Cloud storage resource allocation optimization method and system based on deep learning

Abstract

The invention relates to the technical field of cloud storage, in particular to a cloud storage resource allocation optimization method and system based on deep learning, and the method comprises the steps: obtaining historical resource use data and current resource state data of a cloud storage system; a processing step: constructing a topology group prediction model based on the historical resource usage data; generating a resource demand prediction result according to the topology group prediction model and the current resource state data; constructing a quantum optimization model based on the resource demand prediction result; generating a resource allocation scheme according to the quantum optimization model; constructing a dynamic adjustment model based on the resource allocation scheme and the real-time system state; and according to the dynamic adjustment model, outputting an optimized resource allocation scheme, and by constructing a prediction model based on a topology group, effectively capturing a high-dimensional structure and dynamic characteristics of a resource use mode in the cloud storage system, and significantly improving the accuracy of resource demand prediction.

Description

Technical Field

[0001] The present invention relates to the field of cloud storage technology, and more specifically, to a cloud storage resource allocation optimization method and system based on deep learning. Background Art

[0002] With the rapid development of cloud computing technology, cloud storage systems have become an important infrastructure in today's digital world. However, with the continuous growth of user demand and the explosive growth of data volume, how to efficiently allocate and manage cloud storage resources has become an increasingly prominent issue. Traditional cloud storage resource allocation methods usually adopt static or semi-static strategies, which are difficult to adapt to rapidly changing user needs and complex workload patterns.

[0003] In recent years, some researchers have tried to apply machine learning techniques to the problem of cloud storage resource allocation. These methods usually train prediction models based on historical data and then allocate resources based on the prediction results. Although this method is an improvement over traditional methods, it still has some obvious shortcomings. First, most existing machine learning methods use simple regression or classification models, which are difficult to capture the complex nonlinear dynamic characteristics of cloud storage systems. Second, these methods often only focus on a single resource type (such as storage space) and ignore the mutual influence and constraint relationship between different resources. Furthermore, existing methods usually use a fixed prediction time window, which makes it difficult to balance short-term and long-term resource requirements.

[0004] In addition, most existing resource allocation methods use greedy algorithms or heuristic algorithms to optimize resource allocation schemes. Although these algorithms have fast calculation speeds, they can only obtain local optimal solutions and it is difficult to achieve optimal resource allocation on a global scale. Especially in large-scale cloud storage systems, the combinatorial space of resource allocation is extremely large, and traditional optimization algorithms are difficult to find a satisfactory solution within an acceptable time.

[0005] Finally, existing resource allocation methods generally lack effective mechanisms for handling system dynamics and uncertainty. The workload and user demand of cloud storage systems may fluctuate dramatically due to various factors (such as emergencies, seasonal changes, etc.). Existing methods are often unable to perceive and respond to these changes in a timely manner, resulting in a serious disconnect between resource allocation plans and actual needs, affecting system performance and user experience.

[0006] In view of the above problems, there is an urgent need for a cloud storage resource allocation method that can comprehensively consider multiple resource types, capture the complex dynamic characteristics of the system, achieve global optimization, and quickly adapt to environmental changes. The present invention is an innovative solution to this demand. Summary of the invention

[0007] The cloud storage resource allocation optimization method and system based on deep learning proposed in this invention effectively solves many problems existing in the prior art by innovatively combining cutting-edge technologies such as topology, quantum computing and chaos theory. This method can fully capture the complex dynamic characteristics of cloud storage systems, realize the coordinated optimization of multiple resource types, quickly find the global optimal or approximate optimal solution in a huge solution space, and flexibly respond to dynamic changes in the system.

[0008] The present invention provides a cloud storage resource allocation optimization method based on deep learning, comprising:

[0009] The acquisition steps include:

[0010] Obtain historical resource usage data and current resource status data of the cloud storage system;

[0011] Processing steps include:

[0012] Based on the historical resource usage data, construct a topological group prediction model;

[0013] Generate a resource demand prediction result according to the topology group prediction model and the current resource status data;

[0014] Based on the resource demand prediction results, construct a quantum optimization model;

[0015] Generate a resource allocation plan according to the quantum optimization model;

[0016] Building a dynamic adjustment model based on the resource allocation scheme and real-time system status;

[0017] Output steps include:

[0018] According to the dynamic adjustment model, an optimized resource allocation plan is output.

[0019] Preferably, the obtaining step specifically includes:

[0020] Obtain resource usage logs of the cloud storage system within a preset time period;

[0021] Extract resource usage, timestamp and user request information from the resource usage log; obtain the current CPU usage, memory occupancy and storage space usage of the cloud storage system. Preferably, the construction of the topology group prediction model specifically includes:

[0022] Mapping the historical resource usage data to a topological space;

[0023] Define group actions to describe resource state transfer;

[0024] Construct kernel functions to capture the importance of group elements;

[0025] Based on the Haar measure, a resource demand prediction function is generated.

[0026] Preferably, generating resource demand prediction results specifically includes:

[0027] Inputting the current resource status data into the topology group prediction model;

[0028] Calculate the group integral to get the prediction result;

[0029] The prediction result is normalized.

[0030] Preferably, the construction of the quantum optimization model specifically includes:

[0031] Mapping resource allocation problems to quantum systems;

[0032] Define quantum states to represent different resource allocation schemes;

[0033] Construct Hamiltonian operators to describe system energy and constraints;

[0034] Set the initial quantum state.

[0035] Preferably, the generating resource allocation scheme specifically includes:

[0036] Solve the Schrödinger equation to obtain the quantum state evolution;

[0037] Measuring quantum states at specific points in time;

[0038] Convert the measurement results into specific resource allocation plans.

[0039] Preferably, the constructing of the dynamic adjustment model specifically includes:

[0040] Calculate the system's Lyapunov exponents to characterize instability;

[0041] Calculate the fractal dimension of a system to describe complexity;

[0042] Based on Lyapunov exponent and fractal dimension, a dynamic adjustment function is constructed.

[0043] Preferably, the output optimized resource allocation scheme specifically includes:

[0044] combining the resource demand forecast result and the resource allocation plan;

[0045] Applying the dynamic adjustment function to perform weighting;

[0046] Generate the final resource allocation instructions.

[0047] Preferably, the step of feedback is also included:

[0048] Monitor system performance indicators after resource allocation;

[0049] The performance indicator is fed back to the acquisition step to update the historical resource usage data.

[0050] The cloud storage resource allocation optimization system based on deep learning includes:

[0051] An acquisition module is used to obtain historical resource usage data and current resource status data of the cloud storage system;

[0052] A topology group prediction module, used to construct a topology group prediction model based on the historical resource usage data, and generate a resource demand prediction result according to the topology group prediction model and the current resource status data;

[0053] A quantum optimization module, used to build a quantum optimization model based on the resource demand prediction result, and generate a resource allocation plan according to the quantum optimization model;

[0054] A dynamic adjustment module, used to construct a dynamic adjustment model based on the resource allocation scheme and real-time system status;

[0055] The output module is used to output the optimized resource allocation plan according to the dynamic adjustment model.

[0056] The beneficial effects of the present invention are mainly reflected in the following aspects:

[0057] Specifically, the method of the present invention can effectively capture the high-dimensional structure and dynamic characteristics of resource usage patterns in cloud storage systems by constructing a prediction model based on topological groups, significantly improving the accuracy of resource demand prediction. The introduction of quantum optimization models breaks through the limitations of traditional optimization algorithms, and uses the parallelism and superposition characteristics of quantum computing to quickly explore in a huge solution space, greatly improving the global optimality of resource allocation schemes. In addition, the present invention also introduces a dynamic adjustment mechanism based on chaos theory and fractal geometry, which enables the system to keenly perceive environmental changes and respond quickly, effectively improving the flexibility and adaptability of resource allocation.

[0058] These innovative technical solutions work together and complement each other to form a highly intelligent and adaptive resource allocation system. In this way, the cloud storage system can more accurately predict resource requirements, allocate resources more efficiently, and respond to environmental changes more flexibly. This not only significantly improves resource utilization and reduces operating costs, but also greatly improves system performance and user experience. In particular, in large-scale, highly dynamic, and multi-variable complex cloud storage environments, the advantages of the present invention are more prominent.

[0059] In summary, the present invention provides a brand-new solution for the field of cloud storage resource management, opens up a new way to improve the efficiency, reliability and scalability of cloud storage systems, and has important theoretical significance and broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 The figure is a flow chart of the method of the present invention.

[0061] Figure 2 It is a logic block diagram of the acquisition module of the present invention.

[0062] Figure 3 It is a logic block diagram of the topological group prediction module of the present invention.

[0063] Figure 4 This is a logical block diagram of the quantum optimization module of the present invention.

[0064] Figure 5 It is a logic block diagram of the dynamic adjustment module of the present invention.

[0065] Figure 6 It is a logic block diagram of the output module of the present invention.

[0066] Figure 7 It is a logic block diagram of the feedback module of the present invention. DETAILED DESCRIPTION

[0067] Please refer to Figure 1-7 The present invention provides a cloud storage resource allocation optimization method and system based on deep learning. The method aims to solve the problem that resource allocation in existing cloud storage systems is not accurate and dynamic enough, and achieves more intelligent and efficient resource management by introducing deep learning technology.

[0068] According to one embodiment of the present invention, the method includes an acquisition step, a processing step, and an output step. In the acquisition step, the system first acquires historical resource usage data and current resource status data of the cloud storage system. These data are the basis of the subsequent optimization process and are crucial for accurately predicting future resource requirements.

[0069] Preferably, the historical resource usage data includes the resource usage of the cloud storage system in the past period of time (e.g., the last 30 days), such as CPU usage, memory usage, storage space usage, etc. The current resource status data reflects the real-time status of the system, including but not limited to the current resource utilization, user request queue length, etc.

[0070] In the processing step, the method of the present invention first constructs a topological group prediction model based on historical resource usage data. This is an innovative step that maps the resource usage of the cloud storage system into the topological space and uses group theory methods to describe the changes in resource status.

[0071] Specifically, a topological group G is defined to represent the cloud storage system, and a resource configuration space X. The group action φ:G×X→X is used to describe the resource allocation process.

[0072] Based on this model, a prediction function is constructed:

[0073] f(x)=∫ G K(g)·φ(g,x)dμ(g),

[0074] Among them, K(g) is the kernel function used to capture the importance of group elements; μ(g) is the Haar measure on G; x is the current resource configuration. This function comprehensively considers all possible resource state transitions in the system by integrating all elements of the group G.

[0075] Next, the method generates a resource demand forecast result based on the topology group forecast model and the current resource state data. This step inputs the current system state into the forecast model to obtain a resource demand forecast for a period of time in the future (eg, the next hour).

[0076] After obtaining the resource demand prediction results, the method of the present invention further constructs a quantum optimization model. This is another innovation point, mapping the resource allocation problem into the quantum system and using the advantages of quantum computing to find the optimal solution. Where |i> represents the i-th resource allocation scheme, α i is the complex amplitude.

[0077] At the same time, the Hamiltonian operator is constructed:

[0078] H=H 0 +λH 1 ,

[0079] Among them, H 0 represents the basic energy of the system, H 1 represents the resource allocation constraint, and λ is the Lagrange multiplier.

[0080] Based on the quantum optimization model, this method generates a resource allocation plan by solving the Schrödinger equation:

[0081]

[0082] The solution is:

[0083]

[0084] By measuring |ψ(t)>, the optimal resource allocation plan can be obtained.

[0085] In order to cope with the dynamics and uncertainty of the cloud environment, the method of the present invention also introduces a dynamic adjustment model. The model is built based on the resource allocation scheme and the real-time system status, and uses the concepts of chaos theory and fractal geometry to describe the complexity and instability of the system.

[0086] Specifically, the Lyapunov exponent is defined as

[0087]

[0088] Where δx(t) represents the deviation of the system state at time t. At the same time, the fractal dimension D is introduced to describe the complexity of the system:

[0089]

[0090] where N(∈) is the number of boxes of size ∈ required to cover the system.

[0091] Based on these indicators, a dynamic adjustment function is constructed:

[0092]

[0093] In the output step, this method outputs the optimized resource allocation plan based on the dynamic adjustment model. The final resource allocation result is:

[0094] R(t)=(1-A(t))·R p (t)+A(t)·R q (t),

[0095] Among them, R p (t) is the prediction result of the topological group prediction model, R q (t) is the optimization result of the quantum optimization model.

[0096] One advantage of the method of the present invention is that it makes comprehensive use of cutting-edge mathematical tools such as topology, quantum mechanics and chaos theory to make the resource allocation process more accurate and dynamic. The topological group prediction model can capture the overall structure and dynamic characteristics of the system, the quantum optimization model can quickly find the optimal solution in the huge solution space, and the dynamic adjustment model ensures that the system can quickly respond to environmental changes.

[0097] In addition, this method also includes obtaining the resource usage log of the cloud storage system within a preset time period and extracting key data such as resource usage, timestamps, and user request information. These data provide a solid foundation for subsequent prediction and optimization processes. At the same time, the system also obtains the current CPU usage, memory occupancy, and storage space usage in real time to ensure that resource allocation decisions are based on the latest system status.

[0098] It is worth noting that this method adopts innovative steps when constructing the topological group prediction model. First, the historical resource usage data is mapped to the topological space, which makes it possible to understand the resource usage pattern from a higher dimension. Then, by defining group actions to describe resource state transitions, this method can capture complex resource dynamic changes. Next, a kernel function is constructed to capture the importance of group elements, which helps to identify key resource state transitions. Finally, a resource demand prediction function is generated based on the Haar measure, which ensures the mathematical rigor of the prediction process.

[0099] In general, the method provided by the present invention realizes the intelligent and dynamic allocation of cloud storage resources through deep learning technology and advanced mathematical tools, effectively improves resource utilization, reduces operating costs, and ensures the stability and consistency of user experience. In a preferred embodiment of the present invention, the process of generating resource demand forecast results further includes inputting current resource status data into the topology group forecasting model. This step ensures that the forecasting process fully considers the latest status of the system, thereby improving the accuracy and timeliness of the forecast.

[0100] Specifically, the current resource status data may include key indicators such as CPU usage, memory occupancy, storage space usage, and network bandwidth utilization. These indicators reflect the real-time operating status of the cloud storage system at the prediction time. By inputting these data into the topological group prediction model, this method can more accurately evaluate the position of the current state of the system in the topological space, thereby providing a more reliable starting point for subsequent predictions.

[0101] Next, the method of the present invention obtains the prediction result by calculating the group integral. This step involves performing an integral operation on the function defined on the topological group. Preferably, the Monte Carlo method can be used to approximate this integral. For example, N points can be randomly sampled on the group (where N can be determined according to the required accuracy and computing resources, and a value between 1000 and 10000 can usually be selected), and then the average of the function values ​​at these points is calculated as an approximate value of the integral.

[0102] After obtaining the prediction results, this method will also normalize them. The purpose of this step is to convert the prediction results to a standardized scale for subsequent processing and comparison. Usually, the minimum-maximum normalization method can be used to linearly map the prediction results to the [0,1] interval. The specific formula is as follows:

[0103]

[0104] Among them, x is the original predicted value, x min and x max are the minimum and maximum values ​​in the prediction results, respectively, normalizedis the normalized result.

[0105] In another embodiment of the present invention, the process of constructing a quantum optimization model is further refined. First, the resource allocation problem is mapped into a quantum system. The core idea of ​​this step is to encode each possible resource allocation scheme into a base state of a quantum state. For example, if the system has n resource units and m tasks, then an n×m binary matrix can be used to represent an allocation scheme, and this matrix is ​​encoded into a quantum state.

[0106] Next, we define quantum states to represent different resource allocation schemes. Here, we can use quantum superposition states to represent multiple possible allocation schemes at the same time. For example, for a simple system, we can define:

[0107] |ψ>=α|00>+β|01>+γ|10>+δ|11>,

[0108] Among them, |00>,|01>,|10>,|11> represent four different resource allocation schemes, and α, β, γ, and δ are complex amplitudes that satisfy |α| 2 +|β| 2 +|γ| 2 +|δ| 2 =1.

[0109] Then, this method constructs the Hamiltonian operator to describe the system energy and constraints. The Hamiltonian operator usually consists of two parts: the problem Hamiltonian and the driving Hamiltonian. The problem Hamiltonian encodes the objective function and constraints of the optimization problem, while the driving Hamiltonian is used to drive the system evolution. For example, for the resource allocation problem, it can be defined as:

[0110] H=H p +H d

[0111] Among them, H p For the Hamiltonian problem, H d To drive the Hamiltonian.

[0112] Finally, set the initial quantum state. Usually, you can choose the ground state of the system as the initial state, or choose a uniform superposition state. For example:

[0113]

[0114] Wherein, N is the total number of possible resource allocation schemes. When generating a resource allocation scheme, the method of the present invention first solves the Schrödinger equation to obtain the quantum state evolution. This step can be achieved by numerical methods, such as using the Runge-Kutta method. Specifically, time can be discretized and then the evolution operator is applied at each time step:

[0115]

[0116] Among them, Δt is the time step, which can be selected according to the required accuracy, usually 1 / 100 to 1 / 1000 of the system characteristic time scale.

[0117] Next, the quantum state is measured at a specific time point. The choice of this time point is critical, and it is usually necessary to determine the best measurement time through multiple experiments. The measurement process will project the quantum state onto a certain ground state, corresponding to a specific resource allocation scheme.

[0118] Finally, the measurement results are converted into specific resource allocation plans. This step involves decoding the measurement results of the quantum state into actual resource allocation instructions. For example, if the measured state is |1010>, it may correspond to allocating the 1st and 3rd resource units to a specific task.

[0119] Through the above steps, the method of the present invention can take advantage of quantum computing to quickly find a near-optimal resource allocation solution in a huge solution space, thereby significantly improving the resource utilization efficiency of the cloud storage system. In another embodiment of the present invention, the process of constructing a dynamic adjustment model is further refined. First, the method calculates the Lyapunov exponent of the system to characterize instability. The Lyapunov exponent is an important indicator for measuring the sensitivity of a dynamic system to initial conditions. In a cloud storage system, this index can reflect the severity of changes in resource demand.

[0120] Specifically, the calculation of the Lyapunov exponent can be based on the time series data of the system state. For example, the CPU usage rate can be selected as a representative of the system state. Assuming that the CPU usage rates at time t and t+Δt are x(t) and x(t+Δt) respectively, the Lyapunov exponent can be estimated by the following formula:

[0121]

[0122] Where N is the number of sampling points, Δx 0 is the initial perturbation size. Usually, N = 1000 to 10000 can be selected to obtain a more stable estimation result.

[0123] Next, this method calculates the fractal dimension of the system to describe the complexity. The fractal dimension reflects the self-similarity of the system at different scales and can be used to quantify the complexity of the resource usage pattern of the cloud storage system. A common method for calculating the fractal dimension is the box counting method. The specific steps are as follows:

[0124] 1. Divide the system state space into grids of size ε.

[0125] 2. Calculate the number of non-empty grids N(ε).

[0126] 3. Repeat steps 1 and 2, using different sizes of ε.

[0127] 4. Plot the graph of log(N(ε)) versus log(1 / ε). The slope is the estimate of the fractal dimension.

[0128] Preferably, the value of ε may be selected to range from 1 / 100 to 1 / 10 of the characteristic scale of the system to capture the system behaviors at different scales.

[0129] Based on Lyapunov exponent and fractal dimension, this method constructs a dynamic adjustment function. The design goal of this function is to strike a balance between system stability and adaptability. One possible construction method is:

[0130]

[0131] Among them, k is a parameter for adjusting the steepness of the function, and θ is a threshold parameter. By adjusting these parameters, the response speed of the system to changes can be controlled. For example, when k = 0.1, θ = 5, the system will respond moderately to moderate changes. In a preferred embodiment of the present invention, the process of outputting an optimized resource allocation plan includes several key steps. First, the method combines the resource demand forecast results and the resource allocation plan. This step aims to balance the long-term trend of the forecast and the short-term effect of the optimization. Next, a dynamic adjustment function is applied for weighting. Specifically, the following formula can be used:

[0132] R(t)=(1-A(t))·R p (t)+A(t)·R q (t),

[0133] Among them, Rp(t) is the result of the prediction model, Rq(t) is the result of the quantum optimization model, and A(t) is the dynamic adjustment function defined above. This weighted approach allows the system to dynamically adjust between stability and flexibility.

[0134] Finally, the method generates the final resource allocation instructions. These instructions may include specific operations such as creation, deletion or migration of virtual machines, allocation or recovery of storage space, etc. Preferably, these instructions should be generated in a format that can be directly executed by the cloud storage system, such as JSON or YAML format.

[0135] The method of the present invention also includes a feedback step, which is essential for continuously optimizing system performance. In this step, the system performance indicators after resource allocation are first monitored. These indicators may include but are not limited to:

[0136] 1. Response time: The average response time should be kept below 100ms.

[0137] 2. Throughput: The number of requests that the system can handle per second, which may be between 1,000 and 10,000 depending on the size of the system.

[0138] 3. Resource utilization: The average utilization of various resources (CPU, memory, storage) should be maintained between 60% and 80%.

[0139] 4. Quality of Service (QoS) default rate: should be less than 0.1%.

[0140] The specific thresholds of these performance indicators may vary depending on the specific application scenario, and system administrators can adjust them according to actual needs.

[0141] After collecting these performance indicators, this method feeds them back to the acquisition step to update the historical resource usage data. This closed-loop design ensures that the system can continuously learn and adapt to new usage patterns and demand changes. For example, if the utilization rate of a certain type of resource is found to be consistently low, the system may reduce the weight of this type of resource in future allocations.

[0142] According to a preferred embodiment of the present invention, the cloud storage resource allocation optimization system based on deep learning includes the following key modules:

[0143] Acquisition module 1 is used to obtain historical resource usage data and current resource status data of the cloud storage system. This module is responsible for collecting raw data from various data sources and performing necessary preprocessing operations such as data cleaning and standardization. Acquisition module 1 can regularly extract relevant information from the log files, monitoring systems, and user request records of the cloud storage system to ensure that the system has sufficient data to support subsequent prediction and optimization processes.

[0144] The topological group prediction module 2 is used to build a topological group prediction model based on historical resource usage data, and generate resource demand prediction results based on the model and current resource status data. This module implements the topological group model construction process described above, including steps such as mapping resource usage data to topological space, defining group effects, and constructing kernel functions. The topological group prediction module 2 is also responsible for running the constructed model to generate resource demand forecasts for a period of time in the future (such as the next hour).

[0145] Quantum optimization module 3 is used to build a quantum optimization model based on the resource demand prediction results and generate a resource allocation plan based on the model. This module first maps the resource allocation problem into the quantum system, defines quantum states to represent different resource allocation plans, and constructs Hamiltonian operators to describe the system energy and constraints. Then, quantum optimization module 3 solves the Schrödinger equation and obtains the optimized resource allocation plan by measuring the quantum state.

[0146] The dynamic adjustment module 4 is used to construct a dynamic adjustment model based on the resource allocation scheme and the real-time system status. This module calculates the Lyapunov index and fractal dimension of the system to evaluate the instability and complexity of the system. Based on these indicators, the dynamic adjustment module 4 constructs a dynamic adjustment function to balance between the prediction results and the optimization results.

[0147] The output module 5 is used to output the optimized resource allocation plan according to the dynamic adjustment model. This module comprehensively considers the prediction results of the topological group prediction model and the optimization results of the quantum optimization model, applies the dynamic adjustment function for weighting, and finally generates executable resource allocation instructions. These instructions may include specific operations such as the creation, deletion or migration of virtual machines, the allocation or recovery of storage space, etc.

[0148] These modules work closely together to form a complete closed-loop system. The acquisition module 1 provides the necessary input data, the topology group prediction module 2 and the quantum optimization module 3 are responsible for prediction and optimization respectively, the dynamic adjustment module 4 ensures that the system can adapt to environmental changes, and the output module 5 converts the optimization results into actual executable operations. Through this modular design, the system of the present invention can flexibly cope with various complex cloud storage scenarios and achieve efficient allocation and utilization of resources.

[0149] Preferably, the system may also include a feedback module 6 for monitoring system performance indicators after resource allocation and feeding back these indicators to the acquisition module 1. This design enables the system to continuously learn and adapt to new usage patterns and demand changes, further improving the accuracy and efficiency of resource allocation.

[0150] In general, the cloud storage resource allocation optimization system based on deep learning provided by the present invention realizes the intelligent and dynamic allocation of cloud storage resources by comprehensively applying advanced technologies such as topology, quantum computing and chaos theory. The system can effectively improve resource utilization, reduce operating costs, and ensure the stability and consistency of user experience, providing a powerful resource management tool for cloud storage service providers.

[0151] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, replacement, and improvement made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A cloud storage resource allocation optimization method based on deep learning, characterized in that: include: The acquisition steps include: Obtain historical resource usage data and current resource status data of the cloud storage system; Processing steps include: Based on the historical resource usage data, construct a topological group prediction model; Generate a resource demand prediction result according to the topology group prediction model and the current resource status data; Based on the resource demand prediction results, construct a quantum optimization model; Generate a resource allocation plan according to the quantum optimization model; Building a dynamic adjustment model based on the resource allocation scheme and real-time system status; Output steps include: According to the dynamic adjustment model, an optimized resource allocation plan is output.

2. The method according to claim 1, characterized in that The acquisition step specifically includes: Obtain resource usage logs of the cloud storage system within a preset time period; Extracting resource usage, timestamp and user request information from the resource usage log; Get the current CPU usage, memory usage, and storage space usage of the cloud storage system.

3. The method according to claim 1, characterized in that The construction of the topological group prediction model specifically includes: Mapping the historical resource usage data to a topological space; Define group actions to describe resource state transfer; Construct kernel functions to capture the importance of group elements; Based on the Haar measure, a resource demand prediction function is generated.

4. The method according to claim 1, characterized in that: The generating resource demand prediction result specifically includes: Inputting the current resource status data into the topology group prediction model; Calculate the group integral to get the prediction result; The prediction result is normalized.

5. The method according to claim 1, characterized in that The construction of the quantum optimization model specifically includes: Mapping resource allocation problems to quantum systems; Define quantum states to represent different resource allocation schemes; Construct Hamiltonian operators to describe system energy and constraints; Set the initial quantum state.

6. The method according to claim 1, characterized in that The generating resource allocation scheme specifically includes: Solve the Schrödinger equation to obtain the quantum state evolution; Measuring quantum states at specific points in time; Convert the measurement results into specific resource allocation plans.

7. The method according to claim 1, characterized in that The construction of the dynamic adjustment model specifically includes: Calculate the system's Lyapunov exponents to characterize instability; Calculate the fractal dimension of a system to describe complexity; Based on Lyapunov exponent and fractal dimension, a dynamic adjustment function is constructed.

8. The method according to claim 1, characterized in that The output optimized resource allocation scheme specifically includes: combining the resource demand forecast result and the resource allocation plan; Applying the dynamic adjustment function to perform weighting; Generate the final resource allocation instructions.

9. The method according to claim 1, characterized in that: Also includes a feedback step: Monitor system performance indicators after resource allocation; The performance indicator is fed back to the acquisition step to update the historical resource usage data.

10. A cloud storage resource allocation optimization system based on deep learning, characterized in that: include: An acquisition module is used to obtain historical resource usage data and current resource status data of the cloud storage system; A topology group prediction module, used to construct a topology group prediction model based on the historical resource usage data, and generate a resource demand prediction result according to the topology group prediction model and the current resource status data; A quantum optimization module, used to build a quantum optimization model based on the resource demand prediction result, and generate a resource allocation plan according to the quantum optimization model; A dynamic adjustment module, used to construct a dynamic adjustment model based on the resource allocation scheme and real-time system status; The output module is used to output the optimized resource allocation plan according to the dynamic adjustment model.

Images