Cloud mobile phone resource management method and related equipment

Abstract

The invention discloses a cloud mobile phone resource management method and related equipment, and relates to the technical field of cloud mobile phones, and the method comprises the steps: obtaining performance index data collected by a lightweight agent deployed in a virtualization layer; based on the performance index data, performing rolling prediction by using a time sequence prediction model to obtain a resource demand prediction result; based on the resource demand prediction result, a preset cost constraint and a preset service level constraint, performing solving processing through a multi-target optimization model to obtain a target resource allocation scheme; and according to the target resource allocation scheme, executing a resource dynamic allocation operation of the cross-cloud platform by using the multi-cloud adapter. The cloud mobile phone resource dynamic management process with high prediction precision, intelligent scheduling capability and multi-cloud elastic migration capability is constructed through technologies of lightweight data acquisition, time sequence prediction, multi-target optimization, container snapshot migration and the like, the resource utilization efficiency is effectively improved, the cost is reduced, and the service continuity is guaranteed.

Description

Technical Field

[0001] This application relates to the field of cloud phone technology, and more specifically, to a cloud phone resource management method and related equipment. Background Technology

[0002] With the continuous development of cloud computing and mobile internet technologies, cloud phones, as representatives of virtualized mobile terminals, are widely used in various scenarios such as mobile application testing, content distribution, remote work, and game cloudification. Compared with traditional physical devices, cloud phones have advantages such as high-concurrency deployment, centralized management, and rapid elastic expansion, becoming an important support for promoting terminal virtualization and computing power sharing. However, the resource management process behind cloud phones involves the real-time allocation and dynamic scheduling of computing resources, storage resources, and network resources. Especially when facing operating environments with multiple users, multiple instances, and drastic load fluctuations, how to achieve efficient, intelligent, and economical resource scheduling has become a key issue that urgently needs to be addressed.

[0003] In existing technologies, cloud phone resource management often relies on scheduling strategies based on static rules or fixed thresholds. For example, scaling up is triggered when CPU utilization exceeds a certain set value, and scaling down is executed when it falls below the threshold. While these strategies are low-cost to implement, they suffer from slow scheduling response and insufficient resource prediction capabilities, making them unsuitable for the refined resource control needs of sudden business surges or variable scenarios. In other words, related technologies suffer from low scheduling accuracy, slow scheduling response, and weak cross-cloud optimization capabilities in cloud phone resource management. Summary of the Invention

[0004] The summary section of this application introduces a series of simplified concepts, which will be further explained in detail in the detailed description section. The summary section of this application is not intended to limit the key features and essential technical features of the claimed technical solution, nor is it intended to determine the scope of protection of the claimed technical solution.

[0005] The cloud phone resource management method and related equipment provided in this application can construct a dynamic management process for cloud phone resources with high prediction accuracy, intelligent scheduling capabilities and multi-cloud elastic migration capabilities through technologies such as lightweight data collection, time series prediction, multi-objective optimization and container snapshot migration, effectively improving resource utilization efficiency, reducing costs and ensuring service continuity.

[0006] In a first aspect, this application provides a cloud phone resource management method, comprising: acquiring performance indicator data collected by a lightweight agent deployed on the virtualization layer; performing rolling prediction using a time-series prediction model based on the performance indicator data to obtain resource demand prediction results; solving a target resource allocation scheme using a multi-objective optimization model based on the resource demand prediction results, preset cost constraints, and preset service level constraints; and performing cross-cloud platform dynamic resource allocation operations using a multi-cloud adapter according to the target resource allocation scheme, wherein the dynamic resource allocation operations include resource migration based on container state snapshots.

[0007] In some implementations, the step of using a time-series forecasting model to perform rolling forecasting based on the performance index data to obtain resource demand forecasting results includes: standardizing the performance index data to obtain standardized index data; extracting periodic feature factors based on the standardized index data using Fourier transform; and inputting the periodic feature factors and the standardized index data into the time-series forecasting model to perform rolling forecasting to obtain the resource demand forecasting results.

[0008] In some implementations, the step of obtaining the target resource allocation scheme by solving a multi-objective optimization model based on the resource demand forecast results, preset cost constraints, and preset service level constraints includes: establishing a multi-objective optimization model based on the resource demand forecast results, with satisfying the preset cost constraints as the first optimization objective and satisfying the preset service level constraints as the second optimization objective; wherein the preset cost constraints include the difference between the instance on-demand price and the reserved price, cross-regional data transmission costs, and API call costs; and the preset service level constraints include preset maximum latency constraints and preset minimum availability constraints; and solving the multi-objective optimization model using an evolutionary algorithm to generate the target resource allocation scheme.

[0009] In some implementations, the step of performing dynamic resource allocation across cloud platforms using a multi-cloud adapter according to the target resource allocation scheme includes: when the cloud phone resources meet preset elastic scaling conditions, performing resource pre-expansion or resource pre-shrinkage according to a three-level resource pool strategy based on the elastic scaling instructions in the target resource allocation scheme, wherein the three-level resource pool includes a hot resource pool, a warm resource pool, and a cold resource pool; when the cloud phone resources meet preset cross-cloud migration conditions, triggering a cross-cloud migration process based on the cross-cloud migration instructions in the target resource allocation scheme.

[0010] In some implementations, the cloud phone resource management method further includes: obtaining a multi-dimensional cost element set based on the dynamic resource allocation operation and performing hierarchical accounting processing to generate a structured cost accounting report; calculating the total explicit cost based on preset pricing rules, wherein the explicit cost elements include instance computing fees, cross-regional data transmission fees, and API call fees; quantifying the implicit cost impact value through a latency loss model, wherein the implicit cost elements include resource switching latency loss value and service interruption risk weight value; integrating the total explicit cost value and the implicit cost impact value to generate a comprehensive cost index; inputting the comprehensive cost index and real-time resource allocation status data into a visualization engine, and dynamically rendering a cost distribution heatmap, a resource pool status topology map, and a service level agreement compliance dashboard through a management console.

[0011] In some implementations, acquiring performance metrics data collected by a lightweight agent deployed on the virtualization layer includes: collecting virtual CPU utilization, graphics processor rendering latency, and audio / video encoding queue depth metrics at preset time intervals through the lightweight agent; and determining the virtual CPU utilization, graphics processor rendering latency, and audio / video encoding queue depth metrics as the performance metrics data.

[0012] Secondly, this application also provides a cloud phone resource management device, comprising: a data acquisition unit for acquiring performance index data collected by a lightweight agent deployed on the virtualization layer; a resource prediction unit for performing rolling prediction based on the performance index data using a time-series prediction model to obtain resource demand prediction results; a scheme acquisition unit for obtaining a target resource allocation scheme by solving a multi-objective optimization model based on the resource demand prediction results, preset cost constraints, and preset service level constraints; and a resource allocation unit for performing cross-cloud platform dynamic resource allocation operations using a multi-cloud adapter according to the target resource allocation scheme, wherein the dynamic resource allocation operation includes resource migration based on container state snapshots.

[0013] Thirdly, this application also provides an electronic device, including: a memory and a processor, wherein the processor is configured to implement the steps of the cloud mobile phone resource management method described in the first aspect when executing a computer program stored in the memory.

[0014] Fourthly, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the cloud phone resource management method described in the first aspect.

[0015] Fifthly, this application also provides a computer program product, including a computer program or computer executable instructions, which, when executed by a processor, implement the steps of the cloud mobile phone resource management method provided in the embodiments of this application.

[0016] In summary, this application utilizes a lightweight agent deployed at the virtualization layer to collect key performance indicators during cloud phone operation at high frequency. Based on the collected data, a time-series prediction model is built for rolling predictions, effectively improving the accuracy and real-time performance of resource demand predictions. Compared to traditional static management methods that rely on fixed threshold triggers, this approach can proactively detect resource usage trends, providing a basis for subsequent scheduling decisions and preventing performance bottlenecks. By introducing a multi-objective optimization model and comprehensively considering multiple factors such as resource prediction results, cost constraints (e.g., instance pricing), and service level constraints (e.g., latency, availability), it can generate an optimal resource allocation scheme with balance and adaptability. It can dynamically balance resource investment and service quality according to business needs, achieving intelligent and differentiated resource allocation and improving the overall system efficiency. Furthermore, by employing a multi-cloud adapter to connect mainstream cloud platforms and combining container state snapshot technology, it enables dynamic migration and resource allocation of service instances across different cloud platforms. This effectively overcomes the limitations of single-cloud platform resource bottlenecks, improving system flexibility and fault tolerance. In cases of resource scarcity or sudden cost increases in the target cloud, it can automatically trigger cross-cloud deployment, ensuring service continuity and high business availability. In summary, the cloud phone resource management method provided in this application constructs a dynamic management process for cloud phone resources with high prediction accuracy, intelligent scheduling capabilities, and multi-cloud elastic migration capabilities through technologies such as lightweight data collection, time series prediction, multi-objective optimization, and container snapshot migration. This effectively improves resource utilization efficiency, reduces costs, and ensures service continuity. Attached Figure Description

[0017] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit this specification. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0018] Figure 1 A flowchart illustrating a cloud phone resource management method provided in an embodiment of this application;

[0019] Figure 2 This is a schematic diagram of the composition structure of a cloud mobile phone resource management device provided in an embodiment of this application;

[0020] Figure 3 This is a schematic diagram of the composition structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0021] The terms used in the specification, claims, and drawings of this application, such as "first," "second," "third," "fourth," etc. (if any), are used to distinguish similar objects and not to describe a specific order or sequence. Therefore, it is to be understood that these terms can be used interchangeably where appropriate, allowing the described embodiments to be used in different orders, unless specifically required by the illustrations or description. Furthermore, the terms "is" and "has," and any variations thereof, are intended to cover, non-exclusively, all possible constituent elements. For example, a process, method, system, product, or apparatus comprising several steps or units is not necessarily limited to the steps or units explicitly listed, but may also include other steps or units not explicitly listed, or steps or units inherent to the process, method, product, or apparatus.

[0022] In this application, a "module" or "unit" refers to a computer program or part of a computer program that has a specific function and works in conjunction with other related parts to achieve a predetermined goal. These modules or units can be implemented by software, hardware (e.g., processing circuitry or memory), or a combination of both. One or more processors or memories can implement one or more modules or units. Furthermore, each module or unit can also be part of a larger module or unit.

[0023] The technical solutions of this application will be described in detail below with reference to the accompanying drawings of the embodiments. It should be noted that the described embodiments are only a part of this application, and not all embodiments. In the following description, the "some embodiments" mentioned are only a subset of all possible embodiments, which may be the same or different subsets, and different embodiments can be combined with each other without conflict.

[0024] Figure 1 This is a flowchart illustrating a cloud phone resource management method provided in an embodiment of this application. For example, see [link to example]. Figure 1 The cloud phone resource management method provided in this application embodiment may include the following steps 101 to 104:

[0025] Step 101: Obtain performance metric data collected by the lightweight agent deployed in the virtualization layer;

[0026] In some examples, a lightweight agent is deployed on the Quick Emulator (QEMU) virtualization layer. This agent establishes a real-time communication channel with the virtualization layer kernel module to continuously collect multiple performance metrics at intervals of no more than five seconds. The collected metrics may include virtual central processing unit (vCPU) utilization, reflecting the real-time load of virtual computing resources; graphics processing unit (GPU) rendering latency, used to measure the response efficiency of graphics processing tasks; and audio / video encoding queue depth, reflecting the queuing status of audio / video data processing. During data collection, the agent performs initial cleaning of the raw metrics, removing outliers caused by momentary network jitter, and uses exponential smoothing denoising to process the continuously collected metric sequences. By dynamically adjusting the smoothing coefficient (ranging from 0 to 1), the timeliness and stability of the data are balanced. The processed performance metric data is then sent to the resource status database via an encrypted transmission channel, providing standardized input data for subsequent predictive analysis steps and ensuring the accuracy of subsequent model training and prediction results.

[0027] By implementing step 101, a lightweight agent is deployed at the virtualization layer, enabling high-frequency, low-overhead collection of key resource indicators such as vCPU utilization, GPU rendering latency, and audio / video queue depth. Compared to traditional monitoring systems, this approach is closer to the underlying resources, providing more accurate and real-time reflection of the cloud phone instance's operating status, offering refined data support for subsequent resource prediction and scheduling, and reducing additional system load.

[0028] Step 102: Based on performance index data, use a time series forecasting model to perform rolling forecasts to obtain resource demand forecast results;

[0029] In some examples, the performance metrics data collected in step 101 can first be standardized. By eliminating the differences in units between different metrics, metrics such as virtual CPU utilization, graphics processor rendering latency, and audio / video encoding queue depth are uniformly mapped to a numerical range of 0 to 1, ensuring data consistency and comparability. The standardized data is then input into a time-series prediction model. The model's input features include not only the real-time collected metric sequence but also periodic features extracted through Fourier transform, thereby capturing the regular patterns of resource usage changes over hourly cycles. Simultaneously, it incorporates the rate of change in resource utilization at adjacent time points, i.e., sudden load gradients, to identify short-term load fluctuations. The time-series prediction model employs a rolling prediction mechanism, setting a default sliding time window of thirty minutes. This means that the prediction window is updated at regular intervals based on the latest metric data, regenerating the resource demand prediction results for the next thirty minutes. To ensure the model's prediction accuracy, it is updated using weekly incremental training. During each incremental training, 20% of the data is used for A / B testing. The model's performance is verified and parameters are fine-tuned by comparing the test results. The predicted resource demand results will be stored in the resource status database in real time, providing accurate resource demand prediction basis for subsequent optimization decision-making modules, and ensuring that the resource allocation plan can adapt to potential load changes in advance.

[0030] By implementing step 102, a time-series prediction model, such as ARIMA, LSTM, or Prophet, is constructed, and a sliding time window is applied for rolling prediction. This can effectively capture the trend and cyclical changes in resource usage, especially the ability to detect sudden traffic or abnormal loads. Compared with traditional methods based on static thresholds or historical averages, the prediction is more dynamic and adaptive, and can detect load growth trends in advance, thus making resource allocation more forward-looking and proactive.

[0031] Step 103: Based on the resource demand forecast results, preset cost constraints, and preset service level constraints, the target resource allocation scheme is obtained by solving the problem through a multi-objective optimization model.

[0032] In some examples, based on the resource demand forecast results obtained in step 102, and combined with preset cost and service level constraints, a multi-objective optimization model is constructed with the goal of minimizing total cost and maximizing service availability. This model comprehensively considers explicit cost factors such as the difference between the on-demand price and the reserved price of instances, cross-regional data transfer costs, and application programming interface (API) call costs, while also incorporating implicit cost factors such as resource switching latency loss and service interruption risk weights, forming a multi-dimensional cost constraint system. Regarding service level constraints, the model sets maximum latency constraints and minimum availability constraints, corresponding to the upper limit requirement for cloud phone service response time and the lower limit guarantee for system uptime, respectively. During the solution process, a non-dominated sorting genetic algorithm (NSGA-III) based on an elite retention strategy is used to solve the multi-objective optimization model. By simulating a biological evolution process, a Pareto optimal solution set is generated, with each solution representing the optimal trade-off between cost and service level. To balance optimization accuracy and computational efficiency, the algorithm performs global optimization every hour to consider the allocation of all resources and local adjustments every five minutes to respond to real-time load changes. Finally, it selects the target resource allocation scheme that meets the preset preferences from the Pareto solution set. This scheme includes elastic scaling instructions and cross-cloud migration instructions, which are used to guide subsequent dynamic resource allocation operations.

[0033] By implementing step 103, the predicted resource demand results are used as input, and combined with cost constraints and service level requirements in actual business scenarios, a multi-objective optimization model is constructed. This model can generate Pareto optimal resource allocation schemes that meet various constraints, achieving the dual objectives of maximizing resource utilization and controlling costs while ensuring service quality. Thus, it possesses intelligence, balance, and business adaptability.

[0034] Step 104: According to the aforementioned target resource allocation scheme, use the multi-cloud adapter to perform cross-cloud platform dynamic resource allocation operations, wherein the dynamic resource allocation operations may include resource migration based on container state snapshots;

[0035] In some examples, based on the target resource allocation scheme generated in step 103, a multi-cloud adapter integrating Terraform and Crossplane protocol conversion functions is used to perform dynamic resource allocation operations across cloud platforms. This adapter is compatible with the application programming interfaces of mainstream cloud service providers, enabling unified conversion and execution of instructions between heterogeneous cloud platforms. Dynamic resource allocation operations encompass two core scenarios: elastic scaling and cross-cloud migration. In the elastic scaling scenario, based on the elastic scaling instructions in the target solution, resources are pre-expanded or pre-shrunk using a three-tiered resource pool strategy (hot resource pool, warm resource pool, and cold resource pool). The hot resource pool is used to respond to sudden loads in real time, the warm resource pool maintains minute-level scheduling capabilities, and the cold resource pool stores offline backup resources. Dynamic transfer between resource pools is achieved through a preset probability threshold triggering mechanism. In the cross-cloud migration scenario, when the price fluctuation of the target cloud platform is detected to exceed a set threshold, the system automatically triggers a migration process based on container state snapshots. Container state snapshots are generated using Checkpoint / Restore in Userspace (CRIU) technology. In an overlay network environment built on a Virtual Extensible Local Area Network (VXLAN), point-to-point distribution of images can be achieved using the Dragonfly tool. Domain Name System records are updated synchronously to complete seamless traffic switching. The interruption time of the cloud phone service during the entire migration process is controlled within three seconds, ensuring the continuity of the user experience.

[0036] By implementing step 104, the multi-cloud adapter component is integrated, enabling seamless resource invocation and automatic scheduling across multiple cloud service platforms. In scenarios with resource scarcity, sudden load surges, or high costs, container state snapshots can facilitate rapid and uninterrupted cross-cloud resource migration, enhancing the system's elastic scheduling capabilities, business continuity, and fault tolerance. This significantly reduces the need for manual intervention and effectively addresses the complex resource management challenges in multi-cloud environments.

[0037] In summary, this application's embodiments utilize a lightweight agent deployed in the virtualization layer to achieve high-frequency collection of key performance indicators during cloud phone operation. Based on the collected data, a time-series prediction model is constructed for rolling predictions, effectively improving the accuracy and real-time performance of resource demand predictions. Compared to traditional static management methods relying on fixed threshold triggers, this approach can proactively detect resource usage trends, providing a basis for subsequent scheduling decisions and preventing performance bottlenecks. By introducing a multi-objective optimization model and comprehensively considering multiple factors such as resource prediction results, cost constraints (e.g., instance pricing), and service level constraints (e.g., latency, availability), an optimal resource allocation scheme can be generated, exhibiting balance and adaptability. It can dynamically balance resource investment and service quality according to business needs, achieving intelligent and differentiated resource allocation and improving the overall system operating efficiency. Employing a multi-cloud adapter to connect mainstream cloud platforms and combining container state snapshot technology, it enables dynamic migration and resource allocation of service instances across different cloud platforms. This effectively overcomes the limitations of single-cloud platform resource bottlenecks, enhancing system flexibility and fault tolerance. In cases of target cloud resource scarcity or sudden cost increases, cross-cloud deployment can be automatically triggered, ensuring service continuity and high business availability. In summary, the cloud phone resource management method provided in this application constructs a dynamic management process for cloud phone resources with high prediction accuracy, intelligent scheduling capabilities, and multi-cloud elastic migration capabilities through technologies such as lightweight data collection, time-series prediction, multi-objective optimization, and container snapshot migration. This effectively improves resource utilization efficiency, reduces costs, and ensures service continuity.

[0038] In some embodiments, step 102 may include: standardizing the performance index data to obtain standardized index data; extracting periodic feature factors based on the standardized index data using Fourier transform; and inputting the periodic feature factors and standardized index data into a time series forecasting model for rolling forecasting to obtain resource demand forecasting results.

[0039] In some examples, the collected performance metrics data can first be standardized. By using a min-max scaling method, metrics with different dimensions, such as virtual CPU utilization, graphics processor rendering latency, and audio / video encoding queue depth, can be uniformly mapped to a numerical range of 0 to 1, eliminating the weight imbalance caused by differences in units. The processed data is the standardized metric data. Subsequently, based on this standardized metric data, Fourier transform can be used to convert the time-domain signal to the frequency domain. By analyzing the peak frequencies in the signal spectrum, periodic patterns of resource usage can be identified, and a periodic feature factor on an hourly basis can be extracted. This factor can quantify the fluctuation patterns of resource demand at different times of the day (e.g., morning peak, midday trough, evening peak). After feature extraction, the periodic feature factor and the standardized metric data are used as input parameters to a time-series prediction model. The model can employ a Long Short-Term Memory (LSTM) network. Using Short-Term Memory (LSTM) as the core architecture, combined with a sliding time window of 30 minutes by default, the system performs rolling predictions. Every five minutes, the input sequence is updated based on the latest standardized indicator data and periodic feature factors to regenerate the resource demand prediction results for the next 30 minutes. This ensures that the prediction results can adapt to the dynamic changes in resource usage patterns in real time, providing accurate predictive basis for subsequent resource allocation decisions.

[0040] Through the implementation of the above embodiments, the collected performance index data is standardized to ensure data consistency and comparability across different index dimensions. At the same time, Fourier transform is used to extract periodic feature factors, thereby enhancing the model's ability to perceive the periodicity of resource usage. These features, as inputs, work together to form a time-series prediction model, making the prediction results more accurate and robust. This model is particularly suitable for dealing with the dynamic changes in cloud phone resources during specific time periods (such as high concurrency at night or low load during working hours), thereby improving the foresight and adaptability of the overall scheduling process.

[0041] In some embodiments, step 103 may include: establishing a multi-objective optimization model based on the aforementioned resource demand forecast results, with satisfying a preset cost constraint as the first optimization objective and satisfying a preset service level constraint as the second optimization objective. The preset cost constraint may include the difference between the instance on-demand price and the reserved price, cross-regional data transfer costs, and API call costs. The preset service level constraint may include a preset maximum latency constraint and a preset minimum availability constraint. An evolutionary algorithm is used to solve the multi-objective optimization model to generate a target resource allocation scheme.

[0042] In some examples, during the specific implementation of step 103, a multi-objective optimization model can be constructed based on the resource demand forecast results generated in step 102, with satisfying preset cost constraints as the primary optimization objective and satisfying preset service level constraints as the secondary optimization objective. The preset cost constraints cover the difference between the on-demand purchase price and the reserved purchase price of cloud service instances, the costs incurred during cross-regional data transmission, and the fees incurred during application programming interface (API) calls. These cost elements collectively constitute the quantitative constraint standard for resource allocation costs in the model. The preset service level constraints include preset maximum latency constraints and preset minimum availability constraints. The former corresponds to the maximum allowed time limit from request to response completion for the cloud phone service, while the latter corresponds to the minimum percentage requirement for the system to maintain normal operation within a specified time. Both together ensure the quality and stability of the cloud phone service. During the model solving phase, an evolutionary algorithm based on an elite retention strategy is used to iteratively solve the multi-objective optimization model. This algorithm simulates the selection, crossover, and mutation operations in the biological evolution process, and selects non-dominated solutions in each generation of the population. After multiple rounds of iteration, a Pareto optimal solution set containing various trade-off schemes is generated. From this set, the solution that simultaneously satisfies the dual constraints of cost and service level is selected as the final target resource allocation scheme. This scheme covers specific scheduling parameters such as the allocation ratio of resources across different cloud platforms, instance type selection, and elastic scaling threshold.

[0043] Through the implementation of the above embodiments, during the resource optimization phase, key cost elements in actual business operations (such as instance type price differences, cross-regional transmission costs, and API costs) and key service level indicators (such as maximum latency and minimum availability) are fully integrated to construct a realistic and complete multi-objective optimization model, which is then solved globally using an evolutionary algorithm. This modeling and solving method can generate multiple Pareto optimal resource allocation schemes that meet business needs, allowing the system to select and execute them, thus achieving an optimal balance between resource costs and service experience.

[0044] In some embodiments, step 104 may include: when cloud phone resources meet preset elastic scaling conditions, performing resource pre-expansion or resource pre-shrinking according to the three-level resource pool strategy based on the elastic scaling instructions in the aforementioned target resource allocation scheme, wherein the three-level resource pool may include a hot resource pool, a warm resource pool, and a cold resource pool; when cloud phone resources meet preset cross-cloud migration conditions, triggering the cross-cloud migration process based on the cross-cloud migration instructions in the target resource allocation scheme.

[0045] In some examples, during the specific implementation of step 104, when the real-time monitoring data of cloud phone resources meets the preset elastic scaling conditions, resource pre-expansion or pre-shrinkage operations can be performed according to the elastic scaling instructions included in the target resource allocation scheme, following a three-level resource pool strategy consisting of a hot resource pool, a warm resource pool, and a cold resource pool. The hot resource pool stores active resources that can respond in real time, enabling instance scheduling to cope with sudden loads within seconds. The warm resource pool stores resources in standby mode, which can be started and put into use within minutes. The cold resource pool is used to store offline resources, activated only through a wake-up mechanism when long-term resource demand increases. When the virtual central processing unit utilization exceeds 85% for three consecutive minutes or the graphics processor rendering latency exceeds a preset threshold, a pre-expansion operation is triggered, scheduling resources from the warm resource pool to supplement the hot resource pool. If the warm resource pool is insufficient, some resources from the cold resource pool are woken up. When the resource utilization is below 30% for ten consecutive minutes, a pre-shrinkage operation is performed, migrating some hot resource pool resources to the warm or cold resource pool to reduce idle costs. When the price fluctuation of the cloud platform hosting the cloud phone resources exceeds a preset threshold, or its performance indicators consistently fail to meet service level constraints, the preset cross-cloud migration conditions are met. The system will then initiate the cross-cloud migration process according to the cross-cloud migration instructions in the target resource allocation plan. This process first establishes an overlay network between the source and target cloud platforms using a Virtual Extensible Local Area Network (VXLAN) to ensure the privacy and continuity of data transmission. Then, it utilizes the Dragonfly tool for point-to-point distribution of container images, improving image transmission efficiency. Finally, it completes a seamless switch of traffic from the source instance to the target instance by updating Domain Name System records. Throughout the process, Checkpoint / Restore in Userspace (CRIU) technology is used to generate container state snapshots, ensuring consistency of service status before and after the migration.

[0046] The implementation of the above embodiments provides two core operation paths for dynamic resource scheduling: one is elastic scaling triggered by changes in resource load, which uses a three-level resource pool (hot / warm / cold) strategy to achieve on-demand pre-expansion and scaling down, thereby improving the flexibility of resource allocation; the other is cross-cloud migration operations triggered by cost or resource bottlenecks, which improves the resource collaboration capabilities between platforms. This design can meet users' resource elasticity needs while automatically avoiding high-cost or low-availability resource areas, thereby ensuring the stability, cost efficiency, and multi-cloud elasticity of cloud phone services.

[0047] In some embodiments, the aforementioned cloud phone resource management method may further include: obtaining a multi-dimensional cost element set based on the aforementioned dynamic resource allocation operation and performing hierarchical accounting processing to generate a structured cost accounting report; calculating the total explicit cost based on preset pricing rules, wherein explicit cost elements may include instance computing fees, cross-regional data transmission fees, and API call fees; quantifying the impact value of implicit costs through a latency loss model, wherein the aforementioned implicit cost elements may include resource switching latency loss values ​​and service interruption risk weight values; integrating the total explicit cost and the impact value of implicit costs to generate a comprehensive cost index; inputting the comprehensive cost index and real-time resource allocation status data into a visualization engine, and dynamically rendering a cost distribution heatmap, a resource pool status topology map, and a service level agreement compliance dashboard through a management console.

[0048] In some examples, based on the data generated by the aforementioned dynamic resource allocation operations, a multi-dimensional cost element set containing both explicit and implicit elements can be automatically obtained. This set is then processed hierarchically according to cost attributes and impact scope, ultimately generating a structured cost accounting report covering cost composition, fluctuation trends, and optimization suggestions. During the accounting process, the total explicit cost is first calculated based on preset pricing rules (which integrate publicly available pricing standards and agreement discount terms from various cloud platforms). Explicit cost elements include the computing fees for cloud server instances (charged based on the number of virtual CPU cores and runtime), cross-regional data transfer fees (a fixed rate per GB of data transfer), and Application Programming Interface (API) call fees (priced in tiers based on the number of calls). Simultaneously, a self-developed latency loss model quantifies the impact of implicit costs. This model converts latency loss during resource switching into equivalent costs (such as a service experience degradation coefficient per millisecond of latency) and assigns corresponding weight values ​​to service interruption risks based on historical fault data. Both of these factors together constitute the implicit cost elements. After separately calculating explicit and implicit costs, the system uses a weighted fusion algorithm to integrate the total explicit cost and the impact of implicit costs into a single comprehensive cost indicator. This indicator reflects both actual capital expenditure and the economic impact of service quality degradation. Finally, the comprehensive cost indicator and real-time resource allocation status data (including the number of instances in each resource pool, load rate, and migration progress) are simultaneously input into the visualization engine. The engine dynamically renders a cost distribution heatmap (using color gradients to show the cost proportion of different cloud platforms), a resource pool status topology map (intuitively presenting the resource flow relationship of the three-tier resource pool), and a Service Level Agreement (SLA) compliance dashboard (real-time display of the deviation of indicators such as latency and availability from constraint thresholds), providing administrators with a comprehensive view of cost and resource status monitoring.

[0049] Through the implementation of the above embodiments, after resource scheduling is executed, explicit costs such as instance computing fees, cross-region transmission fees, and API calls are collected. At the same time, implicit costs such as switching latency and service interruption risks are combined to construct a complete multi-dimensional cost accounting mechanism, and generate structured comprehensive cost indicators. These indicators are not only visualized on the management console, but also support display in the form of heat maps, topology maps, and service level agreement compliance dashboards, providing strong support for operation and maintenance and decision-making. This greatly improves the observability and transparency of resource costs and service quality of the cloud phone platform, and enhances its refined management capabilities.

[0050] In some embodiments, step 101 may include: collecting virtual CPU utilization, graphics processor rendering latency, and audio / video encoding queue depth metrics at preset time intervals via a lightweight agent; and determining the virtual CPU utilization, graphics processor rendering latency, and audio / video encoding queue depth metrics as performance indicator data.

[0051] In some examples, during the specific implementation of step 101, a lightweight agent deployed in the QuickEmulator (QEMU) virtualization layer establishes a communication channel with the performance monitoring interface of the kernel space to continuously collect three key performance indicators at preset time intervals of no more than five seconds. The virtual central processing unit (vCPU) utilization is calculated by reading CPU time slice statistics from the / proc / stat file and combining them with clock interrupt frequency; this indicator reflects the real-time load pressure on virtual computing resources. The graphics processing unit (GPU) rendering latency is measured by inserting a timestamp probe into the graphics rendering pipeline to measure the time interval from the submission of the rendering command to the completion of the frame buffer update; this indicator directly affects the smoothness of the cloud phone interface. The audio / video encoding queue depth is obtained by querying the task queue length of the FFmpeg encoding engine, reflecting the queuing status of audio / video data processing. During the collection process, the agent uses non-blocking I / O technology to acquire the three indicators in parallel, ensuring data time synchronization. Simultaneously, it performs sliding window filtering on the raw data to eliminate outliers caused by instantaneous network jitter or system scheduling fluctuations. After data collection is completed, the system encapsulates the three preprocessed metrics—virtual CPU utilization, graphics processor rendering latency, and audio / video encoding queue depth—into a unified performance metric data object. This object is then transmitted to the resource status database via a Secure Sockets Layer (SSL) encrypted channel, providing foundational data support for subsequent predictive analysis.

[0052] Through the implementation of the above embodiments, the key content and sampling method of collecting performance indicators have been clarified: the lightweight agent obtains key indicators such as virtual CPU utilization, GPU rendering latency, and audio and video encoding queue depth at fixed time intervals. These indicators are directly related to the real-time computing performance, graphics responsiveness, and audio and video smoothness of cloud phones. As the basic data for performance prediction and resource scheduling, they not only have real-time performance and representativeness, but also accurately reflect the changing trends of user experience, providing a stable and reliable data source for subsequent predictive analysis and resource management.

[0053] Furthermore, as an implementation of the aforementioned method embodiments, this application also provides a cloud phone resource management device for implementing the aforementioned method embodiments. This device embodiment corresponds to the aforementioned method embodiments. For ease of reading, this cloud phone resource management device embodiment will not repeat the details of the aforementioned method embodiments one by one, but it should be understood that the device in this application embodiment can correspondingly implement all the contents of the aforementioned method embodiments. For example... Figure 2 As shown, the cloud phone resource management device 20 includes: a data acquisition unit 201, a resource prediction unit 202, a scheme acquisition unit 203, and a resource allocation unit 204. The data acquisition unit 201 acquires performance indicator data collected by a lightweight agent deployed in the virtualization layer. The resource prediction unit 202 performs rolling predictions based on the aforementioned performance indicator data using a time-series prediction model to obtain resource demand prediction results. The scheme acquisition unit 203, based on the resource demand prediction results, preset cost constraints, and preset service level constraints, performs solution processing through a multi-objective optimization model to obtain a target resource allocation scheme. The resource allocation unit 204 performs dynamic resource allocation operations across cloud platforms using a multi-cloud adapter according to the target resource allocation scheme. The dynamic resource allocation operations may include resource migration based on container state snapshots.

[0054] This application also provides a computer-readable storage medium storing computer-executable instructions or computer programs, which, when executed by a processor, will cause the processor to perform any step of the cloud phone resource management method provided in this application.

[0055] In some embodiments, the computer-readable storage medium may be a random access memory (RAM), a read-only memory (ROM), flash memory, a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM); or it may be a variety of devices that include one or any combination of the above-mentioned memories.

[0056] In some embodiments, computer-executable instructions may take the form of programs, software, software modules, scripts, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

[0057] In some embodiments, computer-executable instructions may, but do not necessarily, correspond to files in a file system, and may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a HyperText Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple co-located files (e.g., files that store one or more modules, subroutines, or code sections).

[0058] In some embodiments, computer-executable instructions may be deployed to execute on an electronic device, or on multiple electronic devices located at one location, or on multiple electronic devices distributed across multiple locations and interconnected via a communication network.

[0059] like Figure 3 As shown, this application also provides an electronic device 30, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor. When the processor 320 executes the computer program 311, it implements any step of the cloud mobile phone resource management method described above.

[0060] This application also provides a computer program product comprising a computer program or computer-executable instructions stored in a computer-readable storage medium. The processor of an electronic device reads the computer program or computer-executable instructions from the computer-readable storage medium and executes the computer program or computer-executable instructions, causing the electronic device to perform any step of the cloud phone resource management method described above.

[0061] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A cloud phone resource management method, characterized in that, include: Obtain performance metrics data collected by a lightweight agent deployed in the virtualization layer; Based on the performance index data, rolling forecasts are performed using a time-series forecasting model to obtain resource demand forecast results. Based on the resource demand forecast results, preset cost constraints, and preset service level constraints, the target resource allocation scheme is obtained by solving the problem through a multi-objective optimization model. According to the target resource allocation scheme, a multi-cloud adapter is used to perform dynamic resource allocation operations across cloud platforms, wherein the dynamic resource allocation operations include resource migration based on container state snapshots.

2. The cloud phone resource management method according to claim 1, characterized in that, The step of using a time-series forecasting model to perform rolling forecasts based on the performance index data to obtain resource demand forecast results includes: The performance index data is standardized to obtain standardized index data; Based on the standardized index data, periodic feature factors are extracted using Fourier transform; The periodic characteristic factors and the standardized index data are input into the time series prediction model for rolling prediction to obtain the resource demand prediction results.

3. The cloud phone resource management method according to claim 1, characterized in that, The process of obtaining the target resource allocation scheme by solving a multi-objective optimization model based on the resource demand forecast results, preset cost constraints, and preset service level constraints includes: Based on the resource demand forecast results, a multi-objective optimization model is established with the preset cost constraint as the first optimization objective and the preset service level constraint as the second optimization objective. The preset cost constraint includes the difference between the instance on-demand price and the reserved price, the cross-regional data transmission cost, and the API call cost. The preset service level constraint includes the preset maximum latency constraint and the preset minimum availability constraint. An evolutionary algorithm is used to solve the multi-objective optimization model to generate the target resource allocation scheme.

4. The cloud phone resource management method according to any one of claims 1, characterized in that, The step of performing cross-cloud platform dynamic resource allocation operations using a multi-cloud adapter according to the target resource allocation scheme includes: When the cloud phone resources meet the preset elastic scaling conditions, based on the elastic scaling instructions in the target resource allocation scheme, the resources are pre-expanded or pre-shrunk according to the three-level resource pool strategy. The three-level resource pool includes a hot resource pool, a warm resource pool, and a cold resource pool. When the cloud phone resources meet the preset cross-cloud migration conditions, the cross-cloud migration process is triggered based on the cross-cloud migration instructions in the target resource allocation scheme.

5. The cloud phone resource management method according to claim 1, characterized in that, The cloud phone resource management method also includes: Based on the aforementioned dynamic resource allocation operation, a multi-dimensional cost element set is obtained and hierarchical accounting processing is performed to generate a structured cost accounting report. The total explicit cost is calculated based on preset pricing rules, wherein the explicit cost elements include instance computing fees, cross-regional data transfer fees, and API call fees; The impact of implicit costs is quantified through a latency loss model, wherein the implicit cost elements include resource switching latency loss and service interruption risk weight. By combining the total explicit cost value and the impact value of implicit cost, a comprehensive cost index is generated; The comprehensive cost indicators and real-time resource allocation status data are input into the visualization engine, and the cost distribution heatmap, resource pool status topology map, and service level agreement compliance dashboard are dynamically rendered through the management console.

6. The cloud phone resource management method according to any one of claims 1 to 5, characterized in that, The acquisition of performance metric data collected by the lightweight agent deployed in the virtualization layer includes: The lightweight agent collects virtual CPU utilization, graphics processor rendering latency, and audio / video encoding queue depth metrics at preset time intervals. The virtual CPU utilization, the graphics processor rendering latency, and the audio / video encoding queue depth are determined as the performance indicator data.

7. A cloud phone resource management device, characterized in that, include: The data acquisition unit is used to acquire performance metric data collected by a lightweight agent deployed in the virtualization layer; The resource prediction unit is used to perform rolling predictions based on the performance index data using a time-series prediction model to obtain resource demand prediction results. The scheme acquisition unit is used to obtain the target resource allocation scheme by solving the resource demand prediction results, preset cost constraints and preset service level constraints through a multi-objective optimization model. The resource allocation unit is used to perform cross-cloud platform dynamic resource allocation operations using a multi-cloud adapter according to the target resource allocation scheme, wherein the dynamic resource allocation operation includes resource migration based on container state snapshots.

8. An electronic device, comprising: The memory and processor are characterized in that the processor, when executing a computer program stored in the memory, implements the steps of the cloud phone resource management method as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the cloud mobile phone resource management method as described in any one of claims 1 to 6.

10. A computer program product comprising a computer program or computer-executable instructions, characterized in that, When the computer program or computer-executable instructions are executed by the processor, the steps of the cloud mobile phone resource management method as described in any one of claims 1 to 6 are implemented.

Images