, but the atrous convolutional neural network model has not completed the specified number of training times, the value of the basic computing resource requirement information of the active task in the previous time slice will still be used as the predicted value of the basic computing resource requirement of the active task in the current time slice. However, at the end of the time slice, the prediction model based on the atrous convolutional neural network will be trained once after normalization using the recorded historical data and the newly obtained true value of the basic computing resource requirement of the active task.

[0109] If the recorded historical data meets the requirements of prediction using the atrous convolutional neural network model, and the atrous convolutional neural network model has completed the specified number of training times, the historical data will be normalized using the specified number of historical data in the window size and then input into the atrous convolutional neural network model to predict the basic computing resource requirement information of the active task in this time slice. The output value of the model is inversely normalized to be the predicted value. After the actual value of the basic computing resource requirement of the active task in this time slice is obtained, the atrous convolutional neural network model is trained again.

[0110] In one embodiment, the system performance measurement information provided by this embodiment includes the average response time ART of the system time slice t and average energy consumption AEC t .

[0111]

[0112] Among them, L t represents the set of tasks that leave the system at the tth time slice, I t represents the tth time slice, s(I t ) represents time slice I t RT(a) is the time from when the task becomes active to when the task is completed. represents the power function of the ith node, represents the maximum power that the i-th node can achieve.

[0113] The scheduling decision determination method for the fog computing system provided in this embodiment adopts a prediction model based on a void convolutional neural network, which gives the method the ability to predict the computing resource requirements of active tasks; adopts an agent model based on a deep neural network, which can capture the complex characteristics of the impact of scheduling decisions on the fog computing system, giving the method a powerful ability to evaluate scheduling decisions; adopts a quasi-Newton optimization algorithm, which gives the method the ability to quickly find excellent scheduling decisions; can make excellent scheduling decisions within an acceptable time, help improve the performance of the fog computing system, and has broad application prospects in the field of fog computing task scheduling.

[0114] Based on the above embodiment, this embodiment provides an example of a scheduling decision determination method using the above fog computing system, which can be specifically performed with reference to the following steps:

[0115] Step 1: Before the fog computing system is run, the historical scheduling evaluation data of the fog computing system is obtained, and the deep proxy model (i.e., the performance evaluation model) is trained based on the historical scheduling evaluation data to obtain a trained deep proxy model;

[0116] Step 2: Record the basic resource requirement information of active tasks in the fog computing system in each time slice;

[0117] Step 3: Input the basic resource requirement information of the active task into the prediction model based on the dilated convolutional neural network to predict the basic resource requirement information of the active task in the current time slice;

[0118] See Figure 2 The structural diagram of the fog computing system scheduling method based on the dilated convolutional neural network and the quasi-Newton algorithm is shown. At the beginning of the current time slice (i.e., the t-th time slice), the historical information of the computing resource requirements of the active tasks (i.e., the resource requirement information of the t-1th time slice or before the time slice) is input into the prediction model based on the dilated convolutional neural network. Each active task has a corresponding computing resource considered in a prediction model, and the computing resource requirement information of the active task in the t-th time slice can be predicted.

[0119] See Figure 3 The structural schematic diagram of the prediction model based on the atrous convolutional neural network and the deep proxy model shown in the figure, the prediction model based on the atrous convolutional neural network includes two atrous convolutional layers, a pooling layer and a linear layer connected in sequence using the tach activation function. After the historical information of the computing resource requirements of the active tasks is input into the prediction model based on the atrous convolutional neural network, the computing resource requirement information of the active tasks in the tth time slice is predicted.

[0120] Step 4: Integrate the basic resource demand information of the current time slice, the node resource information of the fog computing system, and the latest scheduling decision obtained during operation into the trained deep proxy model;

[0121] like Figure 2 As shown in FIG, the computing resource requirement information of the active task in the tth time slice, the optimal scheduling strategy obtained last time, and the node computing resource information are input into the trained deep proxy model, and the output of the deep proxy model is the estimated system time slice performance evaluation.

[0122] like Figure 3 As shown in FIG. 1 , the deep proxy model (i.e., the above-mentioned performance evaluation model) includes a fully connected neural network with three hidden layers, the first and second hidden layers use the softplus activation function, the third hidden layer uses the tanhshrink activation function, and the output layer uses the sigmoid activation function. The deep proxy model can output the system time slice performance metric.

[0123] Step 5: Use the output of the deep proxy model as the loss value, and use the quasi-Newton algorithm to optimize the scheduling decision through multiple wash cycles to obtain a new scheduling decision.

[0124] Step 6, determine whether the new scheduling decision meets the preset requirements. If so, the new scheduling decision is used as the final scheduling decision. If not, return to execute step 5.

[0125] Corresponding to the scheduling decision determination method of the fog computing system provided in the above embodiment, an embodiment of the present invention provides a fog computing system, which includes a processor and a memory, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the steps of the method provided in the above embodiment are implemented.

[0126] An embodiment of the present invention provides a computer-readable medium, wherein the computer-readable medium stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the method described in the above embodiment.

[0127] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system described above can refer to the corresponding process in the aforementioned embodiment, and will not be repeated here.

[0128] In addition, in the description of the embodiments of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0129] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0130] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0131] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for determining scheduling decisions in a fog computing system, characterized in that: include: Step S102, obtaining resource demand information of active tasks in each time slice during the operation of the fog computing system; Step S104, inputting the resource demand information into a resource demand information prediction model to predict the resource demand information of the active task in the current time slice; Step S106, obtaining the scheduling decision of the previous time slice, inputting the resource demand information of the current time slice, the node resource information of the fog computing system and the optimal scheduling decision of the previous time slice into a pre-trained performance evaluation model to obtain system performance measurement information; wherein the performance evaluation model is trained based on a sample set marked with system performance measurement information, and the sample set is constructed based on the resource demand information, the node resource information and the scheduling decision of each time slice; Step S108, optimizing the optimal scheduling decision of the previous time slice based on the system performance measurement information to obtain the optimal scheduling decision of the current time slice; The above steps S104 to S108 are repeatedly executed until the operation of the fog computing system ends.

2. The method according to claim 1, characterized in that The step of inputting the resource demand information of the current time slice, the node resource information of the fog computing system and the optimal scheduling decision of the previous time slice into the pre-trained performance evaluation model to obtain system performance measurement information includes: A splicing matrix is ​​constructed based on the resource demand information of the current time slice, the node resource information of the fog computing system and the optimal scheduling decision of the previous time slice, and the splicing matrix is ​​input into a pre-trained performance evaluation model to obtain system performance measurement information within the implementation period of the optimal scheduling decision of the previous time slice.

3. The method according to claim 2, characterized in that Also includes: Obtaining scheduling evaluation data of the fog computing system; wherein the scheduling evaluation data includes resource demand information, node resource information, scheduling decision and system performance measurement information of each time slice in which the scheduling decision is implemented; A splicing matrix is ​​constructed based on the resource demand information, node resource information and scheduling decision of each time slice to form a sample set, and the sample set marked with the system performance measurement information is input into a neural network model for training to obtain the performance evaluation model.

4. The method according to claim 3, characterized in that The step of obtaining the scheduling evaluation data of the fog computing system includes: Obtaining historical scheduling evaluation data of the fog computing system; or, The computing resource requirement information of the task is extracted from the container tracking data set, and fog computing simulation is performed on the computing resource requirement information based on the random scheduler. The resource requirement information, node resource information, scheduling decision and system performance measurement information of each time slice in which the scheduling decision is implemented are recorded to obtain the scheduling evaluation data of the fog computing system.

5. The method according to claim 1, characterized in that The step of optimizing the optimal scheduling decision of the previous time slice based on the system performance measurement information to obtain the optimal scheduling decision of the current time slice includes: The system performance measurement information is used as a loss value, and multiple rounds of iterative optimization are performed on the optimal scheduling decision of the previous time slice based on the quasi-Newton algorithm and the loss value, until a new scheduling decision obtained by iteration meets the preset requirements; The new scheduling decision that meets the preset requirements is used as the optimal scheduling decision for the current time slice.

6. The method according to claim 5, characterized in that The preset requirement includes that the number of iterations reaches a first preset number, or that the number of times the new scheduling decision obtained by iteration is the same as the optimal scheduling decision of the previous time slice reaches a second preset number.

7. The method according to claim 2, characterized in that The step of inputting the resource demand information into a resource demand information prediction model to predict the resource demand information of the active task in the current time slice includes: Obtain historical resource demand information, input the historical resource demand information into the trained resource demand information prediction model, predict the resource demand information of the active task in the current time slice, and re-train the resource demand information prediction model based on the resource demand information of the current time slice.

8. The method according to claim 7, characterized in that The resource demand information prediction model is a dilated convolutional neural network model; The performance evaluation model is a fully connected neural network model.

9. The method according to any one of claims 1 to 8, characterized in that: The system performance metric information includes average response time and average energy consumption.

10. A fog computing system, characterized in that: include: processor and storage device; The storage device stores a computer program, which, when executed by the processor, executes the method according to any one of claims 1 to 9.

Images