Cloud computing management system and method

Abstract

The invention relates to the technical field of information transmission, in particular to a cloud computing management system and method. The method comprises the following steps: acquiring user activity data, and predicting a traffic demand according to the user activity data to obtain live broadcast predicted traffic data; obtaining cloud platform resource allocation data, and performing elastic expansion to obtain a resource allocation configuration table; performing content distribution according to the resource allocation configuration table and the live broadcast prediction flow data to obtain live broadcast content distribution data; performing live broadcast content multi-preference layer caching on the live broadcast content distribution data to obtain a cache storage optimization table; acquiring real-time live broadcast network bandwidth data, and performing video coding according to the real-time live broadcast network bandwidth data to obtain a video quality strategy table; and obtaining real-time resource allocation data of the cloud platform, and optimizing cloud resource allocation according to the real-time resource allocation data of the cloud platform to obtain optimized cloud resource allocation data. The cloud computing resource use efficiency and the communication efficiency can be improved.

Description

Technical Field

[0001] The present invention relates to the field of information transmission technology, and in particular to a cloud computing management system and method. Background Art

[0002] E-commerce live streaming platforms combine the advantages of traditional e-commerce and social media, allowing anchors to display products and interact with viewers through live video to promote product sales. E-commerce live streaming platforms are faced with growing user traffic, massive video streaming data, and changing business needs. In order to ensure the stable operation of the platform, improve user experience, and effectively utilize computing resources, e-commerce live streaming platforms usually need to rely on large-scale cloud computing platforms for computing resource management. Cloud computing can provide e-commerce live streaming platforms with elastic computing, storage, and content distribution services, enabling them to cope with the challenges of traffic fluctuations and data processing requirements during live streaming. Although cloud computing provides strong technical support for e-commerce live streaming platforms, existing cloud computing management methods usually adopt a static resource allocation model, that is, the use of cloud resources is pre-determined. This method may be able to meet the needs of e-commerce live streaming platforms in the short term, but with the continuous changes in platform traffic, especially during large-scale promotional activities or special live events, the demand for resources often fluctuates violently. Due to the lag in resource allocation, the platform faces insufficient resources during peak periods, resulting in live streaming freezes, reduced user experience, and even system downtime. During trough periods, excess resources will lead to resource waste and increase operating costs. Summary of the invention

[0003] Based on this, it is necessary for the present invention to provide a cloud computing management system and method to solve at least one of the above technical problems.

[0004] To achieve the above object, a cloud computing management method includes the following steps:

[0005] Step S1: Obtain user activity data, and predict the traffic demand during the user's active time period based on the user activity data to obtain live broadcast predicted traffic data;

[0006] Step S2: Obtain cloud platform resource allocation data, and dynamically and elastically expand the cloud platform live broadcast resource demand based on the live broadcast predicted traffic data to obtain a resource allocation configuration table;

[0007] Step S3: Classify and distribute the content according to the user's live broadcast dependency according to the resource allocation configuration table and the live broadcast predicted traffic data to obtain live broadcast content distribution data; cache the live broadcast content distribution data in multiple preference layers to obtain a cache storage optimization table, and upload it to the cloud platform to perform the content caching task;

[0008] Step S4: obtaining real-time live network bandwidth data, and performing bit rate stream adaptive video encoding according to the real-time live network bandwidth data, obtaining a video quality strategy table, and uploading it to the cloud platform to execute the video encoding task;

[0009] Step S5: Obtain the real-time resource allocation data of the cloud platform, and optimize the cloud resource configuration according to the real-time resource allocation data of the cloud platform, obtain the optimized cloud resource configuration data, and upload it to the cloud platform to execute the resource configuration task.

[0010] The present invention can realize dynamic scheduling and optimization of cloud computing resources, greatly improving the resource utilization rate and user experience of the platform. First, the user activity data is obtained and predicted, and the flow demand of the user's active time period is accurately predicted, so that the platform can prepare for the flow in advance, avoiding resource tension or waste caused by burst flow. On the basis of the traditional static resource allocation method, the dynamic elastic expansion of resource allocation can effectively cope with the drastic fluctuations of live broadcast flow, avoid the jamming phenomenon during the peak period, and ensure the stability of the live broadcast quality. In the trough period, the waste of resources is reduced, the operating cost is optimized, and the economic benefits of the platform are ensured. Next, through the optimization of user live broadcast dependency content classification distribution and multi-preference layer cache, not only the resource utilization efficiency is improved, but also the user's viewing experience is improved. High-dependency users can obtain the required content with higher priority and reduce delay, while low-dependency users can also obtain better viewing quality under cache optimization, further improving the load capacity and content distribution efficiency of the platform. At the same time, the cache preheating strategy also ensures that the content can respond quickly during high-flow periods, reducing the delay and jamming of live broadcast. In terms of video encoding, through real-time network bandwidth data monitoring and bit rate stream adaptive adjustment, the platform can flexibly adjust the video quality according to the network status of the user's device to ensure the stability and fluency of the video stream. Whether it is a user with poor network or a user with sufficient bandwidth, they can obtain suitable viewing quality according to the actual situation, improve user satisfaction, and reduce user loss caused by video quality problems. Real-time resource allocation optimization and adjustment further improve the resource utilization efficiency of the platform. Through real-time monitoring and optimization of cloud platform resources, the platform can adjust resource configuration according to the current load, avoid resource waste, and ensure that cloud resources are optimally allocated under different loads. This not only improves the operating efficiency of the platform, but also reduces unnecessary cost expenditures, ensuring the stable operation of the platform during large-scale promotional activities or special live events. Overall, the present invention not only optimizes the resource management of the e-commerce live broadcast platform by realizing dynamic and intelligent resource scheduling, content distribution and video quality adjustment, but also greatly improves the user experience and the operating efficiency of the platform, and has significant technical and commercial value.

[0011] Optionally, step S1 specifically includes:

[0012] Step S11: acquiring user activity data, and performing data preprocessing on the user activity data to obtain user activity data to be analyzed;

[0013] Step S12: extracting user activity features from the user activity data to obtain user activity time data and user live broadcast interaction data;

[0014] Step S13: dividing the user activity time periodically according to the user activity time data to obtain the user activity time periodicity data;

[0015] Step S14: performing user activity pattern recognition based on the user activity time periodicity data and the user live broadcast interaction data to obtain user activity pattern data;

[0016] Step S15: construct a live broadcast traffic prediction model based on the user activity pattern data, and use the live broadcast traffic prediction model to predict the traffic during the user's active time period to obtain live broadcast predicted traffic data.

[0017] The present invention can effectively extract the user's activity time and live interactive behavior characteristics through preprocessing and feature extraction of user activity data, thereby providing clear and structured data for subsequent analysis. This process can reduce data noise and improve the accuracy of analysis. When the user activity time is further divided periodically, the user's active time period can be revealed, which provides a basis for accurately predicting user needs and traffic fluctuations. Based on these periodic data and live interactive features, user activity pattern recognition helps to deeply understand the user's behavioral laws, further improve user portraits, and effectively improve the accuracy of traffic prediction. By constructing a live traffic prediction model and combining the user's active time period for accurate prediction, not only can the platform's resource scheduling be optimized, but also traffic plans can be made in advance to avoid system load problems caused by a sharp increase or decrease in traffic, thereby improving the stability of platform services and user experience. In addition, this process can also help the platform accurately push content, improve user activity, optimize content distribution strategies, and enhance the platform's operational efficiency.

[0018] Optionally, step S14 is specifically:

[0019] Step S141: performing interaction type weight classification on user live broadcast interaction data to obtain live broadcast interaction type weight data;

[0020] Step S142: evaluating the user live broadcast participation degree of the user live broadcast interaction data according to the live broadcast interaction type weight data to obtain the user live broadcast participation degree data;

[0021] Step S143: clustering the user's live broadcast type preference according to the live broadcast participation data to obtain the user's live broadcast type preference data;

[0022] Step S144: performing user participation in live broadcast type time association based on the user activity time periodicity data and the user live broadcast type preference data to obtain live broadcast preference time data;

[0023] Step S145: identifying the user's live broadcast activity participation behavior pattern according to the live broadcast preference time data to obtain live broadcast participation behavior pattern data;

[0024] Step S146: assigning behavior pattern labels to the live broadcast participation behavior pattern data to obtain user activity pattern data.

[0025] The present invention can accurately reflect the level of activity of each user in different types of live broadcasts by weighting and evaluating the user live broadcast interaction data, thereby laying the foundation for accurate user portraits. Clustering these data with user preferences can effectively reveal the user's live broadcast preferences, help the platform push content that meets the user's interests, and improve user experience. Combined with the periodic data analysis of user activity time, we can deeply understand the user's active time period and further optimize the release time and promotion strategy of live broadcast content. On this basis, identifying the user's live broadcast participation behavior pattern can help the platform provide customized services at different active stages of the user, improve user stickiness and platform interaction rate. Based on the allocation of behavioral pattern labels, users can be accurately classified, thereby achieving more targeted operation strategies and improving the overall effect of live broadcast activities and platform activity.

[0026] Optionally, step S2 specifically includes:

[0027] Step S21: Obtain cloud platform resource allocation data, and perform data preprocessing on the cloud platform resource allocation data to obtain resource allocation data to be analyzed;

[0028] Step S22: estimating resource demand during the live broadcast activity according to the live broadcast predicted traffic data, and obtaining resource demand data during the live broadcast;

[0029] Step S23: determining the node transmission boundary conditions of the cloud platform according to the resource allocation data to be analyzed, and obtaining the node transmission boundary conditions;

[0030] Step S24: matching the data transmission resource demand with the resource demand data during the live broadcast based on the node transmission boundary conditions, and obtaining transmission computing resource matching data, transmission bandwidth resource matching data, and transmission storage resource matching data;

[0031] Step S25: setting resource automatic expansion strategy for the transmission computing resource matching data, the transmission bandwidth resource matching data and the transmission storage resource matching data to obtain a resource allocation configuration table.

[0032] By preprocessing the cloud platform resource allocation data, the present invention can clearly grasp the current resource usage and provide data support for subsequent analysis and optimization. Combining the live broadcast prediction traffic data to estimate the resource demand during the live broadcast helps to prepare resources in advance before the event starts to avoid affecting the quality of the live broadcast due to insufficient resources. After determining the node transmission boundary conditions, the resource allocation of each node can be accurately controlled to ensure efficiency and stability during the data transmission process. Matching the resource requirements during the live broadcast can reasonably allocate computing, bandwidth and storage resources, improve resource utilization and reduce resource waste. Finally, by setting the automatic expansion strategy, it can ensure that the platform can make timely adjustments when resource demand fluctuates, and realize dynamic optimization configuration of resources, thereby ensuring the smooth progress of the live broadcast activity and improving user experience and platform service quality.

[0033] Optionally, step S25 is specifically:

[0034] According to a preset resource division ratio, reserved resource division is performed on the transmission computing resource matching data, the transmission bandwidth resource matching data, and the transmission storage resource matching data to obtain reserved resource data, to-be-allocated computing resource data, to-be-allocated bandwidth resource data, and to-be-allocated storage resource data;

[0035] According to the resource demand data during the live broadcast, the live broadcast resource load threshold is set in steps to obtain a live broadcast resource extension step threshold set, wherein the live broadcast resource extension step threshold set includes a computing resource step extension threshold, a bandwidth resource step extension threshold, and a storage resource step extension threshold;

[0036] Based on the computing resource step expansion threshold, computing resource expansion rules are set for the computing resource data to be allocated, and a computing resource expansion strategy is obtained;

[0037] Based on the bandwidth resource step expansion threshold, bandwidth resource expansion rules are set for the bandwidth resource data to be allocated to obtain a bandwidth resource expansion strategy;

[0038] Based on the storage resource step expansion threshold, storage resource expansion rules are set for the storage resource data to be allocated to obtain a storage resource expansion strategy;

[0039] Set resource expansion dynamic elasticity factor according to node transmission boundary conditions;

[0040] According to the resource expansion dynamic elasticity factor, a reserved resource calling rule is set for the reserved resource data to obtain a reserved resource calling strategy;

[0041] Allocate expansion priorities for the reserved resource call strategy, the computing resource expansion strategy, the bandwidth resource expansion strategy, and the storage resource expansion strategy to obtain a resource expansion priority table;

[0042] The resource expansion dynamic elasticity factor is mapped to the resource expansion priority table, and the resource expansion dynamic scheduling strategy is set to obtain the resource allocation configuration table.

[0043] The present invention reserves and calculates various resources to be allocated through a preset resource division ratio, which helps to reasonably allocate resources and avoid excessive occupation or waste of resources. The live broadcast resource load threshold ladder is set and the resource expansion strategy is carried out based on this, providing flexible resource management for live broadcast activities, ensuring that resources can be adjusted in time when traffic fluctuates, and improving system stability and coping capabilities. For each resource type, specific expansion rules are formulated to effectively optimize the resource allocation strategy, so that the platform can intelligently schedule computing, bandwidth and storage resources under different load conditions. By setting a dynamic elasticity factor and combining the reserved resource call rules, the reserved resources can be flexibly called according to actual resource requirements, improving resource utilization efficiency and reducing resource idleness. The generation of an extended priority table helps to provide clear priorities for various resource allocations, thereby achieving optimal resource allocation. Through the implementation of the resource expansion dynamic scheduling strategy, platform resources can be efficiently managed in a dynamically changing environment, ensuring the smooth progress of live broadcast activities, and improving user experience and platform service capabilities.

[0044] Optionally, the user live broadcast dependency content classification distribution in step S3 is specifically:

[0045] Extracting user activity features from the user activity pattern data to obtain user interaction pattern data and user activity time pattern data;

[0046] According to the user live broadcast participation data, the user interaction mode data is divided into interaction mode participation levels to obtain high-participation interaction mode data and low-participation interaction mode data;

[0047] Perform time distribution division according to user activity time pattern data to obtain continuous activity time pattern data and discrete activity time pattern data;

[0048] Perform activity pattern intersection operations on high-engagement interaction pattern data and continuous activity time pattern data to obtain high-dependency activity pattern data; perform activity pattern intersection operations on low-engagement interaction pattern data and continuous activity time pattern data to obtain low-dependency activity pattern data;

[0049] Calculate the average dependency of the user's live content based on the high-dependency activity pattern data and the low-dependency activity pattern data to obtain the user's live content dependency data;

[0050] Performing live content resource demand analysis based on the live broadcast predicted traffic data to obtain predicted live content resource demand data, and performing multi-level live content division on the predicted live content resource demand data based on the resource allocation configuration table to obtain live content division data;

[0051] Based on the user live content dependency data, content-dependency matching is performed on the live content division data to obtain live content distribution data, wherein the live content distribution data includes user high dependency content distribution data and user low dependency content distribution data.

[0052] By performing feature extraction on user activity pattern data, the present invention can more accurately capture the user's interactive behavior and activity time pattern, and provide reliable data support for subsequent analysis. By dividing the degree of participation in the interactive pattern, it is possible to clarify which users have a higher degree of participation in the live content, so as to adjust the content distribution strategy in a targeted manner. The division of time distribution helps to identify the time periods when users are active, and further optimize the timing of live content delivery. The identification of high-dependency and low-dependency activity patterns, combined with the calculation of live content dependency, enables the platform to effectively evaluate the resource needs of different users, thereby optimizing resource allocation. By predicting the resource needs of live content, the resource needs of live content can be more accurately planned to avoid excessive or insufficient resource allocation. The matching of content dependencies ensures the efficient distribution of live content, which can improve the resource utilization efficiency of the platform while meeting user needs, and ensure the fluency of the live experience.

[0053] Optionally, the multi-preference layer caching of live content in step S3 is specifically:

[0054] Design the cache hierarchy based on the cloud platform resource allocation data to obtain the edge cache layer, cloud cache layer, and storage layer;

[0055] Use the edge cache layer to dynamically cache the live content edge node for the user's high-dependency content distribution data to obtain high-dependency content cache data;

[0056] Using the cloud cache layer to dynamically cache the live content in the cloud for the user's low-dependency content distribution data, and obtain low-dependency content cache data;

[0057] Calculate the cache preheating time during the traffic forecast peak period based on the live broadcast forecast traffic data to obtain the cache preheating time period data;

[0058] According to the cache warm-up time period data, the low-dependency content cache data and the high-dependency content cache data are prioritized for high-dependency content caching to obtain a cache layer calling strategy;

[0059] The storage layer is used to optimize the storage layer structure of the cache layer call strategy and the live content distribution data, obtain the cache storage optimization table, and upload it to the cloud platform to perform content caching tasks.

[0060] The present invention performs a cache hierarchy design for cloud platform resource allocation data, which can optimize data storage and access efficiency, and allocate different levels of cache to the edge cache layer, cloud cache layer and storage layer, thereby achieving more efficient data distribution. The edge cache layer dynamically caches high-dependency content, which can increase the user's access speed to live content, reduce latency, and improve the viewing experience; the cloud cache layer caches low-dependency content, effectively reducing bandwidth pressure and optimizing resource utilization. By calculating the cache preheating time through traffic prediction data, cache preparation can be made in advance to ensure smooth playback of content during peak hours. Prioritizing cached content helps to adjust the cache strategy according to real-time needs and user behavior, reduce the bandwidth occupied by content with low cache hit rate, and thus improve the overall service quality. The optimization of the storage layer ensures the rational use of resources in the cache and content distribution process, and further improves the system's responsiveness and resource scheduling efficiency.

[0061] Optionally, step S4 is specifically:

[0062] Step S41: acquiring real-time live network bandwidth data, and performing data preprocessing on the real-time live network bandwidth data to obtain real-time network bandwidth data to be analyzed;

[0063] Step S42: Evaluate the network quality of the user equipment according to the real-time network bandwidth data to be analyzed, and obtain the network quality data of the user equipment;

[0064] Step S43: setting a bit rate threshold for each user based on the user equipment network quality data to obtain user bit rate threshold data;

[0065] Step S44: extracting the network bandwidth dynamic change characteristics from the real-time network bandwidth data to be analyzed, and obtaining the network bandwidth dynamic data;

[0066] Step S45: Perform video adaptive encoding according to the network bandwidth dynamic data and the user bit rate threshold data, obtain a video quality strategy table, and upload it to the cloud platform to execute the video encoding task.

[0067] The present invention can comprehensively evaluate network conditions by acquiring and preprocessing real-time live network bandwidth data, providing an accurate basis for subsequent analysis. The network quality of user equipment is evaluated, and the bit rate of the video stream can be dynamically adjusted to adapt to different network environments, thereby avoiding video freezes and image quality degradation, and improving user experience. The bit rate threshold is set based on the device network quality data, so that each user can obtain the best video quality under their device and network conditions. This strategy can optimize bandwidth usage, ensure that bandwidth resources are reasonably allocated, and avoid ineffective consumption. After analyzing the dynamic change characteristics of real-time network bandwidth, the bandwidth fluctuation trend can be captured more accurately, providing real-time basis for video encoding, and realizing automatic adjustment of video quality. By performing adaptive encoding based on network bandwidth dynamic data and user bit rate threshold data, not only the quality of the video stream is optimized, but also different network bandwidth conditions can be dynamically adapted, the smoothness and stability of video playback are improved, and it is ensured that the cloud platform can perform efficient encoding tasks according to actual needs.

[0068] Optionally, step S5 specifically includes:

[0069] Step S51: acquiring cloud platform real-time resource allocation data, and performing data preprocessing on the cloud platform real-time resource allocation data to obtain real-time resource allocation data to be analyzed;

[0070] Step S52: Calculate the node average usage rate and node response time according to the real-time resource allocation data to be analyzed, and evaluate the resource utilization rate based on the calculated node average usage rate and node response time to obtain real-time resource utilization rate data;

[0071] Step S53: optimizing the low-utilization resource pool for the real-time resource allocation data to be analyzed based on the real-time resource utilization to obtain optimized resource pool allocation data;

[0072] Step S54: Perform load balancing resource dynamic expansion optimization on the optimized resource pool allocation data to obtain optimized cloud resource configuration data, and upload it to the cloud platform to execute the resource configuration task.

[0073] The present invention can ensure the accuracy of resource allocation information by acquiring and preprocessing the real-time resource allocation data of the cloud platform, and provide effective data support for subsequent optimization. The average utilization rate and response time of the computing nodes are combined with these data to evaluate the resource utilization rate, which helps to identify the bottleneck of resource use and monitor the platform load in real time. Based on these evaluation results, the low-utilization resource pool is optimized, which can effectively reduce resource waste and improve the overall resource utilization efficiency. By dynamically expanding and optimizing load balancing and cloud resource configuration, resource allocation can be intelligently adjusted when the demand of the cloud platform fluctuates, avoiding overcrowding or insufficient resources, and ensuring the stability and efficiency of the platform.

[0074] Optionally, this specification also provides a cloud computing management system for executing the cloud computing management method as described above, the cloud computing management system comprising:

[0075] The traffic demand prediction module is used to obtain user activity data and predict the traffic demand during the user's active time period based on the user activity data to obtain live broadcast predicted traffic data;

[0076] The resource demand elastic expansion module is used to obtain the cloud platform resource allocation data, and dynamically and elastically expand the cloud platform live broadcast resource demand according to the live broadcast predicted traffic data to obtain the resource allocation configuration table;

[0077] The live content distribution cache module is used to classify and distribute the content according to the user's live broadcast dependency according to the resource allocation configuration table and the live broadcast predicted traffic data to obtain the live content distribution data; cache the live content distribution data in multiple preference layers to obtain the cache storage optimization table, and upload it to the cloud platform to perform the content caching task;

[0078] The video adaptive encoding module is used to obtain the real-time live network bandwidth data, and perform bit rate stream adaptive video encoding according to the real-time live network bandwidth data, obtain the video quality strategy table, and upload it to the cloud platform to perform the video encoding task;

[0079] The cloud resource configuration module is used to obtain the real-time resource allocation data of the cloud platform, optimize the cloud resource configuration according to the real-time resource allocation data of the cloud platform, obtain the optimized cloud resource configuration data, and upload it to the cloud platform to execute the resource configuration task.

[0080] The cloud computing management system of the present invention can implement any one of the cloud computing management methods of the present invention, and is used to combine the operation and signal transmission media between various modules to complete the cloud computing management method. The internal modules of the system cooperate with each other, thereby improving the utilization efficiency of cloud computing resources and the communication efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0082] Figure 1 A schematic diagram of the steps of the cloud computing management method of the present invention;

[0083] Figure 2 Detailed step flow diagram of step S1 in the present invention;

[0084] Figure 3 Detailed step flow diagram of step S2 in the present invention;

[0085] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0086] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0087] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0088] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0089] To achieve this, please refer to Figures 1 to 3 The present invention provides a cloud computing management method, the method comprising the following steps:

[0090] Step S1: Obtain user activity data, and predict the traffic demand during the user's active time period based on the user activity data to obtain live broadcast predicted traffic data;

[0091] In this embodiment, the user activity log is constructed by collecting the user's interactive data on the e-commerce live broadcast platform in real time (such as viewing time, number of comments, number of shares, number of likes, etc.), and the time series analysis method is applied to predict the log data. To predict the user's active time period traffic demand. For the traffic demand of different user groups, cluster analysis can be performed based on historical behavior data, and the model can be trained using machine learning algorithms (such as support vector machines, random forests, etc.) to predict the live broadcast traffic demand in the future. For example, based on the traffic fluctuation data of a similar time period last month, the traffic demand changes in each time period are predicted to obtain "traffic forecast data", and the peak and trough periods are distinguished in combination with the traffic trend, and the traffic changes are expected to reserve resources.

[0092] Step S2: Obtain cloud platform resource allocation data, and dynamically and elastically expand the cloud platform live broadcast resource demand based on the live broadcast predicted traffic data to obtain a resource allocation configuration table;

[0093] In this embodiment, real-time resource allocation is obtained from the resource management system of the cloud platform, including computing resources (such as CPU, memory), storage resources (such as disk, database) and bandwidth resources (such as upload bandwidth, download bandwidth). By analyzing the predicted live broadcast traffic data, combined with the fluctuation of resource demand during the user's active time period, resources are elastically expanded. For example, if the traffic demand in a certain time period is predicted to be large, the resource allocation of the cloud platform can be adjusted through the automated resource scheduling system, and more computing and bandwidth resources can be allocated in advance to ensure stability and smoothness during the live broadcast. The cloud platform will automatically add computing nodes, expand bandwidth, and generate a "resource allocation configuration table" based on the expected traffic to facilitate the implementation of subsequent resource scheduling.

[0094] Step S3: Classify and distribute the content according to the user's live broadcast dependency according to the resource allocation configuration table and the live broadcast predicted traffic data to obtain live broadcast content distribution data; cache the live broadcast content distribution data in multiple preference layers to obtain a cache storage optimization table, and upload it to the cloud platform to perform the content caching task;

[0095] In this embodiment, the user content is classified by dependency by analyzing each user's live viewing behavior, interaction intensity, and viewing preferences, combined with the data in the resource allocation configuration table. High-dependence users (such as users who interact frequently and watch for a long time) will be assigned high-quality, low-latency content streams, while low-dependence users will be assigned ordinary streams. On this basis, the content will be cached hierarchically according to preference, and edge computing devices will be used to cache hotter content to reduce the latency of cloud requests. In this process, based on real-time traffic predictions, the cache preheating time for each live content is determined, and a "cache storage optimization table" is constructed and uploaded to the cloud platform. The cloud platform performs distribution and caching tasks to ensure that the caching mechanism can be efficiently executed during predicted traffic peaks, thereby improving the viewing experience.

[0096] Step S4: obtaining real-time live network bandwidth data, and performing bit rate stream adaptive video encoding according to the real-time live network bandwidth data, obtaining a video quality strategy table, and uploading it to the cloud platform to execute the video encoding task;

[0097] In this embodiment, the cloud platform will collect the network bandwidth data of live broadcast users in real time, mainly by measuring the upload and download bandwidth of each viewer's device. By analyzing the network bandwidth and delay, the network quality of each user's device is evaluated, and the video stream is further bitrate adaptively encoded according to the network conditions. If the network quality of the user's device is detected to be poor, the video encoding quality will be automatically adjusted, and the video resolution and bit rate will be reduced to adapt to the user's bandwidth, otherwise the quality will be improved. For example, if the user's bandwidth is lower than 1Mbps, the video quality will be automatically adjusted to 360p; if the bandwidth is higher than 10Mbps, it will be automatically upgraded to 1080p HD. Through these dynamic adjustments, a "video quality policy table" is generated, and the policy table is uploaded to the cloud platform to perform encoding tasks to ensure that each user can get the best viewing experience.

[0098] Step S5: Obtain the real-time resource allocation data of the cloud platform, and optimize the cloud resource configuration according to the real-time resource allocation data of the cloud platform, obtain the optimized cloud resource configuration data, and upload it to the cloud platform to execute the resource configuration task.

[0099] In this embodiment, the cloud platform obtains real-time usage data of cloud resources through a pre-deployed resource monitoring system, including the current utilization of computing, storage, and bandwidth. By analyzing these data, it is possible to detect whether there is over-allocation or waste of resources, such as the situation where some nodes have excess computing resources and tight bandwidth resources. Based on the analysis of resource utilization, the cloud platform can automatically adjust resource configuration, for example, reallocate node resources with excess computing to areas with greater bandwidth demand. In order to achieve efficient use of resources, a "dynamic resource scheduling algorithm" is used to optimize each resource, and resource allocation is adjusted in combination with a load balancing strategy. The optimized resource configuration table will be generated and uploaded to the cloud platform to ensure that resource allocation is always in the optimal state.

[0100] Optionally, step S1 specifically includes:

[0101] Step S11: acquiring user activity data, and performing data preprocessing on the user activity data to obtain user activity data to be analyzed;

[0102] In this embodiment, the live cloud platform collects the user's activity data on the e-commerce live platform in real time, including the user's viewing time, interactive behavior (such as likes, comments, sharing, etc.), viewing device information, geographic location, etc. The collected data is cleaned and deduplicated to filter out invalid or abnormal data, such as repeated viewing records or meaningless interactions in a short period of time, so as to obtain the user activity data to be analyzed. For example, by setting all interaction records within a time window (such as 1 hour), frequent click behaviors in a short period of time are removed to ensure the accuracy and effectiveness of the data.

[0103] Step S12: extracting user activity features from the user activity data to obtain user activity time data and user live broadcast interaction data;

[0104] In this embodiment, for the acquired user activity data, feature extraction of user behavior data is performed through algorithms (such as K-means clustering or PCA dimensionality reduction). Specifically, the platform can extract the user's active time period (such as a specific time period of each day, the average length of time users watch live broadcasts, etc.) and interaction frequency (such as the number of comments, likes, and shares made by each user in the live broadcast). By calculating the user's activity frequency, activity time distribution, etc., "user activity time data" and "user live broadcast interaction data" can be generated. For example, a user watches live broadcasts for an average of 3 hours a day and interacts an average of 10 times during the viewing process. The data will be recorded as the user's behavioral characteristics.

[0105] Step S13: dividing the user activity time periodically according to the user activity time data to obtain the user activity time periodicity data;

[0106] In this embodiment, based on the user's activity time data, the platform will perform a periodic analysis of each user's active time period. For example, set a period (such as 7 days or 30 days), use Fourier transform or autoregressive model (ARIMA) to periodically divide the time data, and obtain the peak and trough periods of user activity. Assuming that in the past 7 days, the user's activity reached a peak between 2 pm and 4 pm and 8 pm to 10 pm, then the user's activity time periodic data can be clearly divided into these two time periods as its main active periods.

[0107] Step S14: performing user activity pattern recognition based on the user activity time periodicity data and the user live broadcast interaction data to obtain user activity pattern data;

[0108] In this embodiment, the platform uses the aforementioned extracted "user activity time periodicity data" and "user live broadcast interaction data" to identify user activity patterns through cluster analysis (such as DBSCAN or K-means algorithm). Specifically, the platform can divide users into different activity patterns, such as "high-frequency interaction type" and "low-frequency viewing type", based on the user's activity level and interaction frequency in a specific time period. During this process, the platform will use a clustering algorithm to determine the behavior type of each user and generate corresponding "user activity pattern data". For example, a user who watches live broadcasts for more than 2 hours each time and interacts frequently can be identified as a "high-frequency interaction type", while another user who only watches live broadcasts on holidays and has less interaction may be identified as a "low-frequency viewing type".

[0109] Step S15: construct a live broadcast traffic prediction model based on the user activity pattern data, and use the live broadcast traffic prediction model to predict the traffic during the user's active time period to obtain live broadcast predicted traffic data.

[0110] In this embodiment, the platform predicts future live traffic by constructing a machine learning model (such as random forest, regression analysis or neural network) based on user activity pattern data. By collecting and organizing user activity data, live interaction data and historical traffic data of the platform, feature variables (such as user active time period, interaction frequency, viewing time, etc.) are selected as input. Then, these features are trained using machine learning algorithms (such as regression analysis, random forest or long short-term memory network (LSTM)) to identify the relationship between user behavior patterns and live traffic. During the training process, the known traffic conditions in the historical data are used as labels to gradually optimize the model parameters. The model can predict the live traffic demand for a specific time period in the future based on information such as user activity patterns and time periods, and output the corresponding traffic prediction data for subsequent resource allocation and scheduling. Subsequently, the platform inputs historical activity data into the traffic prediction model, for example, using information such as user activity patterns, interaction intensity, viewing time period, etc., combined with the overall traffic trend of the platform, to predict the user's traffic demand in a certain time period in the future. Assuming that the model shows that the user activity in a certain period of time (such as 3pm to 5pm every day) will reach 20,000 people, and the overall interaction rate is high, the model will predict that the traffic demand in this period is a specific bandwidth and computing resources, and generate "live broadcast predicted traffic data" based on this. The accuracy of model prediction can be further improved through continuous iteration and optimization, such as adjusting model parameters using real-time data, to further improve the accuracy and efficiency of prediction.

[0111] Optionally, step S14 is specifically:

[0112] Step S141: performing interaction type weight classification on user live broadcast interaction data to obtain live broadcast interaction type weight data;

[0113] In this embodiment, the user's live broadcast interaction data is analyzed in detail to identify different types of interactive behaviors, such as "like", "comment", "share", "gift reward", etc. Each interactive behavior will be given different weights according to its contribution to the live broadcast heat and popularity and its impact on user engagement. For example, "comment" can be given a higher weight (such as 0.4) because comments usually represent a deeper user participation, while "like" is given a lower weight (such as 0.2) because like is a simpler interactive behavior. Based on these weights, the platform generates weight data for each interaction type, so that the user's interaction depth can be accurately measured in subsequent analysis.

[0114] Step S142: evaluating the user live broadcast participation degree of the user live broadcast interaction data according to the live broadcast interaction type weight data to obtain the user live broadcast participation degree data;

[0115] In this embodiment, the user's live interaction data is evaluated for participation based on the live interaction type weight data. Specifically, the platform performs weighted summation on each user's interaction data in different live broadcasts. For example, the user's live interaction data in Live Room A includes 50 likes, 10 comments, and 5 rewards. According to the preset weights, these interactions will be weighted and calculated to summarize the user's total participation score. If the weight of likes is 0.2, the weight of comments is 0.4, and the weight of rewards is 0.6, then the user's total participation in Live Room A is (50×0.2+10×0.4+5×0.6), thereby obtaining the user's participation data. This data provides a basis for subsequent behavioral analysis.

[0116] Step S143: clustering the user's live broadcast type preference according to the live broadcast participation data to obtain the user's live broadcast type preference data;

[0117] In this embodiment, user live broadcast type preferences are clustered based on the live broadcast participation data of users in different live broadcasts. Users are clustered according to their participation data, and clustering algorithms such as K-means or DBSCAN are used to divide the live broadcast types into different groups, such as "user high participation live broadcast type", "user medium participation live broadcast type" and "user low participation live broadcast type". For example, if the user has a high participation score in live broadcast room A, he will be classified into the "user high participation live broadcast type" group, while the user has a low score in live broadcast room B, he will be classified into the "user low participation live broadcast type" group. This process helps the platform understand the user's participation tendency in live broadcasts, and then optimize content push and live broadcast activity arrangements.

[0118] Step S144: performing user participation in live broadcast type time association based on the user activity time periodicity data and the user live broadcast type preference data to obtain live broadcast preference time data;

[0119] In this embodiment, the time correlation analysis of user participation in live broadcast types is performed based on the user activity time periodicity data and the user live broadcast type preference data. First, by performing time series analysis on the user live broadcast interaction data, the user's active time period is identified. For example, user A is active from 7 to 9 pm every day and prefers to watch shopping live broadcasts. Based on this time period and live broadcast type preference data, the platform can generate "live broadcast preference time data", which records each user's most active time period and their preferred live broadcast type.

[0120] Step S145: identifying the user's live broadcast activity participation behavior pattern according to the live broadcast preference time data to obtain live broadcast participation behavior pattern data;

[0121] In this embodiment, the platform can identify the live broadcast activity patterns of users through behavioral analysis algorithms (such as decision trees, cluster analysis, etc.). For example, according to the algorithm results, user A tends to watch long entertainment live broadcasts at 8 pm, while user B watches short science and technology live broadcasts at noon. The platform organizes these behavior patterns and extracts features, such as active time, live broadcast duration, interactive behavior frequency, etc., and finally identifies the live broadcast behavior pattern of each user.

[0122] Step S146: assigning behavior pattern labels to the live broadcast participation behavior pattern data to obtain user activity pattern data.

[0123] In this embodiment, detailed behavioral features are extracted based on the user's live broadcast activity behavior (such as viewing time, frequency of participation in interaction, preferred live broadcast content type, etc.). For example, user A watches entertainment live broadcasts for more than 2 hours between 7pm and 9pm, and frequently comments and likes. The platform marks him as a "high-engagement user in evening entertainment". For user B, if he prefers to watch fashion live broadcasts for no more than 30 minutes between 12pm and 2pm, and has less interaction, the platform marks him as a "low-engagement user at noon". The rules for label assignment are based on multi-dimensional behavioral data, such as participation, live broadcast time, interaction intensity, etc. The platform uses data mining algorithms (such as decision trees, random forests) to classify user behaviors and assign multiple labels to each user. The label system may include, but is not limited to: "high participation", "low participation", "prefer shopping", "prefer entertainment", "short-term viewing", "long-term viewing", etc.

[0124] Optionally, step S2 specifically includes:

[0125] Step S21: Obtain cloud platform resource allocation data, and perform data preprocessing on the cloud platform resource allocation data to obtain resource allocation data to be analyzed;

[0126] In this embodiment, resource allocation data is obtained from the cloud platform and preprocessed. Specifically, detailed data on resource allocation is collected from the cloud platform, including the usage of computing resources, storage resources, and network bandwidth. Through data cleaning and format conversion, the timeliness and accuracy of these data are ensured, and they are normalized so that data from different sources have a unified standard. After preprocessing, these data will be converted into a format suitable for analysis to prepare for subsequent resource demand forecasting and expansion strategy design. Taking computing resources as an example, if the original data is stored in different data formats, such as CSV files or database records, they will be uniformly converted into a unified format, such as JSON or a table structure, to facilitate subsequent analysis.

[0127] Step S22: estimating resource demand during the live broadcast activity according to the live broadcast predicted traffic data, and obtaining resource demand data during the live broadcast;

[0128] In this embodiment, the computing, bandwidth, and storage resources required during the live broadcast event are estimated based on the user activity in different time periods (such as the expected number of viewers, interaction frequency, etc.), combined with historical traffic data and prediction algorithms. For example, if the number of users is expected to reach 100,000 during a peak live broadcast period, the required computing resources and bandwidth will be estimated based on historical traffic data, and the storage demand will be calculated at the same time. Assuming that the expected network traffic during the peak period is 10GB per second, the required bandwidth resources will be calculated based on this demand. This process involves using machine learning models, such as regression analysis or time series prediction, to accurately calculate the various resources required during the live broadcast event.

[0129] Step S23: determining the node transmission boundary conditions of the cloud platform according to the resource allocation data to be analyzed, and obtaining the node transmission boundary conditions;

[0130] In this embodiment, the transmission boundary conditions of the cloud platform nodes are determined based on the resource allocation data to be analyzed. Specifically, the transmission capacity of the cloud platform nodes and their capacity limitations will be analyzed based on the allocation of different resources and the current network load. The transmission boundary conditions will include the maximum bandwidth, maximum processing power, and data storage limitations of the node. For example, assuming that the maximum bandwidth of a node is 1Gbps and the computing resources are limited to 50 CPU cores, the resource usage of the node will be limited according to these conditions. In this process, the resource allocation boundary of each node will be determined based on the resource data using an optimization algorithm (such as linear programming or heuristic algorithm) to ensure the effective use of all resources.

[0131] Step S24: matching the data transmission resource demand with the resource demand data during the live broadcast based on the node transmission boundary conditions, and obtaining transmission computing resource matching data, transmission bandwidth resource matching data, and transmission storage resource matching data;

[0132] In this embodiment, a global resource matching model is constructed according to the resource capacity of each node, such as computing power, bandwidth and storage space. For computing resources, a load balancing algorithm (such as weighted polling or least connection algorithm) is used to match according to the resource requirements during the live broadcast (such as live picture processing, computing load of real-time data analysis, etc.) and the computing power of each node. For bandwidth resources, the bandwidth required in each time period is estimated by the network traffic prediction model, and the resource allocation is adjusted according to the maximum bandwidth capacity of each node (such as the maximum transmission rate per second). For example, if the bandwidth demand of a node is 100Mbps, and the maximum bandwidth of the node is 200Mbps, 100Mbps of bandwidth will be allocated to the node. If the bandwidth demand is greater than the maximum carrying capacity of the node, other nodes with relatively idle bandwidth resources will be automatically selected for resource allocation. For storage resources, the generation and storage requirements of cached data during the live broadcast process will be taken into account, and the appropriate storage node will be matched according to the storage capacity of each node to ensure that all content can be accessed in time and avoid delays or losses caused by storage bottlenecks. For example, real-time live content will be cached to nodes with stronger storage capabilities first, and the allocation of storage resources will be adjusted as needed to ensure reliable storage of streaming media data.

[0133] Step S25: setting resource automatic expansion strategy for the transmission computing resource matching data, the transmission bandwidth resource matching data and the transmission storage resource matching data to obtain a resource allocation configuration table.

[0134] In this embodiment, the resource allocation strategy is dynamically adjusted according to the real-time traffic demand, resource utilization rate and node load. For example, if the bandwidth usage of a node is close to the upper limit, the bandwidth resource expansion mechanism will be triggered to automatically allocate additional bandwidth from other nodes or data centers of the cloud platform to meet the demand. In addition, computing resources and storage resources will also be dynamically adjusted according to real-time demand changes. Set an expansion threshold, for example, start the expansion operation when the bandwidth utilization rate exceeds 80%. Ultimately, all resource expansion strategies and configurations will form a resource allocation configuration table and upload it to the cloud platform for execution.

[0135] Optionally, step S25 is specifically:

[0136] According to a preset resource division ratio, reserved resource division is performed on the transmission computing resource matching data, the transmission bandwidth resource matching data, and the transmission storage resource matching data to obtain reserved resource data, to-be-allocated computing resource data, to-be-allocated bandwidth resource data, and to-be-allocated storage resource data;

[0137] In this embodiment, reserved resources are divided for transmission computing resource matching data, transmission bandwidth resource matching data, and transmission storage resource matching data according to a preset resource division ratio. Specifically, the preset resource division ratio is set according to historical traffic data and live broadcast demand. For example, 30% of resources are reserved for computing resources, 40% for bandwidth resources, and 30% for storage resources. During the division process, the resource requirements of each node are calculated, and the resources are allocated to different nodes in proportion through a load balancing algorithm. The remaining resources of each node are used to cope with real-time fluctuations or abnormal demands. For example, during certain high-traffic periods, the allocation ratio of these resources will be dynamically adjusted to meet the load requirements of live broadcasts.

[0138] According to the resource demand data during the live broadcast, the live broadcast resource load threshold is set in steps to obtain a live broadcast resource extension step threshold set, wherein the live broadcast resource extension step threshold set includes a computing resource step extension threshold, a bandwidth resource step extension threshold, and a storage resource step extension threshold;

[0139] In this embodiment, the live broadcast resource load threshold ladder is set according to the resource demand data during the live broadcast. This process sets the ladder of different load thresholds based on the predicted data of the live broadcast traffic. For example, three load ladders can be set: low load (demand less than 50%), medium load (demand between 50%-80%), and high load (demand more than 80%). Each ladder has a different resource expansion strategy. Only the most basic resources are needed in the low load stage; at medium load, additional resources will be added for support; and the high load stage requires a lot of additional computing and bandwidth resources. The live broadcast resource expansion ladder threshold set includes the computing resource ladder expansion threshold, the bandwidth resource ladder expansion threshold, and the storage resource ladder expansion threshold.

[0140] Based on the computing resource step expansion threshold, computing resource expansion rules are set for the computing resource data to be allocated, and a computing resource expansion strategy is obtained;

[0141] In this embodiment, based on the computing resource step expansion threshold, the expansion rules for the allocated computing resource data are set. Resources are allocated through elastic computing instances in the cloud to ensure that computing power meets demand. Computing resource expansion strategies may include measures such as expanding the number of virtual machines and increasing processing power. At the same time, the platform will dynamically monitor resource usage to avoid unnecessary waste of resources. For example, when it is monitored that the computing resource demand exceeds 80%, the elastic computing expansion mechanism will be automatically started, and new virtual machine instances will be added to the cloud platform to expand computing resources. This expansion strategy can also be dynamically adjusted according to the real-time usage of resources. For example, when some nodes are under load, the computing resources of these nodes will be automatically reduced to achieve more efficient resource allocation.

[0142] Based on the bandwidth resource step expansion threshold, bandwidth resource expansion rules are set for the bandwidth resource data to be allocated to obtain a bandwidth resource expansion strategy;

[0143] In this embodiment, the bandwidth resource expansion rules are set according to the bandwidth resource step expansion threshold. When the bandwidth demand reaches the high load stage, the traffic peak is handled by multi-link aggregation or increasing the bandwidth capacity of the data center. Bandwidth resource expansion strategies include dynamically adjusting network routing, increasing cache node bandwidth, etc. to ensure smooth data transmission and avoid bandwidth bottlenecks. For example, when the demand for live broadcast bandwidth surges, the traffic routing will be adjusted or the CDN acceleration service will be enabled according to the preset bandwidth resource expansion strategy to ensure efficient use of bandwidth resources and avoid bandwidth bottlenecks affecting the quality of live broadcasts.

[0144] Based on the storage resource step expansion threshold, storage resource expansion rules are set for the storage resource data to be allocated to obtain a storage resource expansion strategy;

[0145] In this embodiment, storage resource expansion rules are set for the storage resource data to be allocated based on the storage resource step expansion threshold. The storage expansion strategy mainly includes capacity expansion strategy, storage backup strategy, storage data migration strategy and distributed storage optimization strategy. The capacity expansion strategy is to trigger the expansion of storage capacity when the storage demand exceeds the current capacity threshold. This includes adding storage space on existing storage nodes or migrating data to additional storage nodes. In specific operations, the storage allocation is dynamically adjusted according to the popularity of the data, access frequency, etc. For example, for the storage of popular live content, higher-performance storage nodes are preferred for expansion to ensure the smoothness of the user experience. In order to prevent data loss, the storage backup strategy will be started according to the growth of storage resource demand, and the live data will be backed up to the off-site storage node regularly. When the storage load reaches a certain threshold, the backup operation will be executed first to ensure the security and recovery capability of key data. The storage data migration strategy is to start the storage data migration strategy when resources are tight, and migrate cold data to low-cost storage nodes on demand. Specifically, the migration rules are determined based on the frequency of data access, and infrequently accessed data is migrated from fast storage devices to slower storage media, thereby optimizing the use efficiency of storage resources. When storage resources are under high load, the distributed storage optimization strategy is enabled to automatically balance the load by horizontally expanding storage nodes, reduce data storage bottlenecks, and improve access speed.

[0146] Set resource expansion dynamic elasticity factor according to node transmission boundary conditions;

[0147] In this embodiment, the dynamic elasticity factor of resource expansion is set according to the node transmission boundary conditions. The setting of the dynamic elasticity factor will be adjusted according to the resource usage of different nodes. For example, some nodes will set a higher elasticity factor so that they can get more resource support first during peak traffic periods. These elasticity factors can be adjusted according to different requirements for computing, bandwidth, and storage resources. For example, when the demand for computing resources is high, the elasticity factor of the computing node will be increased to ensure that these nodes will be given priority to expand computing resources when under high load.

[0148] According to the resource expansion dynamic elasticity factor, a reserved resource calling rule is set for the reserved resource data to obtain a reserved resource calling strategy;

[0149] In this embodiment, reserved resource calling rules are set for reserved resource data according to the elasticity factor. The reserved resource calling rules will call reserved resources according to a certain priority based on the node resource usage. For example, in a live broadcast event, the demand for computing resources and bandwidth will suddenly increase, and reserved resources will be called to ensure that there will be no shortage of resources. At the same time, the calling priority of storage resources is lower, and resource scheduling will be carried out according to the urgency of storage demand.

[0150] Allocate expansion priorities for the reserved resource call strategy, the computing resource expansion strategy, the bandwidth resource expansion strategy, and the storage resource expansion strategy to obtain a resource expansion priority table;

[0151] In this embodiment, the expansion priority is allocated according to the reserved resource call strategy, the computing resource extension strategy, the bandwidth resource extension strategy, and the storage resource extension strategy. The reserved resource call strategy plays a vital role in the resource extension priority table. The reserved resource call strategy is to set a priority allocation rule based on the actual demand and the reserved resource amount. For example, reserved computing resources and bandwidth resources usually have a higher call priority because these resources are essential for the real-time and smoothness of live broadcast activities. Although storage resources also need to be expanded, they are usually at a lower priority, especially when resource demand will not increase sharply in the short term. Therefore, in the resource extension priority table, the role of the reserved resource call strategy is to ensure that during the peak period of real-time traffic, computing and bandwidth resources are prioritized, and the call of storage resources is postponed to a lower load stage. This enables the platform to control resource allocation more accurately, avoid excessive resource consumption, and ensure user experience. Specifically, computing resources and bandwidth resources will be assigned a higher priority according to their importance during the peak period of live broadcast traffic, while the expansion of storage resources will be assigned a lower priority. The resource extension priority table will be dynamically adjusted according to the actual load situation to ensure that the platform can efficiently handle different types of resource requirements.

[0152] The resource expansion dynamic elasticity factor is mapped to the resource expansion priority table, and the resource expansion dynamic scheduling strategy is set to obtain the resource allocation configuration table.

[0153] In this embodiment, the dynamic elasticity factor of resource expansion is mapped to the resource expansion priority table, and the dynamic scheduling strategy of resource expansion is further set. The elasticity factor will be dynamically adjusted according to the load conditions of different nodes, and these factors will be mapped to the resource expansion priority table. Specifically, the expansion elasticity factor of each resource type (computing, bandwidth, storage) will be calculated, and these factors determine the priority when resources are allocated. For example, if the computing resource demand of a node is very high, and the bandwidth demand is low, the elasticity factor of the computing resource will be set to a higher value (e.g., 1.5), and the elasticity factor of the bandwidth resource will be set to a lower value (e.g., 0.8). This mapping will adjust the scheduling mode of platform resources in real time and flexibly allocate resources according to load changes. Through this mapping, the platform can dynamically adjust the scheduling strategy of resources when the resource load changes, such as realizing the dynamic expansion or contraction of different resources through the load balancing algorithm, ensuring that the platform can flexibly respond to the actual load demand and ensure the quality and stability of the live broadcast service. In specific implementation, resource allocation will be adjusted through real-time data traffic analysis to avoid excessive concentration or waste of resources. For example, during the peak period of live broadcast, the demand for computing and bandwidth resources may surge, and computing and bandwidth resources will be prioritized, while during low-load periods, the platform will reduce unnecessary resource expansion to reduce operating costs. Ultimately, through this scheduling strategy, the platform can achieve efficient resource management and ensure the stability and smoothness of live broadcast activities.

[0154] Optionally, the user live broadcast dependency content classification distribution in step S3 is specifically:

[0155] Extracting user activity features from the user activity pattern data to obtain user interaction pattern data and user activity time pattern data;

[0156] In this embodiment, feature extraction is performed on user activity pattern data to obtain user interaction pattern data and user activity time pattern data. Specifically, interaction patterns (such as likes, comments, shares, etc.) and activity time patterns (such as active time periods in live broadcasts) are extracted from user activity data. For example, by analyzing the frequency of user interaction during a live broadcast, the types of live broadcasts and interactive behaviors in which users frequently participate can be identified to obtain interaction pattern data. At the same time, through time analysis, the time periods in which users are active can be identified to obtain activity time pattern data (for example, users usually participate in interactions between 8 and 10 p.m.).

[0157] According to the user live broadcast participation data, the user interaction mode data is divided into interaction mode participation levels to obtain high-participation interaction mode data and low-participation interaction mode data;

[0158] In this embodiment, a participation threshold is set. For example, by calculating the number of likes, comments, and shares of each user in the live broadcast, the interactive participation score is obtained. Users with scores higher than the set threshold will be classified as "high participation interaction mode", while users with scores lower than the threshold will be classified as "low participation interaction mode". For example, users actively comment and share in live broadcast A, and obtain a high participation score, so they are classified as high participation interaction mode.

[0159] Perform time distribution division according to user activity time pattern data to obtain continuous activity time pattern data and discrete activity time pattern data;

[0160] In this embodiment, the division is performed by analyzing the continuity of user activity time. For example, if a user remains active for a long time during the live broadcast, it will be divided into a continuous activity time mode; and if the user's activity time is scattered and the intervals are long, it will be divided into a discrete activity time mode. In this way, the activity patterns of different users can be captured more accurately, helping the platform to better predict the time window of user activity.

[0161] Perform activity pattern intersection operations on high-engagement interaction pattern data and continuous activity time pattern data to obtain high-dependency activity pattern data; perform activity pattern intersection operations on low-engagement interaction pattern data and continuous activity time pattern data to obtain low-dependency activity pattern data;

[0162] In this embodiment, by combining the relationship between user interaction patterns and activity time, the user group with a high interaction frequency and a long active time during the live broadcast is analyzed and defined as a high-dependence activity pattern. On the contrary, users with low participation and scattered time activities are defined as low-dependence activity patterns. For example, the user has a high interaction frequency and a relatively concentrated activity time in live broadcast A, and the system classifies it as a high-dependence activity pattern. Although the user interacts in live broadcast B, his participation in activities is relatively scattered, which belongs to the low-dependence activity pattern.

[0163] Calculate the average dependency of the user's live content based on the high-dependency activity pattern data and the low-dependency activity pattern data to obtain the user's live content dependency data;

[0164] In this embodiment, the average dependency of the user's live content is calculated based on the high-dependence activity mode data and the low-dependence activity mode data to obtain the user's live content dependency data. Specifically, by calculating the degree of each user's dependency on different live content (such as viewing time, interaction frequency, etc.), the average dependency of the user on the same type of live content is calculated to obtain the dependency score on each type of live content. This score helps the platform evaluate which content is more attractive to user groups with high and low dependency, and then optimize the content distribution strategy.

[0165] Performing live content resource demand analysis based on the live broadcast predicted traffic data to obtain predicted live content resource demand data, and performing multi-level live content division on the predicted live content resource demand data based on the resource allocation configuration table to obtain live content division data;

[0166] In this embodiment, live broadcast prediction traffic data, including parameters such as user visits and viewing time, is used to perform detailed resource demand calculations. For example, the bandwidth demand, computing resource demand, and storage resource demand of specific live broadcast content within a certain time period are predicted based on the user's viewing behavior. Assume that the prediction results show that the bandwidth required for a large-scale promotional event live broadcast is 6Gbps, the computing resources require a 10-core CPU, and the storage demand is 15TB. Based on these prediction data, the demand intensity of different live broadcast content for various resources is further analyzed to obtain live broadcast content resource demand data. Then, based on the resource allocation configuration table, the live broadcast content is divided into multiple levels according to different resource demand levels. For example, content with higher resource demand will be marked as "high priority content", content with moderate demand will be marked as "medium priority content", and content with low demand will be marked as "low priority content" to facilitate subsequent resource scheduling and allocation.

[0167] Based on the user live content dependency data, content-dependency matching is performed on the live content division data to obtain live content distribution data, wherein the live content distribution data includes user high dependency content distribution data and user low dependency content distribution data.

[0168] In this embodiment, matching is performed according to each user's dependence on different types of live content. For example, users with high dependence on live content usually have long viewing time and high interaction frequency. Therefore, high-dependence content with high resource demand and strong interactivity will be distributed to such users first. Assuming that user A shows a high interaction frequency and a long viewing time in multiple live broadcasts of the same type, the user will be classified as a "high-dependence" user, and the same type of live content with large resource demand will be recommended to him, such as limited-time promotion live broadcasts or celebrity live broadcasts. On the contrary, for the user's low-dependence live content, low-dependence content with small resource demand and weak interactivity will be recommended based on its low interaction frequency and short viewing time, such as general product display live broadcasts or previews of educational live content. Finally, the generated live content distribution data includes user high-dependence content distribution data and user low-dependence content distribution data, and the matching rules ensure the accurate delivery of content to improve user experience and optimize resource utilization efficiency.

[0169] Optionally, the multi-preference layer caching of live content in step S3 is specifically:

[0170] Design the cache hierarchy based on the cloud platform resource allocation data to obtain the edge cache layer, cloud cache layer, and storage layer;

[0171] In this embodiment, the cache hierarchy is designed according to the cloud platform resource allocation data to obtain the edge cache layer, cloud cache layer and storage layer. Specifically, according to different cache requirements and resource allocation situations, the cache is divided into three levels: edge cache layer, cloud cache layer and storage layer. The edge cache layer is mainly responsible for caching nodes closer to the user to reduce latency; the cloud cache layer is responsible for processing cache content with lower demand, and the storage layer is used to optimize large-scale storage and long-term data management. Assuming that the edge cache layer can accommodate a maximum of 10TB of content, and the cloud cache layer provides a 30TB cache capacity, the cache capacity of each layer will be dynamically adjusted based on the platform's real-time analysis and allocation of resources.

[0172] Use the edge cache layer to dynamically cache the live content edge node for the user's high-dependency content distribution data to obtain high-dependency content cache data;

[0173] In this embodiment, the user's focus is analyzed based on the real-time audience data and interactive behavior of the live broadcast, and the live broadcast content with high user dependence is identified. Based on the predicted traffic data of the live broadcast, the edge node dynamically adjusts its cache content and capacity, and pre-loads relevant high-dependency content (such as celebrity live broadcasts, popular events, etc.) during the time period when the traffic is expected to be large to ensure smooth live broadcast. For example, before a large live broadcast event starts, the edge node loads high-dependency content 30 minutes in advance and adjusts the cache data according to the number of real-time users.

[0174] Using the cloud cache layer to dynamically cache the live content in the cloud for the user's low-dependency content distribution data, and obtain low-dependency content cache data;

[0175] In this embodiment, the cloud cache layer is used to dynamically cache the live content in the cloud for the user's low-dependency content distribution data to obtain low-dependency content cache data. For low-dependency content (such as live news broadcasts, conventional educational content, etc.), these contents are dynamically cached to the cloud. By analyzing the user's viewing behavior in real time, it is identified which content has less user demand, and cache management is performed based on the timeliness and demand fluctuations of the content. The cloud cache capacity is adjusted according to the content characteristics. The cache time of low-dependency content can be set to 24 hours or longer, and low-dependency content that is refreshed frequently can be refreshed within 48 hours.

[0176] Calculate the cache preheating time during the traffic forecast peak period based on the live broadcast forecast traffic data to obtain the cache preheating time period data;

[0177] In this embodiment, the cache preheating time during the predicted traffic peak is calculated based on the live broadcast predicted traffic data to obtain the cache preheating time period data. By analyzing the historical data of past live broadcast activity traffic, the upcoming peak traffic can be accurately predicted. For example, the number of spectators at a music concert is expected to reach a peak in 30 minutes. According to historical data, the peak of traffic is calculated to occur 15 minutes before the start of the event, so the cache preheating time period is set to 15 minutes. During this period, high-frequency live content will be loaded first, and the cache layer will complete content loading within this time.

[0178] According to the cache warm-up time period data, the low-dependency content cache data and the high-dependency content cache data are prioritized for high-dependency content caching to obtain a cache layer calling strategy;

[0179] In this embodiment, the cache layer call strategy is obtained by prioritizing the low-dependency content cache data and the high-dependency content cache data according to the cache warm-up time period data. Specifically, based on the time period data of the traffic forecast peak period, priority will be set for the cached content. For example, 15 minutes before the peak of the live broadcast of a concert, high-dependency content related to the concert (such as concert previews, singer interviews, etc.) will be loaded into the edge cache layer first, and low-dependency content (such as other live broadcasts and general information content) will be loaded later in the cloud cache. This strategy can ensure that when the number of user visits increases significantly, the core live content can be played smoothly first.

[0180] The storage layer is used to optimize the storage layer structure of the cache layer call strategy and the live content distribution data, obtain the cache storage optimization table, and upload it to the cloud platform to perform content caching tasks.

[0181] In this embodiment, the storage layer is used to optimize the storage layer structure of the cache layer call strategy and the live content distribution data, obtain the cache storage optimization table, and upload it to the cloud platform to perform the content caching task. The storage layer optimizes the structure of the cache call strategy according to the content distribution requirements. Specifically, the storage requirements of each live content are analyzed, and the most appropriate storage resources are allocated to different cache layers based on factors such as content type, user viewing time, and user activity. The cache structure of high-dependency content can use multi-copy storage, while low-dependency content can use compressed storage to save space. The data table after storage optimization will be uploaded to the cloud platform to ensure that the cache task is smoothly executed during the actual live broadcast.

[0182] Optionally, step S4 is specifically:

[0183] Step S41: acquiring real-time live network bandwidth data, and performing data preprocessing on the real-time live network bandwidth data to obtain real-time network bandwidth data to be analyzed;

[0184] In this embodiment, real-time live network bandwidth data is obtained through the live broadcast platform, and the real-time live network bandwidth data is preprocessed to obtain real-time network bandwidth data to be analyzed. In this process, the live network bandwidth is monitored in real time, and the data of each node, such as the download speed, delay and other parameters of each live audience terminal, are collected and cleaned and normalized. For example, the bandwidth data will be cut into time granularities such as seconds and minutes and outliers will be removed to ensure the accuracy and availability of the data. After preprocessing, a real-time network bandwidth data set to be analyzed is obtained, which includes the real-time bandwidth information of each time period and each node.

[0185] Step S42: Evaluate the network quality of the user equipment according to the real-time network bandwidth data to be analyzed, and obtain the network quality data of the user equipment;

[0186] In this embodiment, the user device network quality is evaluated based on the real-time network bandwidth data to be analyzed to obtain the user device network quality data. Based on the preprocessed data, the platform evaluates the quality of each user's device network, and scores mainly based on parameters such as the user's download rate, network delay, and packet loss rate. For example, if the user's download rate is lower than a certain threshold (such as 2Mbps), it is considered that the network quality is poor; if the packet loss rate is greater than 10%, it is considered that there are obvious problems with the network quality. These evaluation results will form a "network quality score data", which will further provide a basis for the bit rate adjustment in subsequent steps.

[0187] Step S43: setting a bit rate threshold for each user based on the user equipment network quality data to obtain user bit rate threshold data;

[0188] In this embodiment, a bit rate threshold suitable for each user's network conditions is set based on the user's network quality score. For example, if a user's network quality is good (rate greater than 5Mbps), the bit rate threshold is set to 4Mbps; if the network is poor (such as less than 1Mbps), the bit rate threshold is set to 0.5Mbps. In this way, the platform can provide each user with a dynamically adapted bit rate level to ensure smooth video playback while reducing freezes or delays. The setting of the bit rate threshold not only takes into account the current network quality, but also makes dynamic adjustments based on factors such as user device performance and video content complexity.

[0189] Step S44: extracting the network bandwidth dynamic change characteristics from the real-time network bandwidth data to be analyzed, and obtaining the network bandwidth dynamic data;

[0190] In this embodiment, through time series analysis, the trend characteristics of bandwidth fluctuations are extracted from the real-time bandwidth data to be analyzed. Specifically, the platform uses sliding window technology to perform segmented analysis on the bandwidth data, calculates the average, maximum and minimum bandwidth values ​​for each time period, and detects the bandwidth fluctuation frequency, identifying the periodicity of bandwidth peaks and valleys. For example, if the bandwidth fluctuates greatly and the bandwidth drops rapidly within a certain time period, it will be marked as a "bandwidth fluctuation" state and prepare to switch the encoding strategy. The bandwidth dynamic data will become the basis for video quality optimization and help the system make real-time adjustments.

[0191] Step S45: Perform video adaptive encoding according to the network bandwidth dynamic data and the user bit rate threshold data, obtain a video quality strategy table, and upload it to the cloud platform to execute the video encoding task.

[0192] In this embodiment, the platform sets the video adaptive encoding strategy by combining the bandwidth dynamic data and the bit rate threshold. When the bandwidth is at its peak and meets the user bit rate threshold, the platform provides high-definition video quality (such as 1080p); if the bandwidth drops to a lower level and the user bit rate threshold is strictly limited, the video quality will be dynamically downgraded (such as 720p, 480p) to ensure smooth video playback. In addition, the encoding strategy will be adjusted in different time periods according to bandwidth fluctuations to avoid video freezes or delays. Finally, the encoding strategy table generated by the platform will be uploaded to the cloud platform to guide the encoding system to adjust the quality of the user's video stream in real time.

[0193] Optionally, step S5 specifically includes:

[0194] Step S51: acquiring cloud platform real-time resource allocation data, and performing data preprocessing on the cloud platform real-time resource allocation data to obtain real-time resource allocation data to be analyzed;

[0195] In this embodiment, the allocation of resources such as CPU, memory, storage, bandwidth, etc. of each node is obtained from the resource management system of the cloud platform in real time through the API interface. The collected data includes information such as the current load, resource consumption, and allocation of each node. In order to ensure the accuracy and integrity of the data, the system will format the data, such as unifying the timestamp and unit (such as GB, Mbps), and remove invalid or abnormal data (such as data points that failed to read or timed out), to ensure that the cleaned resource allocation data set to be analyzed is finally obtained.

[0196] Step S52: Calculate the node average usage rate and node response time according to the real-time resource allocation data to be analyzed, and evaluate the resource utilization rate based on the calculated node average usage rate and node response time to obtain real-time resource utilization rate data;

[0197] In this embodiment. By averaging the real-time consumption data of resources such as CPU, memory, bandwidth, etc. of each node, the system obtains the average resource utilization rate of each node. Next, the platform will also calculate the response time of the node (i.e., the time delay for the node to process the request) to judge the responsiveness of the resource. Based on these two data (node ​​utilization rate and response time), the platform evaluates the utilization rate of the overall resources, and calculates the real-time resource utilization data according to the standardized calculation formula (for example, average node utilization rate = (node ​​CPU utilization rate + node memory utilization rate + node bandwidth utilization rate) / 3), thereby obtaining an evaluation result of whether the resource utilization is efficient.

[0198] Step S53: optimizing the low-utilization resource pool for the real-time resource allocation data to be analyzed based on the real-time resource utilization to obtain optimized resource pool allocation data;

[0199] In this embodiment, for the low resource utilization nodes calculated in real time in the platform (such as the CPU utilization of some nodes is lower than the set threshold of 30%), the resources in these low-utilization resource pools are reconfigured according to the resource pool optimization algorithm. Specifically, the resources in the low-load resource pool will be reallocated to the nodes with high resource utilization. During the optimization process, the configuration of the resource pool is adjusted according to the preset optimization rules (for example, the storage space in the storage resource pool with lower load is preferentially allocated to the computing nodes with greater demand), to ensure that the resource usage of each node is more balanced and to improve the overall resource utilization efficiency.

[0200] Step S54: Perform load balancing resource dynamic expansion optimization on the optimized resource pool allocation data to obtain optimized cloud resource configuration data, and upload it to the cloud platform to execute the resource configuration task.

[0201] In this embodiment, the resource pool will be further load-balanced according to the optimized resource pool allocation data obtained. If the load of some resource pools is still high, a dynamic expansion strategy will be triggered, for example, automatically increasing the number of computing nodes or storage nodes, or transferring part of the load to idle nodes. The load distribution of the resource pool is optimized through a dynamic expansion mechanism (for example, starting more virtual machine instances or automatically adjusting resource allocation parameters) to ensure the overall performance and response speed of the cloud platform. After the optimization is completed, the platform generates new resource configuration data and uploads it to the resource management system of the cloud platform to ensure real-time update and execution of resource configuration.

[0202] Optionally, this specification also provides a cloud computing management system for executing the cloud computing management method as described above, the cloud computing management system comprising:

[0203] The traffic demand prediction module is used to obtain user activity data and predict the traffic demand during the user's active time period based on the user activity data to obtain live broadcast predicted traffic data;

[0204] The resource demand elastic expansion module is used to obtain the cloud platform resource allocation data, and dynamically and elastically expand the cloud platform live broadcast resource demand according to the live broadcast predicted traffic data to obtain the resource allocation configuration table;

[0205] The live content distribution cache module is used to classify and distribute the content according to the user's live broadcast dependency according to the resource allocation configuration table and the live broadcast predicted traffic data to obtain the live content distribution data; cache the live content distribution data in multiple preference layers to obtain the cache storage optimization table, and upload it to the cloud platform to perform the content caching task;

[0206] The video adaptive encoding module is used to obtain the real-time live network bandwidth data, and perform bit rate stream adaptive video encoding according to the real-time live network bandwidth data, obtain the video quality strategy table, and upload it to the cloud platform to perform the video encoding task;

[0207] The cloud resource configuration module is used to obtain the real-time resource allocation data of the cloud platform, optimize the cloud resource configuration according to the real-time resource allocation data of the cloud platform, obtain the optimized cloud resource configuration data, and upload it to the cloud platform to execute the resource configuration task.

[0208] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0209] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A cloud computing management method, characterized in that: The following steps are involved: Step S1: Obtain user activity data, and predict the traffic demand during the user's active time period based on the user activity data to obtain live broadcast predicted traffic data; Step S2: Obtain cloud platform resource allocation data, and dynamically and elastically expand the cloud platform live broadcast resource demand based on the live broadcast predicted traffic data to obtain a resource allocation configuration table; Step S3: Classify and distribute the content according to the user's live broadcast dependency according to the resource allocation configuration table and the live broadcast predicted traffic data to obtain live broadcast content distribution data; cache the live broadcast content distribution data in multiple preference layers to obtain a cache storage optimization table, and upload it to the cloud platform to perform the content caching task; Step S4: obtaining real-time live network bandwidth data, and performing bit rate stream adaptive video encoding according to the real-time live network bandwidth data, obtaining a video quality strategy table, and uploading it to the cloud platform to execute the video encoding task; Step S5: Obtain the real-time resource allocation data of the cloud platform, and optimize the cloud resource configuration according to the real-time resource allocation data of the cloud platform, obtain the optimized cloud resource configuration data, and upload it to the cloud platform to execute the resource configuration task.

2. The cloud computing management method according to claim 1, characterized in that: Step S1 is specifically as follows: Step S11: acquiring user activity data, and performing data preprocessing on the user activity data to obtain user activity data to be analyzed; Step S12: extracting user activity features from the user activity data to obtain user activity time data and user live broadcast interaction data; Step S13: dividing the user activity time periodically according to the user activity time data to obtain the user activity time periodicity data; Step S14: performing user activity pattern recognition based on the user activity time periodicity data and the user live broadcast interaction data to obtain user activity pattern data; Step S15: construct a live broadcast traffic prediction model based on the user activity pattern data, and use the live broadcast traffic prediction model to predict the traffic during the user's active time period to obtain live broadcast predicted traffic data.

3. The cloud computing management method according to claim 2, characterized in that: Step S14 is specifically as follows: Step S141: performing interaction type weight classification on user live broadcast interaction data to obtain live broadcast interaction type weight data; Step S142: evaluating the user live broadcast participation degree of the user live broadcast interaction data according to the live broadcast interaction type weight data to obtain the user live broadcast participation degree data; Step S143: clustering the user's live broadcast type preference according to the live broadcast participation data to obtain the user's live broadcast type preference data; Step S144: performing user participation in live broadcast type time association based on the user activity time periodicity data and the user live broadcast type preference data to obtain live broadcast preference time data; Step S145: identifying the user's live broadcast activity participation behavior pattern according to the live broadcast preference time data to obtain live broadcast participation behavior pattern data; Step S146: assigning behavior pattern labels to the live broadcast participation behavior pattern data to obtain user activity pattern data.

4. The cloud computing management method according to claim 1, characterized in that: Step S2 is specifically as follows: Step S21: Obtain cloud platform resource allocation data, and perform data preprocessing on the cloud platform resource allocation data to obtain resource allocation data to be analyzed; Step S22: estimating resource demand during the live broadcast activity according to the live broadcast predicted traffic data, and obtaining resource demand data during the live broadcast; Step S23: determining the node transmission boundary conditions of the cloud platform according to the resource allocation data to be analyzed, and obtaining the node transmission boundary conditions; Step S24: matching the data transmission resource demand with the resource demand data during the live broadcast based on the node transmission boundary conditions, and obtaining transmission computing resource matching data, transmission bandwidth resource matching data, and transmission storage resource matching data; Step S25: setting resource automatic expansion strategy for the transmission computing resource matching data, the transmission bandwidth resource matching data and the transmission storage resource matching data to obtain a resource allocation configuration table.

5. The cloud computing management method according to claim 4, characterized in that: Step S25 is specifically as follows: According to a preset resource division ratio, reserved resource division is performed on the transmission computing resource matching data, the transmission bandwidth resource matching data, and the transmission storage resource matching data to obtain reserved resource data, to-be-allocated computing resource data, to-be-allocated bandwidth resource data, and to-be-allocated storage resource data; According to the resource demand data during the live broadcast, the live broadcast resource load threshold is set in steps to obtain a live broadcast resource extension step threshold set, wherein the live broadcast resource extension step threshold set includes a computing resource step extension threshold, a bandwidth resource step extension threshold, and a storage resource step extension threshold; Based on the computing resource step expansion threshold, computing resource expansion rules are set for the computing resource data to be allocated, and a computing resource expansion strategy is obtained; Based on the bandwidth resource step expansion threshold, bandwidth resource expansion rules are set for the bandwidth resource data to be allocated to obtain a bandwidth resource expansion strategy; Based on the storage resource step expansion threshold, storage resource expansion rules are set for the storage resource data to be allocated to obtain a storage resource expansion strategy; Set resource expansion dynamic elasticity factor according to node transmission boundary conditions; According to the resource expansion dynamic elasticity factor, a reserved resource calling rule is set for the reserved resource data to obtain a reserved resource calling strategy; Allocate expansion priorities for the reserved resource call strategy, the computing resource expansion strategy, the bandwidth resource expansion strategy, and the storage resource expansion strategy to obtain a resource expansion priority table; The resource expansion dynamic elasticity factor is mapped to the resource expansion priority table, and the resource expansion dynamic scheduling strategy is set to obtain the resource allocation configuration table.

6. The cloud computing management method according to claim 1, characterized in that: The content classification distribution of user live broadcast dependency in step S3 is specifically as follows: Extracting user activity features from the user activity pattern data to obtain user interaction pattern data and user activity time pattern data; According to the user live broadcast participation data, the user interaction mode data is divided into interaction mode participation levels to obtain high-participation interaction mode data and low-participation interaction mode data; Perform time distribution division according to user activity time pattern data to obtain continuous activity time pattern data and discrete activity time pattern data; Perform activity pattern intersection operations on high-engagement interaction pattern data and continuous activity time pattern data to obtain high-dependency activity pattern data; Perform activity pattern intersection operation on low-engagement interaction pattern data and continuous activity time pattern data to obtain low-dependency activity pattern data; Calculate the average dependency of the user's live content based on the high-dependency activity pattern data and the low-dependency activity pattern data to obtain the user's live content dependency data; Performing live content resource demand analysis based on the live broadcast predicted traffic data to obtain predicted live content resource demand data, and performing multi-level live content division on the predicted live content resource demand data based on the resource allocation configuration table to obtain live content division data; Based on the user live content dependency data, content-dependency matching is performed on the live content division data to obtain live content distribution data, wherein the live content distribution data includes user high dependency content distribution data and user low dependency content distribution data.

7. The cloud computing management method according to claim 6, characterized in that: The live content multi-preference layer cache in step S3 is specifically as follows: Design the cache hierarchy based on the cloud platform resource allocation data to obtain the edge cache layer, cloud cache layer, and storage layer; Use the edge cache layer to dynamically cache the live content edge node for the user's high-dependency content distribution data to obtain high-dependency content cache data; Using the cloud cache layer to dynamically cache the live content in the cloud for the user's low-dependency content distribution data, and obtain low-dependency content cache data; Calculate the cache preheating time during the traffic forecast peak period based on the live broadcast forecast traffic data to obtain the cache preheating time period data; According to the cache warm-up time period data, the low-dependency content cache data and the high-dependency content cache data are prioritized for high-dependency content caching to obtain a cache layer calling strategy; The storage layer is used to optimize the storage layer structure of the cache layer call strategy and the live content distribution data, obtain the cache storage optimization table, and upload it to the cloud platform to perform content caching tasks.

8. The cloud computing management method according to claim 1, characterized in that: Step S4 is specifically as follows: Step S41: acquiring real-time live network bandwidth data, and performing data preprocessing on the real-time live network bandwidth data to obtain real-time network bandwidth data to be analyzed; Step S42: Evaluate the network quality of the user equipment according to the real-time network bandwidth data to be analyzed, and obtain the network quality data of the user equipment; Step S43: setting a bit rate threshold for each user based on the user equipment network quality data to obtain user bit rate threshold data; Step S44: extracting the network bandwidth dynamic change characteristics from the real-time network bandwidth data to be analyzed, and obtaining the network bandwidth dynamic data; Step S45: Perform video adaptive encoding according to the network bandwidth dynamic data and the user bit rate threshold data, obtain a video quality strategy table, and upload it to the cloud platform to execute the video encoding task.

9. The cloud computing management method according to claim 8, characterized in that: Step S5 is specifically as follows: Step S51: acquiring cloud platform real-time resource allocation data, and performing data preprocessing on the cloud platform real-time resource allocation data to obtain real-time resource allocation data to be analyzed; Step S52: Calculate the node average usage rate and node response time according to the real-time resource allocation data to be analyzed, and evaluate the resource utilization rate based on the calculated node average usage rate and node response time to obtain real-time resource utilization rate data; Step S53: optimizing the low-utilization resource pool for the real-time resource allocation data to be analyzed based on the real-time resource utilization to obtain optimized resource pool allocation data; Step S54: Perform load balancing resource dynamic expansion optimization on the optimized resource pool allocation data to obtain optimized cloud resource configuration data, and upload it to the cloud platform to execute the resource configuration task.

10. A cloud computing management system, characterized in that: Used to execute the cloud computing management method according to claim 1, the cloud computing management system comprises: The traffic demand prediction module is used to obtain user activity data and predict the traffic demand during the user's active time period based on the user activity data to obtain live broadcast predicted traffic data; The resource demand elastic expansion module is used to obtain the cloud platform resource allocation data, and dynamically and elastically expand the cloud platform live broadcast resource demand according to the live broadcast predicted traffic data to obtain the resource allocation configuration table; The live content distribution cache module is used to classify and distribute the content according to the user's live broadcast dependency according to the resource allocation configuration table and the live broadcast predicted traffic data to obtain the live content distribution data; cache the live content distribution data in multiple preference layers to obtain the cache storage optimization table, and upload it to the cloud platform to perform the content caching task; The video adaptive encoding module is used to obtain the real-time live network bandwidth data, and perform bit rate stream adaptive video encoding according to the real-time live network bandwidth data, obtain the video quality strategy table, and upload it to the cloud platform to perform the video encoding task; The cloud resource configuration module is used to obtain the real-time resource allocation data of the cloud platform, optimize the cloud resource configuration according to the real-time resource allocation data of the cloud platform, obtain the optimized cloud resource configuration data, and upload it to the cloud platform to execute the resource configuration task.

Images