Joint optimization method and system for data center
Through the joint optimization method of data centers combining gray model and bird flock algorithm, the problem of low load prediction and resource allocation efficiency is solved, high-precision load prediction and dynamic resource optimization are achieved, energy consumption is reduced and response speed is improved.
Patent Information
- Application Number
- CN202411609282.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-12
- Publication Date
- 2025-03-04
AI Technical Summary
The existing data center optimization methods are difficult to predict load changes and low resource allocation efficiency with high accuracy, unable to effectively reduce system energy consumption, and the response speed is not enough to meet the dynamic adjustment needs.
A joint optimization method based on gray model (GM) and bird flock algorithm is adopted to establish a time series through load history data, load prediction is used to optimize and adjust parameters through bird flock algorithm to achieve dynamic optimization configuration of resources.
It improves load prediction accuracy, optimizes resource allocation efficiency, reduces data center energy consumption, improves system response speed, and achieves an efficient balance between performance and energy consumption.
Smart Images
Figure CN119557089A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data center management, in particular to the technical field of data center joint optimization and resource prediction and adjustment. Background Art
[0002] With the rapid development of information technology, data centers are core facilities for information storage, processing and transmission, and their operational management efficiency is crucial. Traditional data centers mostly use static or manual configuration methods for load management, which has obvious limitations in practical applications. On the one hand, the load of data centers usually has obvious time series characteristics. How to use historical data to predict future load changes is an important means to improve the management efficiency of data centers. On the other hand, most current optimization strategies rely on simple load balancing or regular capacity expansion strategies, which make it difficult to control energy consumption and resource waste while ensuring performance.
[0003] Among the existing optimization methods, prediction models based on time series and intelligent optimization algorithms have high application value in data center load prediction and resource optimization. However, these technologies still have the following problems in the actual application of data centers: most existing models rely only on simple time series analysis or traditional prediction algorithms, which are difficult to adapt to the complex load changes in data centers; traditional optimization algorithms often ignore the correlation between resources at all levels in the data center and lack overall optimization strategies, resulting in low resource allocation efficiency and inability to effectively reduce system energy consumption; when the load surges or resources are scarce, the existing methods are not responsive enough and are difficult to meet the actual needs of dynamic adjustment of data centers.
[0004] In order to solve the above problems, the present invention proposes a joint optimization method based on the grey model (GM) and the bird flock algorithm. A time series is established through historical load data, and the prediction ability of the GM model and the parameter optimization ability of the bird flock algorithm are combined to achieve high-precision prediction of data center load and dynamic optimization configuration of resources. Summary of the invention
[0005] The technical problem to be solved by the embodiments of the present invention is to provide a data center joint optimization method and system, which takes into account the time series characteristics of the data center load and the dynamic optimization requirements of resource allocation, can achieve high-precision load prediction and optimize resource allocation efficiency, thereby achieving an efficient balance between performance and energy consumption in the data center.
[0006] In order to solve the above technical problems, an embodiment of the present invention provides a data center joint optimization method and system, the method comprising the following steps:
[0007] S100 obtains the load conditions of the data center in the historical period to form a data center load data set, and obtains a time series through the data center load data set. The load data set includes but is not limited to processor load, memory load, number of virtual machines, disk load, and network traffic. The data center load data set is sorted, and a fixed time interval is set. The data of each fixed time interval in the historical period is sorted to form a time series. The time series is X={x1,x2,x3,...,xn}, where n represents the data of the nth time interval.
[0008] S200 establishes a GM model. The GM model includes a GM model difference equation and a GM model prediction equation. The time series obtains the GM model difference equation through the GM model. The GM model difference equation includes adjustment parameters. Establishing the GM model includes:
[0009] S201 marks the time series as the original sequence, recorded as:
[0010] x (0) =(x (0) (1),x (0) (2),…,x (0) (n));
[0011] S202 obtains a cumulative sequence based on the original sequence, and the formula is:
[0012]
[0013] This means that for each time instant i, we replace x in the original sequence (0) (1),x (0) (2),…,x (0) (i) Accumulate and obtain a new sequence x (1) , recorded as a cumulative series;
[0014] S203 obtains the grey derivative of a cumulative sequence, the formula is:
[0015] dx (1) (i) = x (1) (i) - x (1) (i-1)i=2,3,…,n
[0016] Among them, dx (1) (i) represents the grey derivative at the i-th moment, x (1) (i) and x (1) (i-1) represents the value of the cumulative sequence at the i-th and i-1-th moments respectively;
[0017] S204 constructs the mean generating sequence:
[0018]
[0019] z (1) (i) is the i-th element of the mean generating sequence, x (1) (i) and x (1) (i―1) is a cumulative sequence x (1) The values of the i-th and i-1-th in ;
[0020] S205 uses the mean value generating series and grey derivatives to establish the GM model difference equation:
[0021] dx (1) (i)+a·z (1) (i) = b
[0022] where dx (1) (i) is the grey derivative, z (1) (i) is the mean generating sequence, a and b are adjustment parameters.
[0023] S300 uses the bird flock algorithm to obtain the optimal value of the adjustment parameters, including:
[0024] S301 initializes the bird flock, each particle represents possible adjustment parameters a and b. During initialization, each dimension of the particle will randomly generate a certain value as the search starting point. Each particle has a position and a speed, and the speed is the rate at which the particle updates its position;
[0025] S302 builds on the speed formula:
[0026]
[0027] Construct a formula about position:
[0028]
[0029] in, is the velocity of the particle at time t+1, is the position of the particle at time t+1, w is the inertia weight, c1 and c2 are learning factors, r1 and r2 are between [0,
[0030] 1], is the optimal position of the particle in the individual, is the global optimal position.
[0031] Step S303 also includes:
[0032] S3031 sets the physical layer and service layer in the Bayesian network. The physical layer includes but is not limited to the energy consumption of the physical machine, the memory load rate of the physical machine, and the CPU resource load rate of the physical machine. The service layer includes but is not limited to the total node load, resource requirements, failure rate, and bandwidth limitation.
[0033] S3032 establishes a multi-layer Bayesian network, the service layer Bayesian network S = {X, Y, P}, where X represents a node set, Y represents an arc set, P represents a conditional probability set, and a physical layer Bayesian network B, which only includes a node set. A mapping relationship is established between the nodes in the physical layer Bayesian network B and the nodes in X in the service layer Bayesian network S, the mapping relationship set is T, and the multi-layer Bayesian network is represented by {S, B, T}
[0034] S3033 uses physical layer data prediction to obtain a first probability confidence M1, with a learning factor c1=3*M1, and uses service layer data prediction to obtain a second probability confidence M2, with a learning factor c2=3*M2.
[0035] S400 brings the optimal value into the GM model prediction equation in the GM model to obtain the prediction series, uses the learning factors c1 and c2 to output the optimal combination of ab through the speed and position formula in the bird flock algorithm, and predicts and evaluates through the GM model. The prediction formula is:
[0036]
[0037] Get the predicted series X through the prediction formula (1) .
[0038] S500 executes the data center optimization strategy based on the prediction sequence, where the specific mechanism is that the target value of RMSE value less than 30% is calculated based on the sequence X (1) Adjust the optimization execution strategy of the data center;
[0039] Optimization execution strategies include but are not limited to expanding physical machines, limiting network traffic, adjusting virtual machines, etc.
[0040] S600 adjusts the Bayesian model according to the evaluation result if the RMSE value is less than the target value of 30%.
[0041] On the other hand, an embodiment of the present invention provides a data center joint optimization system, which includes: a data center load data set acquisition module, a GM model module, a bird flock algorithm module, a predicted quantity acquisition module, and an optimization strategy execution module. The data center acquisition module executes: acquiring the load conditions of the data center in a historical period to form a data center load data set, and obtaining a time series through the data center load data set; the GM model module executes: establishing a GM model, the GM model includes a GM model differential equation and a GM model prediction equation, and the time series obtains the GM model differential equation through the GM model, and the GM model differential equation includes an adjustment parameter: the bird flock algorithm module executes: using the bird flock algorithm to obtain the optimal value of the adjustment parameter; the predicted quantity acquisition module executes: bringing the optimal value into the GM model prediction equation in the GM model to obtain a predicted number series; the optimization strategy execution module executes: executing the data center optimization strategy according to the predicted sequence.
[0042] Implementing the embodiments of the present invention has the following beneficial effects:
[0043] Improve load forecasting accuracy: Through the time series analysis method based on the grey model (GM), the load changes of the data center can be predicted more accurately, thereby effectively improving the forecasting accuracy and meeting the needs of complex load scenarios.
[0044] Optimize resource allocation efficiency: Combine the bird flock algorithm to optimize the adjustment parameters of the GM model, so that resource allocation can be more efficient based on load forecasting, maximize data center resource utilization, and reduce resource waste.
[0045] Reduce data center energy consumption: Through the joint application of intelligent prediction and optimization strategies, unnecessary physical resource expansion and power consumption are reduced, significantly reducing the overall energy consumption of the data center.
[0046] Achieve green data center management.
[0047] Improve system response speed: This method can quickly respond and adjust optimization strategies when load surges or resources are scarce, enhance the real-time response capabilities of the data center, and ensure system stability and service quality.
[0048] Multi-level association optimization: By constructing a multi-layer Bayesian network for the physical layer and the service layer,
[0049] It further enhances the dynamic optimization effect of resources, enables effective collaboration between layers, and improves the rationality and flexibility of resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS
[0050] Figure 1 A schematic diagram of a data center joint optimization method provided in an embodiment of the present application. DETAILED DESCRIPTION
[0051] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0052] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.
[0053] like Figure 1 As shown, the method includes: obtaining the load conditions in the historical period of the data center to form a data center load data set, obtaining a time series through the data center load data set, establishing a GM model, the GM model includes a GM model differential equation and a GM model prediction equation, the time series obtains the GM model differential equation through the GM model, the GM model differential equation includes adjustment parameters, using a bird flock algorithm to obtain the optimal value of the adjustment parameter, bringing the optimal value into the GM model prediction equation in the GM model, obtaining a prediction series, and executing a data center optimization strategy according to the prediction series.
[0054] In the following, in conjunction with the accompanying drawings, a data center joint optimization method and system provided by the embodiments of the present application are described in detail through specific embodiments and application scenarios.
[0055] like Figure 1 As shown, the present application provides a data center joint optimization method, which may include the following steps S100 to S600.
[0056] Step S100 obtains the load conditions of the data center in the historical period, forms a data center load data set, and obtains a time series through the data center load data set;
[0057] Specifically, the load data set includes but is not limited to processor load, memory load, number of virtual machines, disk load, and network traffic. The data center load data set is sorted, and a fixed time interval is set. The data of each fixed time interval in the historical period is sorted to form a time series. The time series is X={x1,x2,x3,...,xn}, where n represents the data of the nth time interval.
[0058] Preferably, the fixed time interval can be dynamically adjusted according to the actual operation of the data center to better adapt to the load fluctuation characteristics. For example, when the load changes drastically, a shorter time interval can be set to increase the sampling frequency to ensure that the time series more accurately reflects the load change trend; while when the load sampling is in a relatively stable period, a long time interval can be selected to reduce the amount of data and computing overhead.
[0059] Furthermore, the load data set can select different indicator combinations according to different optimization requirements. For example, in resource utilization efficiency optimization, the focus is on processor load and memory load; in bandwidth optimization, network traffic data can be given priority. Through flexible indicator selection and time interval setting, the time series can more comprehensively and accurately characterize the data center load situation, providing stronger support for subsequent prediction and optimization.
[0060] In addition, in order to improve the robustness of the time series, the abnormal data can be pre-processed in the process of sorting out the historical data, including: smoothing the sudden load peak, or using interpolation methods to fill in the missing data to ensure that the generated time series has high accuracy and continuity.
[0061] Step S200: Establishing a GM model. The GM model includes a GM model differential equation and a GM model prediction equation. The time series obtains the GM model differential equation through the GM model. The GM model differential equation includes adjustment parameters.
[0062] Specifically, it also includes steps S201 to S205:
[0063] S201 marks the time series as the original sequence, recorded as:
[0064] x (0) =(x (0) (1),x (0) (2),…,x (0) (n));
[0065] S202 obtains a cumulative sequence based on the original sequence, and the formula is:
[0066]
[0067] This means that for each time instant i, we replace x in the original sequence (0) (1),x (0) (2),…,x (0) (i) Accumulate and obtain a new sequence x (1) , recorded as a cumulative series;
[0068] S203 obtains the grey derivative of a cumulative sequence, the formula is:
[0069] dx (1) (i) = x (1) (i) - x (1) (i-1)i=2,3,…,n
[0070] Among them, dx (1) (i) represents the grey derivative at the i-th moment, x (1) (i) and x (1) (i-1) represents the value of the cumulative sequence at the i-th and i-1-th moments respectively;
[0071] S204 constructs the mean generating sequence:
[0072]
[0073] z (1) (i) is the i-th element of the mean generating sequence, x (1) (i) and x (1) (i―1) is a cumulative sequence x (1) The values of the i-th and i-1-th in ;
[0074] S205 uses the mean value generating series and grey derivatives to establish the GM model difference equation:
[0075] dx (1) (i)+a·z (1) (i) = b
[0076] where dx (1) (i) is the grey derivative, z (1) (i) is the mean generating sequence, a and b are adjustment parameters.
[0077] Step S300 uses the bird flock algorithm to obtain the optimal value of the adjustment parameter;
[0078] Specifically, step S300 includes S301 to S303, wherein step S303 includes S3031 to S3033:
[0079] S301 initializes the bird flock. Each particle represents possible adjustment parameters a and b. During initialization, each dimension of the particle will randomly generate a certain value as the search starting point.
[0080] Each particle has a position and a velocity, and the velocity is the rate at which the particle updates its position;
[0081] S302 builds on the speed formula:
[0082]
[0083] Construct a formula about position:
[0084]
[0085] in, is the velocity of the particle at time t+1, is the position of the particle at time t+1, w is the inertia weight, c1 and c2 are learning factors, r1 and r2 are random numbers between [0, 1], is the optimal position of the particle in the individual, is the global optimal position.
[0086] S3031 sets the physical layer and service layer in the Bayesian network. The physical layer includes but is not limited to the energy consumption of the physical machine, the memory load rate of the physical machine, and the CPU resource load rate of the physical machine. The service layer includes but is not limited to the total node load, resource requirements, failure rate, and bandwidth limitation.
[0087] S3032 establishes a multi-layer Bayesian network, the service layer Bayesian network S = {X, Y, P}, where X represents a node set, Y represents an arc set, P represents a conditional probability set, and a physical layer Bayesian network B, which only includes a node set. A mapping relationship is established between the nodes in the physical layer Bayesian network B and the nodes in X in the service layer Bayesian network S. The mapping relationship set is T. The multi-layer Bayesian network is represented as
[0088] {S,B,T};
[0089] S3033 uses physical layer data prediction to obtain a first probability confidence M1, with a learning factor c1=3*M1, and uses service layer data prediction to obtain a second probability confidence M2, with a learning factor c2=3*M2.
[0090] Preferably, during the optimization process of the bird flock algorithm from step S301 to S303, the convergence speed and accuracy can be improved by dynamically adjusting the inertia weight w. Specifically, a larger inertia weight is set in the early stage of the algorithm to increase exploration, so that particles can find the global optimal solution in a larger search space; in the later stage of the algorithm, the inertia weight is gradually reduced to make the particles tend to converge locally, thereby improving the accuracy of the algorithm. In addition, an adaptive adjustment mechanism can be introduced to dynamically modify the inertia weight according to the current search state of the particles to further balance the effects of global search and local optimization.
[0091] Furthermore, in order to enhance the prediction effect of the multi-layer Bayesian network, the nodes in the Bayesian network can be weighted, that is, different weights can be assigned according to the importance of different nodes in the physical layer and service layer. For example, in the case of resource constraints, the weights of physical machine energy consumption and memory load rate can be increased, so that the Bayesian network pays more attention to the consumption of key resources, thereby improving the pertinence and effectiveness of the optimization strategy.
[0092] In addition, when setting learning factors c1 and c2, the probability confidence M1 and M2 can be dynamically updated according to the historical data of the data center load, so as to adaptively adjust the size of the learning factor. This dynamic learning factor adjustment mechanism can make the bird flock algorithm more accurately approach the optimal solution, improve the prediction accuracy and system stability, and achieve more efficient optimization configuration of data center resources.
[0093] Step S400: Substitute the optimal value into the GM model prediction equation in the GM model to obtain a prediction series;
[0094] Specifically, the learning factors c1 and c2 are used to output the optimal combination of ab through the speed and position formula in the bird flock algorithm, and the GM model is used for prediction and evaluation. The prediction formula is:
[0095]
[0096] Get the predicted series X through the prediction formula (1)
[0097] Specifically, the method further includes step S410:
[0098] The predicted series is compared with the original series through RMSE threshold evaluation, the formula is:
[0099]
[0100] The RMSE threshold is used to evaluate whether the predicted series is usable.
[0101] Step S500 executes the data center optimization strategy according to the predicted sequence;
[0102] Specifically, the target value of RMSE value less than 30% is based on the series X (1) Adjust the optimization execution strategy of the data center;
[0103] Optimization execution strategies include but are not limited to expanding physical machines, limiting network traffic, adjusting virtual machines, etc.
[0104] Preferably, the optimization execution strategy can be dynamically adjusted according to the real-time load and resource usage of the data center to ensure the maximum resource utilization efficiency and the stability of system performance. Specifically, the following strategies are included:
[0105] Dynamic resource allocation: When the predicted load is higher than the preset threshold, the resource expansion strategy is automatically triggered, such as starting a spare physical server or increasing the number of virtual machines to cope with the sudden increase in load. Conversely, when the load decreases, some resources can be automatically released to reduce unnecessary energy consumption.
[0106] Intelligent network traffic management: When network traffic approaches bandwidth limit, it will prioritize limiting the traffic of low-priority tasks, or distribute the traffic to server nodes with lighter loads through load balancing to avoid network congestion and improve data transmission efficiency.
[0107] Load balancing optimization: Based on the load distribution information provided by the prediction series, dynamically adjust the load distribution of each server and virtual machine to ensure load balance on each node, thereby extending the service life of the equipment and reducing the failure rate.
[0108] Energy efficiency management strategy: When the load is low, reduce energy consumption and improve the overall energy efficiency of the data center by merging computing tasks, shutting down some servers or switching to energy-saving mode.
[0109] Fault warning and redundant backup: Identify potential load bottlenecks and fault risks based on predicted series, start redundant backup strategies in advance, or issue system warnings to reduce the risk of data center downtime and service interruptions.
[0110] Through the above-mentioned preferred optimization execution strategy, the data center can achieve flexible and accurate resource management when the load fluctuates, effectively reduce operating costs and improve service quality.
[0111] Step S600: If the RMSE value is less than the target value of 30%, the Bayesian model is adjusted according to the evaluation result.
[0112] A data center joint optimization system, the data center joint optimization system includes: a data center load data set acquisition module, a GM model module, a bird flock algorithm module, a prediction quantity acquisition module, and an optimization strategy execution module. The data center acquisition module executes: acquiring the load situation of the data center in the historical period to form a data center load data set, and obtaining a time series through the data center load data set; the GM model module executes: establishing a GM model, the GM model includes a GM model differential equation and a GM model prediction equation, and the time series obtains the GM model differential equation through the GM model, and the GM model differential equation includes an adjustment parameter; the bird flock algorithm module executes: using the bird flock algorithm to obtain the optimal value of the adjustment parameter; the prediction quantity acquisition module executes: bringing the optimal value into the GM model prediction equation in the GM model to obtain a prediction series; the optimization strategy execution module executes: executing the data center optimization strategy according to the prediction series.
[0113] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods of each embodiment of the present application.
[0114] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present application and the claims, all of which belong to the scope of protection of the present application.
Claims
1. A data center joint optimization method, characterized in that: The method comprises: S100 obtains the load condition of the data center in the historical period, forms a data center load data set, and obtains a time series through the data center load data set; S200 establishes a GM model, wherein the GM model includes a GM model differential equation and a GM model prediction equation, wherein the time series obtains the GM model differential equation through the GM model, and the GM model differential equation includes an adjustment parameter; S300 uses a bird flock algorithm to obtain an optimal value of the adjustment parameter; S400: Substitute the optimal value into the GM model prediction equation in the GM model to obtain a prediction series; S500 executes a data center optimization strategy according to the prediction sequence.
2. A data center joint optimization method according to claim 1, characterized in that: In step S100, the load data set includes but is not limited to processor load, memory load, number of virtual machines, disk load, and network traffic. The data center load data set is sorted, a fixed time interval is set, and the data of each fixed time interval in the historical period is sorted to form the time series. The time series is X={x1, x2, x3, ..., xn}, where n represents the data of the nth time interval.
3. A data center joint optimization method according to claim 1, characterized in that: The establishing of the GM model in S200 includes: S201 marks the time series as the original sequence, which is recorded as: x (0) =(x (0) (1),x (0) (2),…,x (0) (n)); S202 obtains a cumulative sequence according to the original sequence, the formula is: This means that for each moment i, we replace x in the original sequence (0) (1),x (0) (2),…,x (0) (i) Accumulate and obtain a new sequence x (1) , recorded as a cumulative series; S203 obtains the grey derivative of the primary accumulation sequence, the formula is: dx (1) (i)=x (1) (i)―x (1) (i―1)i=2,3,…,n Among them, dx (1) (i) represents the grey derivative at the i-th moment, x (1) (i) and x (1) (i-1) represents the value of the primary accumulated sequence at the i-th and i-1-th moments respectively; S204 constructs the mean generating sequence: z (1) (i) is the i-th element of the mean generating sequence, x (1) (i) and x (1) (i-1) is the once accumulated sequence x (1) The values of the i-th and i-1-th in ; S205 uses the mean value generation sequence and the grey derivative to establish the GM model differential equation: dx (1) (i)+a·z (1) (i)=b where dx (1) (i) is the grey derivative, z (1) (i) is the mean generating sequence, a and b are the adjustment parameters.
4. A data center joint optimization method according to claim 3, characterized in that: The step S300 specifically includes: S301 initializes the bird flock, each particle represents the possible adjustment parameters a and b, and during initialization, each dimension of the particle will randomly generate a certain value as a search starting point, and each particle has a position and a speed, and the speed is the rate at which the particle updates its position; S302 builds the speed formula: Construct a formula about position: in, is the velocity of the particle at time t+1, is the position of the particle at time t+1, w is the inertia weight, c1 and c2 are learning factors, r1 and r2 are random numbers between [0, 1], is the optimal position of the particle in the individual, is the global optimal position.
5. A data center joint optimization method according to claim 4, characterized in that: The step S303 is also included, and the step S303 further includes: S3031 sets the physical layer and service layer in the Bayesian network, wherein the physical layer includes but is not limited to the energy consumption of the physical machine, the memory load rate of the physical machine, and the CPU resource load rate of the physical machine, and the service layer includes but is not limited to the total node load, resource requirements, failure rate, and bandwidth limitation; S3032 establishes a multi-layer Bayesian network, the service layer Bayesian network S = {X, Y, P}, where X represents a node set, Y represents an arc set, and P represents a conditional probability set, a physical layer Bayesian network B, and the physical layer Bayesian network B only includes a node set, and establishes a mapping relationship between the nodes in the physical layer Bayesian network B and the nodes in X in the service layer Bayesian network S, the mapping relationship set is T, and the multi-layer Bayesian network is represented by {S, B, T} S3033 uses the physical layer data to predict and obtain a first probability confidence M1, where the learning factor c1=3*M1, and uses the service layer data to predict and obtain a second probability confidence M2, where the learning factor c2=3*M2.
6. A data center joint optimization method according to claim 5, characterized in that: The step S400 specifically includes: The learning factors c1 and c2 are used to output the optimal combination of ab through the speed and position formula in the bird flock algorithm, and prediction and evaluation are performed through the GM model. The prediction formula is: The prediction series X is obtained by the prediction formula (1) .
7. The method for joint optimization of data centers according to claim 3, characterized in that: The step S410 is also included: The predicted sequence is compared with the original sequence through RMSE threshold evaluation, and the formula is: Whether the predicted series is available is determined based on the RMSE threshold evaluation.
8. The method for joint optimization of data centers according to claim 7, characterized in that: The S500 also includes: The target value of RMSE value less than 30% is based on the series X (1) Adjusting the optimization execution strategy of the data center; The optimization execution strategy includes but is not limited to expanding the capacity of physical machines, limiting network traffic, adjusting virtual machines, etc.
9. The method for joint optimization of data centers according to claim 7, characterized in that: The step S600 is also included: S600 adjusts the Bayesian model according to the evaluation result if the RMSE value is less than the target value of 30%.
10. A data center joint optimization system, the data center joint optimization system comprising: A data center load data set acquisition module, a GM model module, a bird flock algorithm module, a prediction quantity acquisition module, and an optimization strategy execution module. The data center acquisition module executes: acquiring the load conditions of the data center in the historical period to form a data center load data set, and obtaining a time series through the data center load data set; the GM model module executes: establishing a GM model, the GM model includes a GM model differential equation and a GM model prediction equation, and the time series obtains the GM model differential equation through the GM model, and the GM model differential equation includes an adjustment parameter: the bird flock algorithm module executes: using the bird flock algorithm to obtain the optimal value of the adjustment parameter; the prediction quantity acquisition module executes: bringing the optimal value into the GM model prediction equation in the GM model to obtain a prediction series; the optimization strategy execution module executes: executing the data center optimization strategy according to the prediction sequence.