Server operation and maintenance method and device, equipment and storage medium
By training an integrated failure rate prediction model and maintenance decision model, the server maintenance strategy is automatically determined, which solves the problem of time-consuming and labor-consuming manual formulation of maintenance plans in the existing technology, and improves the efficiency and accuracy of server maintenance.
Patent Information
- Application Number
- CN202410994501.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-12
AI Technical Summary
The existing server operation and maintenance solutions rely on manual maintenance strategies, which is time-consuming and labor-intensive, and cannot effectively prioritize the accident server that has the greatest impact on system reliability.
By training an integrated failure rate prediction model and maintenance decision model, the server's maintenance strategy is automated by using performance parameters and maintenance costs to automate the determination of the server's maintenance strategy, and realize the automation of failure rate prediction and maintenance decision.
Improves the efficiency and accuracy of server repairs, ensures priority to handling the servers that need to be repaired most under resource and time constraints, and reduces the need for manual intervention.
Smart Images

Figure CN118939461A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, and in particular to a server operation and maintenance method, device, equipment and storage medium. Background Art
[0002] Servers are the core of modern data centers and cloud computing environments. The normal operation of servers is crucial to maintaining service quality. If a server fails, it may cause data loss, damage, or even paralysis of the entire system, so regular maintenance of the server is required. When repairing a server, it usually involves identifying the server failure, repairing or replacing parts, and performing necessary tests to ensure that it returns to normal operation. Due to resource and time constraints, it is impossible to repair all failed servers at the same time, so it is necessary to give priority to those servers that have the greatest impact on system reliability.
[0003] In the current server operation and maintenance plan planning methods, most of them use manual methods, that is, according to the performance parameters of the server, manually determine whether the server needs maintenance, which server needs maintenance the most, etc. However, manual maintenance plan formulation requires a lot of manual input, and due to the variety of strategy options, manual maintenance plan formulation cannot cover all possible strategies. Summary of the invention
[0004] The embodiments of the present application provide a server operation and maintenance method, apparatus, device and storage medium, which can automatically formulate a more appropriate maintenance plan.
[0005] In a first aspect, an embodiment of the present application provides a server operation and maintenance method, including:
[0006] Acquire performance parameters of multiple servers to be detected, where the performance parameters are used to characterize the operating conditions of the servers to be detected;
[0007] Inputting the performance parameters of the multiple servers to be detected into a failure rate prediction model to obtain the failure rate of each server to be detected in the multiple servers to be detected;
[0008] Based on the failure rate of each server to be detected, the maintenance cost of each server to be detected, and the overall maintenance cost, determining a maintenance decision that meets the overall maintenance cost through a maintenance decision model;
[0009] Among them, the failure rate prediction model and the maintenance decision model are obtained by integrated training through a preset loss function; the loss function represents the loss value of the predicted failure rate of the failure rate prediction model under the objective function of the maintenance decision model.
[0010] In the embodiment of the present application, by inputting the performance parameters of the server into the integrated model, a maintenance decision for the server can be obtained, and under the condition of meeting the maintenance cost, the number of servers repaired can be maximized, which not only meets the maintenance cost but also can accurately determine which server needs to be repaired, thereby improving the efficiency of server maintenance.
[0011] Optionally, the loss function is determined based on a subgradient of a regret function; the subgradient represents a target difference between a maintenance decision corresponding to a predicted failure rate and a maintenance decision corresponding to a true failure rate under the objective function.
[0012] In an embodiment of the present application, the loss function is determined under the objective function based on the predicted failure rate and the actual failure rate. The failure rate can be combined with the objective function, that is, the failure probability of the server is combined with the maintenance cost, thereby realizing a strategy for repairing the server based on the maintenance cost.
[0013] Optionally, the loss function is determined based on a subgradient of a regret function, including:
[0014] The loss function includes a first target value, a second target value and a third target value;
[0015] The first target value represents the value of the first maintenance decision based on the failure rate difference and the failure rate difference under the objective function; the failure rate difference is determined based on the predicted failure rate and the actual failure rate; the first maintenance decision is the best maintenance decision that satisfies the first failure rate under the objective function;
[0016] The second target value represents the value of the second maintenance decision corresponding to the predicted failure rate and the actual failure rate under the target function; the second maintenance decision is the best maintenance decision that satisfies the actual failure rate under the target function;
[0017] The third target value represents the value of the third maintenance decision based on the actual failure rate and the actual failure rate under the target function.
[0018] In the embodiment of the present application, the loss function is constructed into three parts, which include the parameters of the failure rate prediction model and the parameters of the maintenance decision model. The loss function is jointly constructed by the parameters of the two models, so that the model is trained based on both at the same time, thereby realizing integrated model training.
[0019] Optionally, the loss function is:
[0020] L(y,y′)=-f(z * (αy′-y),αy′-y)+αf(z * (y),y′)-f(z *(y),y)
[0021] Among them, y′ is the predicted failure rate, y is the actual failure rate, α is the interpolation coefficient, and z * is the value of z that maximizes the optimization objective f among all possible values of z.
[0022] Optionally, the objective function is:
[0023]
[0024] Where N is the number of servers, y j To predict the failure rate, z j For decision making.
[0025] The optimal maintenance decision meets the following conditions:
[0026] z * (y) = argmaxf(z,y)
[0027]
[0028] Among them, z * (y) is the optimal value obtained after optimizing the above objective, that is, the value of z that satisfies the constraints and maximizes the optimization objective f among all possible values of z.
[0029] Optionally, the failure rate prediction model is a multi-layer neural network model; wherein the output of the i-th layer is the input of the i+1-th layer;
[0030] a i+1 =σ(W i a i +b i )
[0031] Among them, W i is the weight of the i-th layer, b i is the bias term of the i-th layer, σ is the activation function, a i is the output of the i-th layer, a i+1 is the output of the i+1th layer.
[0032] Optionally, the performance parameters include: usage status of the server CPU, usage status of the server GPU, usage status of the server radiator device, usage status of the server memory, and usage status of the server hard disk.
[0033] In a second aspect, an embodiment of the present application provides a server operation and maintenance device, including:
[0034] An acquisition module, used for acquiring performance parameters of a plurality of servers to be detected, wherein the performance parameters are used for characterizing the operation status of the servers to be detected;
[0035] An input module, used to input the performance parameters of the multiple servers to be detected into the failure rate prediction model to obtain the failure rate of each server to be detected among the multiple servers to be detected;
[0036] A determination module, configured to determine a maintenance decision that meets the overall maintenance cost through a maintenance decision model based on the failure rate of each server to be detected, the maintenance cost of each server to be detected, and the overall maintenance cost;
[0037] Among them, the failure rate prediction model and the maintenance decision model are obtained by integrated training through a preset loss function; the loss function represents the loss value of the predicted failure rate of the failure rate prediction model under the objective function of the maintenance decision model.
[0038] Optionally, the loss function is determined based on a subgradient of a regret function; the subgradient represents a target difference between a maintenance decision corresponding to a predicted failure rate and a maintenance decision corresponding to a true failure rate under the objective function.
[0039] Optionally, the determining module is specifically used to:
[0040] The loss function includes a first target value, a second target value and a third target value;
[0041] The first target value represents the value of the first maintenance decision based on the failure rate difference and the failure rate difference under the objective function; the failure rate difference is determined based on the predicted failure rate and the actual failure rate; the first maintenance decision is the best maintenance decision that satisfies the first failure rate under the objective function;
[0042] The second target value represents the value of the second maintenance decision corresponding to the predicted failure rate and the actual failure rate under the target function; the second maintenance decision is the best maintenance decision that satisfies the actual failure rate under the target function;
[0043] The third target value represents the value of the third maintenance decision based on the actual failure rate and the actual failure rate under the target function.
[0044] Optionally, the loss function is:
[0045] L(y,y′)=-f(z * (αy′-y),αy′-y)+αf(z * (y),y′)-f(z * (y),y)
[0046] Among them, y′ is the predicted failure rate, y is the actual failure rate, α is the interpolation coefficient, and z* is the value of z that maximizes the optimization objective f among all possible values of z.
[0047] Optionally, the objective function is:
[0048]
[0049] Where N is the number of servers, y j To predict the failure rate, z j For decision making.
[0050] The optimal maintenance decision meets the following conditions:
[0051] z * (y) = argmaxf(z,y)
[0052]
[0053] Among them, z * (y) is the optimal value obtained after optimizing the above objective, that is, the value of z that maximizes the optimization objective f among all possible values of z.
[0054] Optionally, the failure rate prediction model is a multi-layer neural network model; wherein the output of the i-th layer is the input of the i+1-th layer;
[0055] a i+1 =σ(W i a i +b i )
[0056] Among them, W i is the weight of the i-th layer, b i is the bias term of the i-th layer, σ is the activation function, a i is the output of the i-th layer, a i+1 is the output of the i+1th layer.
[0057] Optionally, the performance parameters include: usage status of the server CPU, usage status of the server GPU, usage status of the server radiator device, usage status of the server memory, and usage status of the server hard disk.
[0058] In a third aspect, an embodiment of the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any method described in claims 1 to 7 when executing the program.
[0059] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium storing a computer program executable by a computer device, wherein when the program is run on the computer device, the computer device executes the steps of any one of the methods of claims 1 to 7.
[0060] In a fifth aspect, an embodiment of the present application provides a computer program product, which includes a computer program stored on a computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer device, the computer device executes the steps of the above method. BRIEF DESCRIPTION OF THE DRAWINGS
[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0062] Figure 1 A system architecture diagram provided for an embodiment of the present application;
[0063] Figure 2 A flow chart of a server operation and maintenance method provided in an embodiment of the present application;
[0064] Figure 3 A schematic diagram of a process for determining a loss function provided in an embodiment of the present application;
[0065] Figure 4 A schematic diagram of the structure of a server operation and maintenance device provided in an embodiment of the present application;
[0066] Figure 5 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0067] In order to make the purpose, technical solution and beneficial effects of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0068] To facilitate understanding of this solution, the application scenario of this solution is introduced below.
[0069] Servers are crucial to maintaining service quality. When repairing servers, it is impossible to repair all servers at the same time due to resource and time constraints. Usually, the servers that need repair the most are selected manually based on the operating conditions of each server. The servers that need repair the most may be the ones with the lowest cost or the longest failure time, etc. However, manual selection cannot cover all possibilities and is time-consuming. The following is a detailed introduction to the steps of this application:
[0070] First of all, this application needs to train two models, one is the failure rate prediction model and the other is the maintenance decision model. Then, the performance parameters of the server are first input into the failure rate prediction model, and the failure rate of each server is output. Then, the failure rate, maintenance cost and overall maintenance cost of each server are input into the maintenance decision model, and the maintenance decision is output. According to the loss function jointly constructed by the failure rate prediction model and the maintenance decision model, the parameters of the failure rate prediction model are adjusted so that the failure rate prediction model meets the final requirements. After the two models are trained, when there is a server failure, the performance parameters of the failed server are input to obtain the specific server that needs to be repaired. There is no need to manually formulate a maintenance strategy.
[0071] See also Figure 1 , is a system architecture diagram provided in an embodiment of the present application, the system includes: a terminal device 101 and a server 102. The terminal device 101 can be a tablet computer, a laptop computer, a desktop computer, etc., but is not limited thereto. The server 102 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.
[0072] See also Figure 2 , is a flow chart of a server operation and maintenance method provided in an embodiment of the present application, the method is executed by an electronic device, and the electronic device may be a terminal device, such as Figure 2 As shown, the following steps are included:
[0073] Step 201: Acquire performance parameters of multiple servers to be detected, where the performance parameters are used to characterize the operating conditions of the servers to be detected.
[0074] Specifically, performance parameters are used to measure various indicators of the product. Each server has its own performance parameters, and the server's operating status can be seen from the server's performance parameters.
[0075] In some embodiments, the performance parameters include: usage status of the server CPU, usage status of the server GPU, usage status of the server radiator device, usage status of the server memory, and usage status of the server hard disk.
[0076] Specifically, the usage status of the server CPU includes: the power of the server CPU, the usage time of the server CPU, the CPU model of the server, the number of cores, and the number of threads; the usage status of the server GPU includes: the GPU model of the server and the GPU usage status; the usage status of the server radiator device includes: the radiator model of the server; the memory usage status of the server includes: the memory model, memory size, and memory cache size of the server; the hard disk usage status of the server includes: the type (mechanical or solid state) of the server's hard disk storage, size, usage time, and usage status.
[0077] Among them, the power of the server includes the power consumption of other components such as CPU, memory, hard disk, graphics card, etc. The service life of the server, generally speaking, the service life of the server is 3 to 5 years. The longer the server is used, the higher the probability of server failure. The server CPU model can reflect the quality of the server. The server kernel is the core part of the operating system. The kernel includes single kernel, dual kernel, micro kernel, etc. The server thread is the smallest unit that the operating system can perform calculation scheduling. The server's GPU is a high-performance graphics processing unit specially designed for use in server environments, providing powerful computing and graphics processing capabilities. The server's radiator can reduce the operating temperature of the server. The server's memory is where all programs in the computer run. The server's hard disk stores all software and user data.
[0078] Step 202: Input the performance parameters of the multiple servers to be detected into the failure rate prediction model to obtain the failure rate of each server to be detected in the multiple servers to be detected.
[0079] Specifically, after obtaining the performance parameters of the server to be tested, the performance parameters are input into the failure rate prediction model, and the failure rate of the server to be tested is output. The failure rate prediction model is used to detect the probability of server failure, and the failure probability of the server can be determined based on the performance parameters of the server. Assume that there are two servers A and B, where the performance parameters of A are server power 400W, number of cores 8 cores, and server usage time 3 months. The performance parameters of B are server power 300W, number of cores 4 cores, and server usage time 2 years. The performance parameters of servers A and B are respectively input into the failure rate prediction model, and the failure rates of servers A and B can be obtained. Assume that the failure rate of server A is 20% and the failure rate of server B is 80%.
[0080] Step 203: Based on the failure rate of each server to be detected, the maintenance cost of each server to be detected, and the overall maintenance cost, a maintenance decision that meets the overall maintenance cost is determined through a maintenance decision model.
[0081] Among them, the failure rate prediction model and the maintenance decision model are obtained by integrated training through a preset loss function; the loss function represents the loss value of the predicted failure rate of the failure rate prediction model under the objective function of the maintenance decision model.
[0082] Specifically, the maintenance decision model is used to determine whether the server needs maintenance, "1" indicates that maintenance is required, and "0" indicates that maintenance is not required. The failure rate, maintenance cost, and overall maintenance cost of the server to be tested are input into the maintenance decision model, and the output is whether the server needs maintenance.
[0083] Among them, when training the failure rate prediction model and the maintenance decision model, they are integrated through a preset loss function, that is, the loss function includes the failure rate prediction model and the maintenance decision model. The loss function is determined according to the predicted failure rate output by the failure rate prediction model and the maintenance decision output by the maintenance decision model.
[0084] In some embodiments, the loss function is determined based on the subgradient of the regret function; the subgradient represents the direction of the model update for more appropriate failure rate prediction applicable to the operation and maintenance plan. (The optimization objective represents the expectation of the predicted failure rate and the actual failure rate under the objective function).
[0085] Specifically, the model directly learns from the input data to the final output result, that is, end-to-end training has many difficulties, and determining the final decision solution is an optimization problem, and the variables involved are discrete. These two factors lead to the problem that the gradient cannot be transferred during the overall training process of the model, which in turn hinders the effective update of the parameters. Therefore, in order to solve the problem of end-to-end difficulties, the embodiment of the present application proposes to use a loss function to approximately optimize the regret value, thereby improving the quality of decision making.
[0086] The loss function is determined based on the subgradient of the regret function, which represents the expectation of the predicted failure rate and the actual failure rate under the objective function.
[0087] In some embodiments, the loss function is determined based on the subgradient of the regret function, such as Figure 3 As shown, the following steps are included:
[0088] Step 301: The loss function includes a first target value, a second target value, and a third target value.
[0089] Step 302, the first target value represents the value of the first maintenance decision based on the failure rate difference and the failure rate difference under the objective function; the failure rate difference is determined based on the predicted failure rate and the actual failure rate; the first maintenance decision is the best maintenance decision that satisfies the first failure rate under the objective function.
[0090] Step 303, the second target value represents the value of the second maintenance decision corresponding to the predicted failure rate and the actual failure rate under the objective function; the second maintenance decision is the best maintenance decision that meets the actual failure rate under the objective function.
[0091] Step 304: The third target value represents the value of the third maintenance decision based on the actual failure rate and the actual failure rate under the target function.
[0092] Specifically, as shown in formula (1), the loss function includes three target values, namely, the first target value, the second target value, and the third target value. Among them, the first target value includes the failure rate difference, which represents the value of the first maintenance decision corresponding to the failure rate difference under the objective function. The failure rate difference is the difference between the preset predicted failure rate and the actual failure rate. The first maintenance decision is the best maintenance decision under the condition of satisfying the failure rate difference.
[0093] The second target value includes the predicted failure rate and the real failure rate, and represents the value of the second maintenance decision corresponding to the predicted failure rate and the real failure rate under the target function, wherein the second maintenance decision is the best maintenance decision under the real failure rate.
[0094] The third target value includes the real failure rate, which represents the real failure rate and the value of the third maintenance decision corresponding to the real failure rate under the target function.
[0095] In some embodiments, the loss function is:
[0096] L(y,y′)=-f(z * (αy′-y),αy′-y)+αf(z * (y),y′)
[0097] -f(z * (y),y)...Formula (1)
[0098] Among them, y′ is the predicted failure rate, y is the actual failure rate, α is the interpolation coefficient, and z * is the value of z that maximizes the optimization objective f among all possible values of z.
[0099] In the embodiment of the present application, the value of α is 2, but it is not limited thereto, and the value of α can be any value greater than 0. It can be seen that the loss function of the embodiment of the present application is shown in formula (2):
[0100] L(y,y′)=-f(z * (2y′-y),2y′-y)+2f(z * (y),y′)
[0101] -f(z * (y),y)...Formula (2)
[0102] Among them, the failure rate difference is 2y′-y, and the first target value is f(z * (2y′-y),2y′-y), the second target value is 2f(z * (y),y′), the third target value is f(z * (y),y).
[0103] The design of the loss function is derived from the sub-gradient approximation of the regret function. The sub-gradient represents the direction of the model update for a more appropriate prediction of the failure rate applicable to the operation and maintenance plan, so as to approach the optimal decision-making goal. In uncertainty optimization, the regret function is shown in formula (3):
[0104] Regret(y,y′)=f(z′(y′),y)-f(z * ,y)...Formula (3)
[0105] Among them, z* represents the optimal decision under complete and complete failure rate information y. In real scenarios, it is usually impossible to obtain the actual failure rate completely. Therefore, based on the predicted failure rate y' obtained in step 202, it is substituted into step 203 to obtain the corresponding decision z'(y'). The optimization target difference between the current decision z' and the optimal decision z* is the value of the regret function.
[0106] The regret function cannot return continuous and informative gradients to update the model parameters, and therefore cannot directly optimize the above formula. Therefore, the formula (2) of this application is proposed as the loss function. In formula (2), the first target value and the third target value use negative values, because the optimization problem in the original business is a maximization problem, and the loss function often uses a minimization problem. Through the above gradient approximation, when the failure rate of different servers is unknown, the decision target value of the backend can be used as supervision to update the corresponding prediction model and solution model.
[0107] In some embodiments, the objective function is:
[0108]
[0109] Where N is the number of servers, y j To predict the failure rate, z jis the decision-making scheme, and j is the server. The optimal maintenance decision satisfies the following conditions, as shown in formula (5) and formula (6):
[0110] z * (y)=argmaxf(z,y)...Formula (5)
[0111]
[0112] Among them, z * (y) is the optimal value obtained after optimizing the above objectives, C is the overall maintenance cost, W j The maintenance cost for each server to be tested.
[0113] In some embodiments, the failure rate prediction model is a multi-layer neural network model, wherein the output of the i-th layer is the input of the i+1-th layer, as shown in formula (7):
[0114] a i+1 =σ(W i a i +b i )...Formula (7)
[0115] Among them, W i is the weight of the i-th layer, b i is the bias term of the i-th layer, σ is the activation function, σ(a)=max(0,a),a i is the output of the i-th layer, a i+1 is the output of the i+1th layer. 1 =x.
[0116] The present scheme is explained below by way of examples.
[0117] First, in the model training phase, for N servers numbered {1, 2, ..., N}, a certain amount of data set can be obtained containing the data of the above server set (or subset), including the relevant features of each server {x 1 ,x 2 ,…,x N} (including server model, server usage time, size of each server module, CPU, storage utilization, etc.), and corresponding failure rate data {y 1 ,y 2 ,…,y N}, the corresponding solution is {z 1 ,z 2 ,…,z N}. Using these sets of data, we can train with the loss function to get a new prediction-decision model. After the training is completed, we can save the parameters and reuse them later.
[0118] In the testing phase, the existing model parameters are loaded, and a prediction and a decision are made for a group of servers to obtain a group of maintenance plans. Specifically, the prediction uses the failure rate prediction model to obtain a group of failure rate predictions. The decision uses the maintenance decision solution to obtain the server maintenance plan.
[0119] Based on the same technical concept, the embodiment of the present application provides a server operation and maintenance device, such as Figure 4 As shown, the device 400 includes:
[0120] In a first aspect, an embodiment of the present application provides a server operation and maintenance device, including:
[0121] The acquisition module 401 is used to acquire performance parameters of multiple servers to be detected, where the performance parameters are used to characterize the operating conditions of the servers to be detected;
[0122] An input module 402, configured to input the performance parameters of the plurality of servers to be detected into a failure rate prediction model to obtain a failure rate of each of the plurality of servers to be detected;
[0123] A determination module 403 is used to determine a maintenance decision that meets the overall maintenance cost through a maintenance decision model based on the failure rate of each server to be detected, the maintenance cost of each server to be detected, and the overall maintenance cost;
[0124] Among them, the failure rate prediction model and the maintenance decision model are obtained by integrated training through a preset loss function; the loss function represents the loss value of the predicted failure rate of the failure rate prediction model under the objective function of the maintenance decision model.
[0125] Optionally, the loss function is determined based on a subgradient of a regret function; the subgradient represents a target difference between a maintenance decision corresponding to a predicted failure rate and a maintenance decision corresponding to a true failure rate under the objective function.
[0126] Optionally, the determining module 403 is specifically configured to:
[0127] The loss function includes a first target value, a second target value and a third target value;
[0128] The first target value represents the value of the first maintenance decision based on the failure rate difference and the failure rate difference under the objective function; the failure rate difference is determined based on the predicted failure rate and the actual failure rate; the first maintenance decision is the best maintenance decision that satisfies the first failure rate under the objective function;
[0129] The second target value represents the value of the second maintenance decision corresponding to the predicted failure rate and the actual failure rate under the target function; the second maintenance decision is the best maintenance decision that satisfies the actual failure rate under the target function;
[0130] The third target value represents the value of the third maintenance decision based on the actual failure rate and the actual failure rate under the target function.
[0131] Optionally, the loss function is:
[0132] L(y,y′)=-f(z * (αy′-y),αy′-y)+αf(z * (y),y′)-f(z * (y),y)
[0133] Among them, y′ is the predicted failure rate, y is the actual failure rate, α is the interpolation coefficient, and z * is the value of z that maximizes the optimization objective f among all possible values of z.
[0134] Optionally, the objective function is:
[0135]
[0136] Where N is the number of servers, y j To predict the failure rate, z j For decision making.
[0137] The optimal maintenance decision meets the following conditions:
[0138] z * (y) = argmaxf(z,y)
[0139]
[0140] Among them, z * (y) is the optimal value obtained after optimizing the above objective, that is, the value of z that maximizes the optimization objective f among all possible values of z.
[0141] Optionally, the failure rate prediction model is a multi-layer neural network model; wherein the output of the i-th layer is the input of the i+1-th layer;
[0142] a i+1 =σ(W i a i +b i )
[0143] Among them, W i is the weight of the i-th layer, b iis the bias term of the i-th layer, σ is the activation function, a i is the output of the i-th layer, a i+1 is the output of the i+1th layer.
[0144] Optionally, the performance parameters include: usage status of the server CPU, usage status of the server GPU, usage status of the server radiator device, usage status of the server memory, and usage status of the server hard disk.
[0145] Based on the same technical concept, the embodiment of the present application provides a computer device, which may be a terminal or a server, such as Figure 5 As shown, it includes at least one processor 501 and a memory 502 connected to the at least one processor. The specific connection medium between the processor 501 and the memory 502 is not limited in the embodiment of the present application. Figure 5 For example, the processor 501 and the memory 502 are connected via a bus. The bus can be divided into an address bus, a data bus, a control bus, and the like.
[0146] In the embodiment of the present application, the memory 502 stores instructions that can be executed by at least one processor 501, and the at least one processor 501 can execute the steps included in the above-mentioned server operation and maintenance method by executing the instructions stored in the memory 502.
[0147] The processor 501 is the control center of the computer device, and can use various interfaces and lines to connect various parts of the computer device, by running or executing instructions stored in the memory 502 and calling data stored in the memory 502. Optionally, the processor 501 may include one or more processing units, and the processor 501 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 501. In some embodiments, the processor 501 and the memory 502 may be implemented on the same chip, and in some embodiments, they may also be implemented separately on independent chips.
[0148] The processor 501 may be a general-purpose processor, such as a central processing unit (CPU), a digital signal processor, an application-specific integrated circuit (ASIC), a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component, and may implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present application. A general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application may be directly embodied as being executed by a hardware processor, or may be executed by a combination of hardware and software modules in the processor.
[0149] The memory 502 is a non-volatile computer-readable storage medium that can be used to store non-volatile software programs, non-volatile computer executable programs and modules. The memory 502 may include at least one type of storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory, a random access memory (Random Access Memory, RAM), a static random access memory (Static Random Access Memory, SRAM), a programmable read-only memory (Programmable Read Only Memory, PROM), a read-only memory (Read Only Memory, ROM), an electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), a magnetic memory, a disk, an optical disk, etc. The memory 502 is any other medium that can be used to carry or store a desired program code in the form of an instruction or data structure and can be accessed by a computer, but is not limited thereto. The memory 502 in the embodiment of the present application can also be a circuit or any other device that can realize a storage function, for storing program instructions and / or data.
[0150] Based on the same inventive concept, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program that can be executed by a computer device. When the program runs on the computer device, the computer device executes the steps of the above-mentioned server operation and maintenance method.
[0151] Based on the same inventive concept, an embodiment of the present application provides a computer program product, characterized in that the computer program product includes a computer program stored on a computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer device, the computer device executes the steps of the above-mentioned server operation and maintenance method.
[0152] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.
[0153] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0154] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0155] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0156] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.
Claims
1. A server operation and maintenance method, characterized in that: include: Acquire performance parameters of multiple servers to be detected, where the performance parameters are used to characterize the operating conditions of the servers to be detected; Inputting the performance parameters of the multiple servers to be detected into a failure rate prediction model to obtain the failure rate of each server to be detected in the multiple servers to be detected; Based on the failure rate of each server to be detected, the maintenance cost of each server to be detected, and the overall maintenance cost, determining a maintenance decision that meets the overall maintenance cost through a maintenance decision model; Among them, the failure rate prediction model and the maintenance decision model are obtained by integrated training through a preset loss function; the loss function represents the loss value of the predicted failure rate of the failure rate prediction model under the objective function of the maintenance decision model.
2. The method according to claim 1, characterized in that include: The loss function is determined based on the subgradient of the regret function; the subgradient represents the target difference between the maintenance decision corresponding to the predicted failure rate and the maintenance decision corresponding to the actual failure rate under the objective function.
3. The method according to claim 2, characterized in that The loss function is determined based on the subgradient of the regret function, including: The loss function includes a first target value, a second target value and a third target value; The first target value represents the value of the first maintenance decision based on the failure rate difference and the failure rate difference under the target function; the failure rate difference is determined based on the predicted failure rate and the actual failure rate; the first maintenance decision is the best maintenance decision that satisfies the failure rate difference under the target function; The second target value represents the value of the second maintenance decision corresponding to the predicted failure rate and the actual failure rate under the target function; the second maintenance decision is the best maintenance decision that satisfies the actual failure rate under the target function; The third target value represents the value of the third maintenance decision based on the actual failure rate and the actual failure rate under the target function.
4. The method according to claim 3, characterized in that The loss function is: L(y,y′)=-f(z*(αy′-y),αy′-y)+αf(z*(y),y′)-f(z*(y),y) Where y′ is the predicted failure rate, y is the actual failure rate, α is the interpolation coefficient, and z* is the value of z that maximizes the optimization objective f among all possible values of z.
5. The method according to claim 4, characterized in that The objective function is: Where N is the number of servers, y j To predict the failure rate, z j For decision-making solutions; The optimal maintenance decision meets the following conditions: z*(y)=argmaxf(z,y) Among them, z*(y) is the optimal value obtained after optimizing and solving the above objectives, that is, the value of z that maximizes the optimization objective f among all possible values of z.
6. As claimed in claim 1 - 5. Any of the methods described above, characterized in that include: The failure rate prediction model is a multi-layer neural network model; wherein the output of the i-th layer is the input of the i+1-th layer; a i+1 =σ(W i a i +b i ) Among them, W i is the weight of the i-th layer, b i is the bias term of the i-th layer, σ is the activation function, a i is the output of the i-th layer, a i+1 is the output of the i+1th layer.
7. The method according to claim 6, characterized in that include: The performance parameters include: the usage status of the server CPU, the usage status of the server GPU, the usage status of the server radiator device, the usage status of the server memory, and the usage status of the server hard disk.
8. A server operation and maintenance device, characterized in that: include: An acquisition module, used for acquiring performance parameters of a plurality of servers to be detected, wherein the performance parameters are used for characterizing the operation status of the servers to be detected; An input module, used to input the performance parameters of the multiple servers to be detected into the failure rate prediction model to obtain the failure rate of each server to be detected among the multiple servers to be detected; A determination module, configured to determine a maintenance decision that meets the overall maintenance cost through a maintenance decision model based on the failure rate of each server to be detected, the maintenance cost of each server to be detected, and the overall maintenance cost; Among them, the failure rate prediction model and the maintenance decision model are obtained by integrated training through a preset loss function; the loss function represents the loss value of the predicted failure rate of the failure rate prediction model under the objective function of the maintenance decision model.
9. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any one of the methods of claims 1 to 7 when executing the program.
10. A computer-readable storage medium, characterized in that: It stores a computer program executable by a computer device. When the program is run on the computer device, the computer device executes the steps of any method described in claims 1 to 7.
Citation Information
Cited By
Server availability scheme combination determination method and electronic equipment
CN120872505A
