Method and system for optimizing content delivery network of OTT platform

Abstract

The application discloses an OTT platform content distribution network optimization method and system, relates to the wireless communication related technical field, and the method comprises the following steps: constructing an OTT service operation model according to a full-amount service operation characteristic set; performing dynamic node grading on the content distribution network; performing server deployment bottleneck prediction on the distribution network grading model, and adjusting the multi-level server deployment of the content distribution network; introducing a server deployment evaluation channel to perform deployment evaluation optimization; performing iterative neighborhood optimization on the deployment adjustment optimization space, and performing grading deployment optimization of the content distribution network. The technical problems of lacking global dynamic optimization capability, being difficult to cope with the content distribution network performance bottleneck and low resource utilization efficiency caused by the complex dynamic characteristics of OTT services in the prior art are solved, the technical effects of realizing dynamic and accurate grading deployment optimization of the content distribution network, improving network resource utilization efficiency, service quality and network operation stability are achieved.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, specifically to an OTT platform content delivery network optimization method and system. Background Technology

[0002] With the rapid development of internet video services, OTT (Over-The-Top) platforms have become the main channel for users to access audio and video content. OTT platforms provide content distribution services to users through the public internet, without relying on traditional cable TV or telecommunications operator networks, offering high flexibility and scalability. However, when facing massive user requests, high-concurrency traffic, and diverse business scenarios, the content delivery network (CDN) of OTT platforms may experience performance bottlenecks due to the spatiotemporal unevenness of user access behavior, dynamic changes in content popularity, and limited network resources, thus affecting user experience and service quality. Traditional content distribution uses a static node deployment strategy, lacking the ability to dynamically perceive and adaptively adjust to business operation characteristics, making it difficult to effectively cope with dynamic business loads and network environment changes. Furthermore, existing content delivery networks cannot dynamically classify and optimize server deployment based on business operation characteristics, affecting the operational stability of content delivery network services.

[0003] Therefore, current technologies lack global dynamic optimization capabilities, making it difficult to address the performance bottlenecks and low resource utilization efficiency of content delivery networks caused by the complex dynamic characteristics of OTT services. Summary of the Invention

[0004] This application provides an OTT platform content delivery network optimization method and system, which solves the technical problems of existing technologies that lack global dynamic optimization capabilities and are unable to cope with the performance bottlenecks and low resource utilization efficiency of content delivery networks caused by the complex dynamic characteristics of OTT services. It achieves the technical effect of realizing dynamic, precise, hierarchical deployment optimization of content delivery networks, improving network resource utilization efficiency, service quality, and network operation stability.

[0005] This application provides a method for optimizing the content delivery network of an OTT platform. The method includes: constructing an OTT service operation model based on the full service operation feature set of the OTT platform; dynamically classifying the nodes of the OTT platform's content delivery network to obtain a distribution network classification model, wherein the distribution network classification model includes a core node layer, a regional node layer, and an edge node layer; predicting server deployment bottlenecks in the distribution network classification model based on the OTT service operation model to establish a multi-level deployment bottleneck map; adjusting the server deployment of the content delivery network at multiple levels based on the multi-level deployment bottleneck map to establish a server deployment adjustment space; introducing a server deployment evaluation channel to perform deployment evaluation and optimization in the server deployment adjustment space to obtain a deployment adjustment optimization space; performing iterative neighborhood optimization in the deployment adjustment optimization space through the server deployment evaluation channel and the server deployment adjustment space to obtain a deployment adjustment optimization result; and performing hierarchical deployment optimization of the content delivery network based on the deployment adjustment optimization result.

[0006] In a possible implementation, the OTT platform content delivery network optimization method further performs the following processing: performing hierarchical state prediction on the distribution network hierarchy model based on the OTT service operation model to obtain a core state prediction sequence, a regional state prediction sequence, and an edge state prediction sequence; collecting server deployment features based on the distribution network hierarchy model to obtain a core deployment scheme, a regional deployment scheme, and an edge deployment scheme; performing multi-dimensional bottleneck prediction on the core deployment scheme based on the core state prediction sequence to obtain a core deployment bottleneck map; performing multi-dimensional bottleneck prediction on the regional deployment scheme based on the regional state prediction sequence to obtain a regional deployment bottleneck map; performing multi-dimensional bottleneck prediction on the edge deployment scheme based on the edge state prediction sequence to obtain an edge deployment bottleneck map; and combining the core deployment bottleneck map and the regional deployment bottleneck map to generate the multi-level deployment bottleneck map.

[0007] In a possible implementation, the OTT platform content delivery network optimization method further performs the following processing: fault tree learning is performed based on the historical server hardware bottleneck event set of the core node layer to obtain a core server hardware bottleneck inference model; the core state prediction sequence and the core deployment scheme are input into the core server hardware bottleneck inference model to obtain core hardware bottleneck paths; server software bottleneck inference is performed on the core deployment scheme based on the core state prediction sequence to obtain core software bottleneck paths; server fault bottleneck inference is performed on the core deployment scheme based on the core state prediction sequence to obtain core fault bottleneck paths; the core hardware bottleneck paths, the core software bottleneck paths, and the core fault bottleneck paths are organized to generate the core deployment bottleneck map.

[0008] In a possible implementation, the OTT platform content delivery network optimization method further performs the following processes: adjusting multiple server deployment parameters of the core deployment scheme based on the core deployment bottleneck map to obtain the core deployment adjustment space; adjusting multiple server deployment parameters of the regional deployment scheme based on the regional deployment bottleneck map to obtain the regional deployment adjustment space; adjusting multiple server deployment parameters of the edge deployment scheme based on the edge deployment bottleneck map to obtain the edge deployment adjustment space; and performing a multi-level deployment adjustment decision combination based on the core deployment adjustment space, the regional deployment adjustment space, and the edge deployment adjustment space to generate the server deployment adjustment space.

[0009] In a possible implementation, the OTT platform content delivery network optimization method further performs the following processing: extracting the g-th deployment adjustment strategy based on the server deployment adjustment space, where g is a positive integer; performing simulation optimization on the content delivery network according to the g-th deployment adjustment strategy to obtain g-th network optimization simulation data; activating the server deployment evaluation channel, where the server deployment evaluation channel's deployment evaluation multivariate indicators include service reliability, service scalability, and service security; inputting the g-th network optimization simulation data into the server deployment evaluation channel to obtain the g-th deployment evaluation sequence; if the g-th deployment evaluation sequence meets the deployment evaluation tolerance condition, adding the g-th deployment adjustment strategy to the deployment adjustment optimization space.

[0010] In a possible implementation, the OTT platform content delivery network optimization method further performs the following processing: weighting based on multi-dimensional deployment evaluation indicators to obtain a deployment quality evaluation coordinate system; performing deployment quality analysis on the deployment adjustment optimization space based on the deployment quality evaluation coordinate system to obtain the current deployment quality distribution; performing neighborhood analysis on the deployment adjustment optimization space based on the server deployment evaluation channel and the server deployment adjustment space, according to the current deployment quality distribution, to obtain a first deployment optimization neighborhood; performing deployment quality analysis on the first deployment optimization neighborhood based on the deployment quality evaluation coordinate system to obtain a first neighborhood deployment quality distribution; continuing to perform neighborhood analysis on the first deployment optimization neighborhood based on the server deployment evaluation channel and the server deployment adjustment space, according to the first neighborhood deployment quality distribution, until obtaining the f-th deployment optimization neighborhood, where f is a predetermined number of neighborhoods; performing global iterative optimization of deployment quality on the deployment adjustment optimization space, the first deployment optimization neighborhood, and up to the f-th deployment optimization neighborhood based on the deployment quality evaluation coordinate system to generate the deployment adjustment optimization result.

[0011] In a possible implementation, the OTT platform content delivery network optimization method further performs the following processes: optimizing the deployment quality of the deployment adjustment optimization space based on the current deployment quality distribution to obtain a current deployment optimization population; performing difference analysis on the server deployment adjustment space based on the current deployment optimization population to obtain a current deployment difference distribution; performing cross-mutation on the server deployment adjustment space based on the current deployment difference distribution to obtain a current deployment adjustment mutation space; and introducing the server deployment evaluation channel to perform deployment evaluation optimization on the current deployment adjustment mutation space to generate the first deployment optimization neighborhood.

[0012] In a possible implementation, the OTT platform content distribution network optimization method further performs the following processing: the full service operation feature set includes user geographic distribution features, user operator distribution features, user terminal access features, content popularity features, and network operation features.

[0013] In a possible implementation, the OTT platform content delivery network optimization method further performs the following processing: the deployment evaluation tolerance conditions include service reliability evaluation constraints, service expansion evaluation constraints, and service security evaluation constraints.

[0014] This application also provides an OTT platform content delivery network optimization system, the system comprising: a runtime model construction module, used to construct an OTT service runtime model based on the full service runtime feature set of the OTT platform; a hierarchical model acquisition module, used to dynamically hierarchize the OTT platform's content delivery network to obtain a hierarchical model, the hierarchical model including a core node layer, a regional node layer, and an edge node layer; a bottleneck prediction module, used to predict server deployment bottlenecks in the hierarchical model based on the OTT service runtime model, and establish a multi-level deployment bottleneck map; a server adjustment module, used to adjust the content delivery network's server deployment at multiple levels based on the multi-level deployment bottleneck map, and establish a server deployment adjustment space; a deployment evaluation and optimization module, used to introduce a server deployment evaluation channel to perform deployment evaluation and optimization in the server deployment adjustment space, and obtain a deployment adjustment optimization space; and a hierarchical deployment optimization module, used to perform iterative neighborhood optimization in the deployment adjustment optimization space through the server deployment evaluation channel and the server deployment adjustment space, obtain deployment adjustment optimization results, and execute hierarchical deployment optimization of the content delivery network based on the deployment adjustment optimization results.

[0015] This application proposes an OTT platform content delivery network (CDN) optimization method and system. The method constructs an OTT service operation model based on a full set of service operation characteristics; dynamically hierarchically classifies the CDN nodes; predicts server deployment bottlenecks in the hierarchical model and adjusts multi-level server deployment; introduces a server deployment evaluation channel for deployment evaluation and optimization; iteratively optimizes the deployment adjustment space to obtain the optimization results, and then performs hierarchical deployment optimization of the CDN. This addresses the technical problems in existing technologies, such as the lack of global dynamic optimization capabilities, making it difficult to cope with the performance bottlenecks and low resource utilization efficiency of the CDN caused by the complex dynamic characteristics of OTT services. It achieves dynamic and accurate hierarchical deployment optimization of the CDN, improving network resource utilization efficiency, service quality, and network operational stability. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a schematic diagram of a method for optimizing the content distribution network of an OTT platform, provided in an embodiment of this application.

[0018] Figure 2 This is a schematic diagram of an OTT platform content delivery network optimization system structure provided in an embodiment of this application.

[0019] Figure labeling: Model building module 10, hierarchical model acquisition module 20, bottleneck prediction module 30, server adjustment module 40, deployment evaluation and optimization module 50, hierarchical deployment optimization module 60. Detailed Implementation

[0020] To further illustrate the technical means and effects adopted by the present invention in order to achieve the intended purpose, the following detailed description is provided in conjunction with the accompanying drawings and preferred embodiments, based on the specific implementation methods, structures, features and effects of the present invention.

[0021] This application provides an embodiment of an OTT platform content delivery network optimization method, such as... Figure 1 As shown, the method includes: Step S100: Construct an OTT service operation model based on the full set of service operation characteristics of the OTT platform.

[0022] Step S100 further includes the following: the full set of service operation characteristics includes user geographic distribution characteristics, user operator distribution characteristics, user terminal access characteristics, content popularity characteristics, and network operation characteristics.

[0023] Preferably, multi-dimensional business data from OTT platforms is acquired, preprocessed through data mining, and then processed using a big data platform for ETL, including handling missing data, removing outliers, resolving user IPs to geographic locations, and unifying timestamps, to obtain a full set of business operation features, including user geographic distribution features, user operator distribution features, user terminal access features, content popularity features, and network operation features. Then, the features from the five dimensions are cross-correlated to create composite features with predictive value. A predictive model is built based on gradient boosting trees or time series models, and trained using historical data from the content delivery network. The model learns to predict target traffic, concurrent users, request types, etc., based on inputs such as time, content ID, and region, ultimately obtaining an OTT business operation model that can accurately quantify and predict the overall business load and user behavior patterns of the platform.

[0024] Preferably, user geographic distribution characteristics refer to the physical geographic location information and statistical patterns of users when accessing OTT services, such as user IP address range mapping, access heat distribution, and regional business profiles such as the total number of users, active users, and access duration within the region, to facilitate the deployment of regional and edge nodes in the content delivery network; user operator distribution characteristics refer to the network service provider information used by users when accessing the Internet, such as ISP affiliation, cross-network traffic analysis, and ISP network quality, to implement intra-network acceleration or multi-line BGP network optimization; user terminal access characteristics refer to the attributes of the devices and software environments used by users to play content, such as smartphones, tablets, smart TVs, PCs, set-top boxes, and operating systems such as iOS, Android, and Windows. The system adapts the player to supported encoding formats, screen resolutions, and bitrates to achieve accurate content distribution and transcoding strategies. Content popularity characteristics refer to quantitative indicators of the frequency and popularity of video content accessed within the platform, such as global / regional content popularity rankings, content lifecycle models, and personalized popularity, used to formulate caching strategies. Network operation characteristics refer to real-time operational status data of the OTT platform's own content distribution network and the public network, such as the load of content distribution network nodes, network quality indicators such as latency, packet loss rate, and bandwidth of each content distribution network node monitored by the user terminal or probe point, traffic peaks and trends, and the proportion of back-to-origin traffic. Among them, the proportion of back-to-origin traffic is the proportion of traffic from edge nodes to the next-level node or origin server to the total traffic, to achieve real-time traffic scheduling, load balancing, and capacity planning.

[0025] Step S200: Perform dynamic node classification on the content delivery network of the OTT platform to obtain a distribution network classification model, which includes a core node layer, a regional node layer, and an edge node layer.

[0026] Preferably, based on the constantly changing business demands output by the OTT service operation model and real-time network telemetry data, the content delivery network of the OTT platform is dynamically hierarchically classified. This involves redefining and defining the functions, capacity, and coverage of server nodes to form a hierarchical network structure. The OTT service operation model's prediction might indicate a surge in demand for content X from mobile users in region A within the next hour. Real-time network telemetry data refers to the CPU, memory, bandwidth, disk I / O, number of connections, and latency to users of each node. Specifically, tasks are allocated based on the node's physical capabilities (bandwidth, storage, etc.) and current performance. A policy engine evaluates the service level, service scope, and upstream / downstream relationships of each CDN node in real time, translating the evaluation results into specific configuration instructions. This forms a three-tiered topology of core, regional, and edge nodes, determining the distribution network hierarchy model. This transforms the physical network into an intelligent scheduling network capable of elastically scaling and adaptively optimizing according to business needs, thereby achieving the goals of reducing latency, increasing cache hit rate, and enhancing network robustness.

[0027] Preferably, the core node layer is the "brain" of the entire CDN network. It directly connects to the origin server of the OTT platform to pull and cache the most complete content library, and pushes and pre-caches popular content to lower-level nodes. At the same time, when edge and regional nodes fail to cache the content, the final back-to-origin traffic converges to the core node. The regional node layer is the "regional distribution center" that connects the top and bottom layers. It is responsible for aggregating user traffic in a large area. It receives user requests and back-to-origin traffic from multiple edge nodes within its coverage area. It stores a more curated and comprehensive content library than the core node and the edge node, usually containing the second most popular content across the entire network. The edge node layer stores all the most popular content within the region and performs load balancing within that region. When an edge node is overloaded or content is not found, requests can be scheduled among other nodes within the region. The edge node layer is a "retail store" directly facing end users, with the core goal of minimizing latency. It is used to provide direct user services, storing only the most popular content and highly personalized popular content for users within the node's coverage area. The cached content is dynamically updated based on local user behavior, and the node's load, health status, cache hit rate, user access latency, and other data are reported to the upper-layer scheduling center in real time.

[0028] Step S300: Based on the OTT service operation model, perform server deployment bottleneck prediction on the distribution network hierarchical model and establish a multi-level deployment bottleneck map.

[0029] Step S300 further includes step S310, performing hierarchical state prediction on the distribution network hierarchical model according to the OTT service operation model to obtain a core state prediction sequence, a regional state prediction sequence, and an edge state prediction sequence; step S320, collecting server deployment features according to the distribution network hierarchical model to obtain a core deployment scheme, a regional deployment scheme, and an edge deployment scheme; step S330, performing multi-dimensional bottleneck prediction on the core deployment scheme according to the core state prediction sequence to obtain a core deployment bottleneck map; step S340, performing multi-dimensional bottleneck prediction on the regional deployment scheme according to the regional state prediction sequence to obtain a regional deployment bottleneck map; step S350, performing multi-dimensional bottleneck prediction on the edge deployment scheme according to the edge state prediction sequence to obtain an edge deployment bottleneck map, and combining the core deployment bottleneck map and the regional deployment bottleneck map to generate the multi-level deployment bottleneck map.

[0030] Preferably, the OTT service operation model is used to predict server deployment bottlenecks in the distribution network hierarchical model. Specifically, the OTT service operation model is used to predict the hierarchical status of the distribution network hierarchical model to determine the service status that each layer of CDN nodes will face in the future, obtaining core status prediction sequences, regional status prediction sequences, and edge status prediction sequences. Among them, the core status prediction sequence is a numerical sequence of key indicators such as total origin return requests, cross-regional scheduling traffic, video transcoding computing load, and global storage I / O that the core node layer will face at multiple time points in the future. The regional status prediction sequence is a prediction of the future inbound traffic, outbound traffic, cache hit rate, and concurrent connection count of each regional node. The edge status prediction sequence is a prediction of the future user request QPS, outbound bandwidth, local cache hit rate, and average latency to users of each edge node.

[0031] Preferably, server deployment characteristics are collected based on the distribution network hierarchy model to determine the actual hardware, software, and network configuration of servers at each level in the current CDN network, thereby obtaining core deployment schemes, regional deployment schemes, and edge deployment schemes. The core deployment scheme describes the configuration parameters of the core node layer servers, such as CPU model / number of cores, memory capacity, storage type / capacity / IOPS, and network card bandwidth, as well as the architecture and redundancy scheme, including uplink bandwidth and connection method with the origin server. The regional / edge deployment scheme is similar to that of the core node layer, but edge nodes focus more on outbound bandwidth and cache capacity, while regional nodes may focus more on connection bandwidth with the core and edge and cache capacity within the region.

[0032] Preferably, multi-dimensional bottleneck prediction for server deployment is performed on the core deployment scheme based on the core state prediction sequence. This involves comparing each predicted value in the core state prediction sequence with the corresponding capability threshold in the core deployment scheme to generate a visualized, structured core deployment bottleneck map. The bottleneck type at each time point in the core layer is marked, and computational and storage I / O bottlenecks are also predicted, such as transcoding task queue backlogs or disk read / write delays. Similarly, multi-dimensional bottleneck prediction for server deployment is performed layer-by-layer and node-by-node for the regional / edge deployment scheme, obtaining regional / edge deployment bottleneck maps respectively. Finally, the core deployment bottleneck map and the regional / edge deployment bottleneck maps are correlated and integrated into a global view to reveal the correlation and transmission paths between bottlenecks, ultimately determining a multi-level deployment bottleneck map to display the causal relationships between all bottleneck points.

[0033] Furthermore, step S330 also includes step S331, performing fault tree learning based on the historical server hardware bottleneck event set of the core node layer to obtain a core server hardware bottleneck inference model; step S332, inputting the core state prediction sequence and the core deployment scheme into the core server hardware bottleneck inference model to obtain the core hardware bottleneck path; step S333, performing server software bottleneck inference on the core deployment scheme based on the core state prediction sequence to obtain the core software bottleneck path; step S334, performing server fault bottleneck inference on the core deployment scheme based on the core state prediction sequence to obtain the core fault bottleneck path; and step S335, organizing the core hardware bottleneck path, the core software bottleneck path, and the core fault bottleneck path to generate the core deployment bottleneck map.

[0034] Preferably, detailed records of all hardware-related performance failures or bottlenecks that have occurred in the core node layer are obtained to determine the historical server hardware bottleneck event set of the core node layer. Each record includes not only the result but also the root cause and intermediate process. A model is built based on decision trees, random forests, or Bayesian networks. Fault tree learning is performed using the historical server hardware bottleneck event set to analyze all possible paths leading to undesirable events such as server downtime. This allows the model to learn various patterns, rules, and probabilities that cause hardware issues. For example, when memory usage is >80% and a large number of small file random read operations occur simultaneously, there is an 85% probability that the disk I / O wait queue will saturate within 5 minutes, leading to service unavailability. A core server hardware bottleneck inference model with embedded fault tree rules is obtained, which can receive the current system state input and infer the most likely hardware bottleneck path along the learned fault chain.

[0035] Preferably, specific bottleneck scenarios that may occur in the future are deduced from three dimensions: hardware, software, and overall failure. Specifically, the core state prediction sequence and core deployment plan are input into the core server hardware bottleneck deduction model, and the model is used to infer and output a dynamic failure chain as the core hardware bottleneck path. For example, the predicted high network packet volume → exceeds the network card's processing capacity → causes a surge in CPU soft interrupts → preempts CPU resources of user-mode processes → ultimately leads to unresponsive scheduling services, even if the overall CPU utilization is not high. Similarly, the core deployment plan is used to perform server software bottleneck deduction based on the core state prediction sequence. That is, logical deduction is performed based on the characteristics and known defects of the software system. Among them, when the number of concurrent connections exceeds a certain threshold, the load balancing software triggers inefficient lock contention. The prediction sequence shows that the number of connections will exceed this threshold, thus obtaining the core software bottleneck path. For example, the number of concurrent connections exceeds 1.5 million → the load balancer enters a lock contention state → the average response time increases exponentially → although server resources are idle, the business throughput drops sharply.

[0036] Preferably, server failure bottlenecks in the core deployment scheme are simulated based on the core state prediction sequence, with a greater focus on system redundancy and reliability. Hypothetical analysis combined with predicted load simulations is used to model the impact of single component failures. For example, it is hypothesized that during peak business hours, a core cache server fails and goes down → traffic is redistributed to other servers → the remaining servers bear a load exceeding their design capacity → triggering a cascading failure effect, leading to the paralysis of the entire core layer service. Finally, the core hardware bottleneck path, core software bottleneck path, and core failure bottleneck path are integrated, including spatiotemporal alignment, correlation analysis, and visualization / structured coding, to determine a structured core deployment bottleneck map. This map depicts the hardware and software configurations that ultimately evolve into system bottlenecks along different "paths" of load pressure.

[0037] Step S400: Adjust the multi-level server deployment of the content delivery network according to the multi-level deployment bottleneck map to establish a server deployment adjustment space.

[0038] Step S400 further includes step S410, adjusting multiple server deployment parameters of the core deployment scheme according to the core deployment bottleneck map to obtain the core deployment adjustment space; step S420, adjusting multiple server deployment parameters of the regional deployment scheme according to the regional deployment bottleneck map to obtain the regional deployment adjustment space; step S430, adjusting multiple server deployment parameters of the edge deployment scheme according to the edge deployment bottleneck map to obtain the edge deployment adjustment space; and step S440, performing a multi-level deployment adjustment decision combination based on the core deployment adjustment space, the regional deployment adjustment space, and the edge deployment adjustment space to generate the server deployment adjustment space.

[0039] Preferably, the content delivery network (CDN) is adjusted at multiple levels based on a multi-level deployment bottleneck map. Specifically, multiple parameters of server deployment are adjusted for the core deployment scheme, regional deployment scheme, and edge deployment scheme, including hardware resource parameters such as bandwidth, computing, storage, and memory; software strategy parameters such as caching strategy, scheduling strategy, and transmission protocol; and topological redundancy parameters such as the number of nodes and redundancy level. This yields core deployment adjustment space, regional deployment adjustment space, and edge deployment adjustment space. Each adjustment space contains multiple feasible adjustment schemes within each level. Then, multi-level deployment adjustment decisions are combined based on the core deployment adjustment space, regional deployment adjustment space, and edge deployment adjustment space. This involves globally coordinating the independent solutions at each level to generate a global, collaborative optimization strategy. Since the adjustment scheme decisions at the three levels influence each other, the adjustment spaces at the three levels are combined using a Cartesian product. This involves selecting multiple adjustment strategies from the core deployment adjustment space, regional deployment adjustment space, and edge deployment adjustment space, and filtering them based on business logic to form a global deployment strategy, generating a server deployment adjustment space containing multiple feasible, global, and multi-level collaborative deployment adjustment strategies.

[0040] Step S500: Introduce a server deployment evaluation channel to perform deployment evaluation and optimization on the server deployment adjustment space to obtain the deployment adjustment optimization space.

[0041] Step S500 further includes step S510, extracting the g-th deployment adjustment strategy according to the server deployment adjustment space, where g is a positive integer; step S520, performing simulation optimization on the content delivery network according to the g-th deployment adjustment strategy to obtain g-th network optimization simulation data; step S530, activating the server deployment evaluation channel, where the server deployment evaluation channel's deployment evaluation multivariate indicators include service reliability, service scalability, and service security; step S540, inputting the g-th network optimization simulation data into the server deployment evaluation channel to obtain the g-th deployment evaluation sequence; step S550, if the g-th deployment evaluation sequence meets the deployment evaluation tolerance condition, adding the g-th deployment adjustment strategy to the deployment adjustment optimization space.

[0042] Preferably, a server deployment evaluation channel is introduced to evaluate and optimize the server deployment adjustment space. Specifically, an arbitrary deployment adjustment strategy is extracted from the server deployment adjustment space as the g-th deployment adjustment strategy, where g is a positive integer representing the current strategy count being evaluated. The g-th deployment adjustment strategy is input into the CDN simulation environment to simulate and optimize the content delivery network. The CDN simulation environment has a built-in OTT service operation model, network topology, and traffic model. The service is simulated to run for a period of time in the CDN simulation environment, and the changes described by the g-th deployment adjustment strategy are applied. The g-th network optimization simulation data is output, which includes various indicators reflecting CDN performance, such as the CPU, memory, disk I / O, and network bandwidth utilization of each node, the cache hit rate and origin pull rate of the entire network, the average latency, stuttering rate, and error rate of user requests, and the distribution of network traffic.

[0043] Preferably, the server deployment evaluation channel is activated to quantitatively score the simulation results. The server deployment evaluation channel includes multiple indicators such as service reliability, service scalability, and service security. Specifically, service reliability measures the stability and continuity of the service, and specific indicators may include service availability, request error rate, and mean time between failures. Low error rates and low latency in the simulation data will result in high scores in this area. Service scalability measures the system's ability to cope with load growth and resource utilization efficiency, and specific indicators may include resource utilization, elastic scaling capability, and service throughput per unit cost. The simulation data shows balanced resource usage without bottlenecks, resulting in high scores in this area. Service security measures the system's ability to resist security threats, and specific indicators may include DDoS protection capability, data encryption strength, and the number of security vulnerabilities. The simulation data can be combined with security policy checks to assess whether the deployment strategy will introduce new security risks.

[0044] Preferably, the server deployment evaluation channel is a multi-task learning regression / classification model that receives network optimization simulation data as input and outputs a quantitative evaluation of service reliability, scalability, and security. The structure consists of an input layer, a core processing layer, and an output decision layer. The input layer is used for feature vectorization, extracting statistical features from the network optimization simulation time-series data using a sliding window, encoding these features, and then standardizing and concatenating them into a high-dimensional feature vector. The core processing layer learns to extract general deep representations useful for all three evaluation objectives from the high-dimensional feature vectors. It contains three independent neural network branches, each optimizing one evaluation metric and learning to map it to its respective metric's score. The output decision layer outputs the deployment evaluation sequence and compares it with preset deployment evaluation tolerance conditions, generating a Boolean decision, including pass / fail. Specifically, simulations of numerous different deployment adjustment strategies are run in historical or synthetic CDN environments. The system collects network optimization simulation data from each simulation, and then domain experts label the data with reliability tags based on service availability, error rate, and fault recovery time. It also labels the data with scalability tags based on resource utilization curves, bottleneck thresholds, and performance degradation under increased load. Finally, it labels the data with security tags based on attack penetration test results, configuration compliance checks, and potential vulnerability scan reports from the simulation. A weighted multi-task loss function is used to divide the dataset into training, validation, and test sets in an 8:1:1 ratio. Using the training set data, the model parameters are optimized through backpropagation to minimize the total loss. On the validation set, the prediction accuracy of each task is monitored, and an early stopping strategy is used to prevent overfitting. Finally, the model's final generalization performance is evaluated on the test set to obtain the final server deployment evaluation channel. This determines the intelligent evaluator, which uses simulation data as input, multi-dimensional expert ratings as supervision signals, and is trained using a multi-task deep learning framework.

[0045] Preferably, the network optimization simulation data of the g-th network is input into the server deployment evaluation channel, and a structured deployment evaluation sequence of the g-th network is output, which includes the score of the strategy on multiple evaluation indicators. The deployment evaluation tolerance condition is a preset minimum admission standard, such as a tolerance condition of reliability greater than 90, scalability greater than 80, and security greater than 85. The deployment evaluation sequence of the g-th network is compared with the tolerance condition, and the deployment strategy that meets the minimum standard for all indicators is regarded as a valuable candidate solution. Finally, it is added to the deployment adjustment optimization space, which contains multiple elite strategies that have passed the initial screening and have been proven to be effective in the simulation environment, thereby significantly improving the efficiency of the overall optimization process and the quality of the final result.

[0046] Step S600: Iterative neighborhood optimization is performed on the deployment adjustment optimization space through the server deployment evaluation channel and the server deployment adjustment space to obtain the deployment adjustment optimization result, and the hierarchical deployment optimization of the content delivery network is performed according to the deployment adjustment optimization result.

[0047] Step S600 further includes step S610, configuring weights according to the multi-dimensional deployment evaluation indicators to obtain a deployment quality evaluation coordinate system; step S620, performing deployment quality analysis on the deployment adjustment optimization space according to the deployment quality evaluation coordinate system to obtain the current deployment quality distribution; step S630, performing neighborhood analysis on the deployment adjustment optimization space based on the server deployment evaluation channel and the server deployment adjustment space, according to the current deployment quality distribution, to obtain a first deployment optimization neighborhood; step S640, performing deployment quality analysis on the first deployment optimization neighborhood based on the deployment quality evaluation coordinate system to obtain a first neighborhood deployment quality distribution; step S650, continuing to perform neighborhood analysis on the first deployment optimization neighborhood based on the server deployment evaluation channel and the server deployment adjustment space, according to the first neighborhood deployment quality distribution, until obtaining the f-th deployment optimization neighborhood, where f is a predetermined number of neighborhoods; step S660, performing global iterative optimization of deployment quality on the deployment adjustment optimization space, the first deployment optimization neighborhood, up to the f-th deployment optimization neighborhood, according to the deployment quality evaluation coordinate system to generate the deployment adjustment optimization result.

[0048] Preferably, weights are assigned to the three deployment evaluation indicators of service reliability, scalability, and security based on business priorities. For example, for financial live streaming, which has extremely high stability requirements, the reliability weight is set very high. Each deployment adjustment strategy corresponds to a point in this three-dimensional space, representing the deployment quality coefficient, which is then abstracted into a deployment quality evaluation coordinate system. Deployment quality analysis is performed on the deployment adjustment optimization space based on the deployment quality evaluation coordinate system. That is, all strategies in the deployment adjustment optimization space are plotted according to their positions in the coordinate system, their distribution is analyzed, the clustered and sparse regions of the current high-quality solutions are identified, the Pareto front is determined, and the current deployment quality distribution is obtained, which includes the deployment quality coefficient corresponding to each deployment adjustment strategy in the deployment adjustment optimization space.

[0049] Preferably, the deployment adjustment optimization space is analyzed based on the current deployment quality distribution. Specifically, a batch of high-quality strategies or those located at the Pareto front are selected from the current deployment quality distribution as seed strategies. A neighborhood is defined for each seed strategy in the server deployment adjustment space. The neighborhood of a strategy refers to the set of all new strategies that can be generated by applying one or more small, feasible adjustment actions, such as adjusting the bandwidth from 40G to 50G or replacing a software parameter. In this way, a large number of new candidate strategies are generated within the neighborhood of the seed strategies. The server deployment evaluation channel is called to evaluate these new strategies and obtain their service reliability, scalability, and security scores. Then, from all the evaluated new strategies, strategies that significantly improve quality or enrich the Pareto front are selected to form the first deployment optimization neighborhood, which contains multiple potential deployment adjustment strategies that may be better than the initial optimization space.

[0050] Preferably, the deployment quality of the first deployment optimization neighborhood is analyzed using the deployment quality evaluation coordinate system to obtain the deployment quality distribution of the first neighborhood. Then, based on the deployment quality distribution of the first neighborhood, the neighborhood of the first deployment optimization neighborhood is analyzed again. Then, using its high-quality strategy as a seed, the exploration and evaluation are carried out again in the neighborhood of its respective adjustment action to generate the second deployment optimization neighborhood. This process is repeated until the f-th deployment optimization neighborhood is obtained, where f is a positive integer representing the predetermined number of neighborhoods. After the iterative search is completed, the preset optimization goal is to achieve the highest overall quality. The global iterative optimization of deployment quality is carried out on the deployment adjustment optimization space, the first deployment optimization neighborhood, and up to the f-th deployment optimization neighborhood using the deployment quality evaluation coordinate system. The best quality is selected as the deployment adjustment optimization result according to the preset optimization goal to obtain the best or near-best deployment scheme, thereby efficiently and intelligently finding the optimal solution to the complex CDN network deployment problem.

[0051] Preferably, the deployment and optimization results are finally transformed into executable instructions for each layer and automatically applied to the actual physical CDN network. These instructions include core layer, regional layer, and edge layer instructions, which are then executed in the order of core layer → regional layer → edge layer. Within the same layer, a batch rolling update strategy is adopted. For example, 10% of the edge nodes are upgraded first, and after observing the stability and effect, the scale is gradually expanded. Specifically, the cloud platform API is called to perform server scaling up, scaling down, and configuration changes. BGP policies, ACL rules, and traffic scheduling weights are adjusted through the SDN controller or router command line. Service images are updated on a rolling basis through the container orchestration platform or new configuration files are issued through the configuration management center. While physical changes are being executed, the configuration management database, monitoring and alarm system, and traffic scheduling system are updated simultaneously. After the changes are executed, the system automatically enters the verification phase, which includes health checks on all changed services, comparison of key performance indicators before and after the changes, verification of whether the expected optimization goals have been achieved, and the real running data of the verification phase is added to the full service running feature set. This achieves dynamic, intelligent, and closed-loop optimization of network resources, improving the security and reliability of the environment.

[0052] Furthermore, step S630 also includes step S631, performing deployment quality optimization on the deployment adjustment optimization space based on the current deployment quality distribution to obtain a current deployment optimization population; step S632, performing difference analysis on the server deployment adjustment space based on the current deployment optimization population to obtain a current deployment difference distribution; step S633, performing crossover mutation on the server deployment adjustment space based on the current deployment difference distribution to obtain a current deployment adjustment mutation space; and step S634, introducing the server deployment evaluation channel to perform deployment evaluation optimization on the current deployment adjustment mutation space to generate the first deployment optimization neighborhood.

[0053] Preferably, based on the current deployment quality distribution, deployment adjustment strategies below the deployment quality threshold are removed from the deployment adjustment optimization space, a batch of high-quality strategies are selected to determine the current deployment optimization population, the parent strategies for subsequent genetic operations, and then the server deployment adjustment space is analyzed for differences based on the current deployment optimization population. Each deployment strategy is treated as a multi-dimensional parameter vector, and the statistical distribution of these parameter vectors in the population is calculated, such as mean / mode, variance / range, to obtain the current deployment difference distribution, which describes the statistical distribution of each deployment parameter in the high-quality population. Then, two parent strategies are randomly selected and some of their parameters are exchanged to generate offspring strategies. Then, some parameters of a single strategy are randomly perturbed, and a large number of new strategies are generated through crossover and mutation operations, forming the current deployment adjustment mutation space. Finally, a server deployment evaluation channel is introduced to perform deployment evaluation optimization on the current deployment adjustment mutation space, that is, the server deployment evaluation channel is called to simulate and evaluate all strategies in this mutation space, and then, based on the deployment quality evaluation coordinate system, deployment strategies that meet the deployment evaluation tolerance conditions are selected to generate the first deployment optimization neighborhood.

[0054] Preferred, In the above text, refer to Figure 1 A method for optimizing an OTT platform content delivery network according to an embodiment of the present invention is described in detail. Next, reference will be made to... Figure 2 This invention describes an OTT platform content delivery network optimization system according to an embodiment of the present invention.

[0055] An OTT platform content delivery network optimization system according to an embodiment of the present invention addresses the technical problems in existing technologies, such as the lack of global dynamic optimization capabilities, difficulty in coping with the complex dynamic characteristics of OTT services leading to performance bottlenecks and low resource utilization efficiency in content delivery networks. It achieves the technical effect of realizing dynamic, precise, hierarchical deployment optimization of the content delivery network, improving network resource utilization efficiency, service quality, and network operational stability. Figure 2 As shown, an OTT platform content delivery network optimization system includes: a runtime model construction module 10, a hierarchical model acquisition module 20, a bottleneck prediction module 30, a server adjustment module 40, a deployment evaluation and optimization module 50, and a hierarchical deployment optimization module 60.

[0056] The system comprises the following modules: a model construction module 10, which constructs an OTT service operation model based on the full service operation feature set of the OTT platform; a hierarchical model acquisition module 20, which dynamically hierarchically classifies the content delivery network of the OTT platform to obtain a hierarchical model of the distribution network, which includes a core node layer, a regional node layer, and an edge node layer; a bottleneck prediction module 30, which predicts server deployment bottlenecks in the hierarchical model of the distribution network based on the OTT service operation model and establishes a multi-level deployment bottleneck map; a server adjustment module 40, which adjusts the server deployment of the content delivery network at multiple levels based on the multi-level deployment bottleneck map and establishes a server deployment adjustment space; a deployment evaluation and optimization module 50, which introduces a server deployment evaluation channel to perform deployment evaluation and optimization in the server deployment adjustment space and obtains a deployment adjustment optimization space; and a hierarchical deployment optimization module 60, which iteratively optimizes the deployment adjustment optimization space through the server deployment evaluation channel and the server deployment adjustment space to obtain deployment adjustment optimization results and performs hierarchical deployment optimization of the content delivery network based on the deployment adjustment optimization results.

[0057] The specific configuration of the bottleneck prediction module 30 will be described in detail below. The bottleneck prediction module 30 further includes: performing hierarchical state prediction on the distribution network hierarchical model based on the OTT service operation model to obtain a core state prediction sequence, a regional state prediction sequence, and an edge state prediction sequence; collecting server deployment features based on the distribution network hierarchical model to obtain a core deployment scheme, a regional deployment scheme, and an edge deployment scheme; performing multi-dimensional bottleneck prediction on the core deployment scheme based on the core state prediction sequence to obtain a core deployment bottleneck map; performing multi-dimensional bottleneck prediction on the regional deployment scheme based on the regional state prediction sequence to obtain a regional deployment bottleneck map; performing multi-dimensional bottleneck prediction on the edge deployment scheme based on the edge state prediction sequence to obtain an edge deployment bottleneck map; and combining the core deployment bottleneck map and the regional deployment bottleneck map to generate the multi-level deployment bottleneck map.

[0058] The specific configuration of the bottleneck prediction module 30 will be described in detail below. The bottleneck prediction module 30 further includes: performing fault tree learning based on the historical server hardware bottleneck event set of the core node layer to obtain a core server hardware bottleneck inference model; inputting the core state prediction sequence and the core deployment scheme into the core server hardware bottleneck inference model to obtain core hardware bottleneck paths; performing server software bottleneck inference on the core deployment scheme based on the core state prediction sequence to obtain core software bottleneck paths; performing server fault bottleneck inference on the core deployment scheme based on the core state prediction sequence to obtain core fault bottleneck paths; and organizing the core hardware bottleneck paths, the core software bottleneck paths, and the core fault bottleneck paths to generate the core deployment bottleneck map.

[0059] The specific configuration of the server adjustment module 40 will be described in detail below. The server adjustment module 40 further includes: adjusting multiple server deployment parameters of the core deployment scheme based on a core deployment bottleneck map to obtain a core deployment adjustment space; adjusting multiple server deployment parameters of the regional deployment scheme based on a regional deployment bottleneck map to obtain a regional deployment adjustment space; adjusting multiple server deployment parameters of the edge deployment scheme based on an edge deployment bottleneck map to obtain an edge deployment adjustment space; and performing a multi-level deployment adjustment decision combination based on the core deployment adjustment space, the regional deployment adjustment space, and the edge deployment adjustment space to generate the server deployment adjustment space.

[0060] The specific configuration of the deployment evaluation and optimization module 50 will be described in detail below. The deployment evaluation and optimization module 50 further includes: extracting the g-th deployment adjustment strategy from the server deployment adjustment space, where g is a positive integer; performing simulated optimization on the content delivery network according to the g-th deployment adjustment strategy to obtain g-th network optimization simulation data; activating the server deployment evaluation channel, where the deployment evaluation multi-dimensional indicators include service reliability, service scalability, and service security; inputting the g-th network optimization simulation data into the server deployment evaluation channel to obtain the g-th deployment evaluation sequence; and if the g-th deployment evaluation sequence meets the deployment evaluation tolerance conditions, adding the g-th deployment adjustment strategy to the deployment adjustment optimization space.

[0061] The specific configuration of the hierarchical deployment optimization module 60 will be described in detail below. The hierarchical deployment optimization module 60 further includes: configuring weights based on multi-dimensional deployment evaluation indicators to obtain a deployment quality evaluation coordinate system; performing deployment quality analysis on the deployment adjustment optimization space based on the deployment quality evaluation coordinate system to obtain the current deployment quality distribution; performing neighborhood analysis on the deployment adjustment optimization space based on the server deployment evaluation channel and the server deployment adjustment space, according to the current deployment quality distribution, to obtain a first deployment optimization neighborhood; performing deployment quality analysis on the first deployment optimization neighborhood based on the deployment quality evaluation coordinate system to obtain a first neighborhood deployment quality distribution; continuing to perform neighborhood analysis on the first deployment optimization neighborhood based on the server deployment evaluation channel and the server deployment adjustment space, according to the first neighborhood deployment quality distribution, until obtaining the f-th deployment optimization neighborhood, where f is a predetermined number of neighborhoods; performing global iterative optimization of deployment quality on the deployment adjustment optimization space, the first deployment optimization neighborhood, and up to the f-th deployment optimization neighborhood based on the deployment quality evaluation coordinate system to generate the deployment adjustment optimization result.

[0062] The specific configuration of the hierarchical deployment optimization module 60 will be described in detail below. The hierarchical deployment optimization module 60 further includes: performing deployment quality optimization on the deployment adjustment optimization space based on the current deployment quality distribution to obtain a current deployment optimization population; performing difference analysis on the server deployment adjustment space based on the current deployment optimization population to obtain a current deployment difference distribution; performing cross-mutation on the server deployment adjustment space based on the current deployment difference distribution to obtain a current deployment adjustment mutation space; and introducing the server deployment evaluation channel to perform deployment evaluation optimization on the current deployment adjustment mutation space to generate the first deployment optimization neighborhood.

[0063] The specific configuration of the operation model construction module 10 will be described in detail below. The operation model construction module 10 further includes: the full service operation feature set includes user geographic distribution features, user operator distribution features, user terminal access features, content popularity features, and network operation features.

[0064] The specific configuration of the deployment evaluation and optimization module 50 will be described in detail below. The deployment evaluation and optimization module 50 further includes: the deployment evaluation tolerance conditions include service reliability evaluation constraints, service scalability evaluation constraints, and service security evaluation constraints.

[0065] The OTT platform content distribution network optimization system provided in this embodiment of the invention can execute the OTT platform content distribution network optimization method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0066] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for optimizing the content delivery network of an OTT platform, characterized in that, include: Based on the full set of operational characteristics of OTT platforms, construct an OTT service operation model; The content delivery network of the OTT platform is dynamically hierarchically classified to obtain a distribution network hierarchical model, which includes a core node layer, a regional node layer, and an edge node layer. Based on the OTT service operation model, server deployment bottlenecks are predicted for the distribution network hierarchical model, and a multi-level deployment bottleneck map is established. Based on the multi-level deployment bottleneck map, the content delivery network is adjusted in a multi-level server deployment manner to establish a server deployment adjustment space. A server deployment evaluation channel is introduced to evaluate and optimize the server deployment adjustment space, thereby obtaining the deployment adjustment optimization space. The deployment adjustment optimization space is iteratively optimized using the server deployment evaluation channel and the server deployment adjustment space to obtain the deployment adjustment optimization result, and the hierarchical deployment optimization of the content delivery network is performed based on the deployment adjustment optimization result.

2. The OTT platform content delivery network optimization method as described in claim 1, characterized in that, Based on the OTT service operation model, server deployment bottlenecks are predicted for the distribution network hierarchical model, and a multi-level deployment bottleneck map is established, including: Based on the OTT service operation model, the distribution network hierarchical model is used to perform hierarchical state prediction to obtain core state prediction sequence, regional state prediction sequence and edge state prediction sequence; Based on the distribution network hierarchical model, server deployment characteristics are collected to obtain core deployment schemes, regional deployment schemes, and edge deployment schemes. Based on the core state prediction sequence, perform multi-dimensional bottleneck prediction of server deployment for the core deployment scheme to obtain a core deployment bottleneck map; Based on the regional state prediction sequence, perform multi-dimensional bottleneck prediction of server deployment for the regional deployment scheme to obtain a regional deployment bottleneck map; Based on the edge state prediction sequence, perform multi-dimensional bottleneck prediction on server deployment of the edge deployment scheme to obtain an edge deployment bottleneck map. Combine the core deployment bottleneck map and the regional deployment bottleneck map to generate the multi-level deployment bottleneck map.

3. The OTT platform content delivery network optimization method as described in claim 2, characterized in that, Based on the core state prediction sequence, multi-dimensional bottleneck prediction of server deployment is performed on the core deployment scheme to obtain a core deployment bottleneck map, including: Based on the historical server hardware bottleneck event set of the core node layer, fault tree learning is performed to obtain the core server hardware bottleneck inference model. The core state prediction sequence and the core deployment scheme are input into the core server hardware bottleneck inference model to obtain the core hardware bottleneck path. Based on the core state prediction sequence, server software bottlenecks are simulated for the core deployment scheme to obtain the core software bottleneck path. Based on the core state prediction sequence, server failure bottlenecks are simulated for the core deployment scheme to obtain the core failure bottleneck path. The core hardware bottleneck paths, the core software bottleneck paths, and the core fault bottleneck paths are organized to generate the core deployment bottleneck map.

4. The OTT platform content delivery network optimization method as described in claim 1, characterized in that, Based on the multi-level deployment bottleneck map, the content delivery network is adjusted in a multi-level server deployment manner to establish a server deployment adjustment space, including: Based on the core deployment bottleneck map, adjust multiple parameters of the server deployment in the core deployment scheme to obtain the core deployment adjustment space; Based on the regional deployment bottleneck map, adjust multiple parameters of the server deployment in the regional deployment plan to obtain the regional deployment adjustment space; Based on the edge deployment bottleneck map, the server deployment parameters of the edge deployment scheme are adjusted to obtain the adjustment space for edge deployment; The server deployment adjustment space is generated by combining multi-level deployment adjustment decisions based on the core deployment adjustment space, the regional deployment adjustment space, and the edge deployment adjustment space.

5. The OTT platform content delivery network optimization method as described in claim 1, characterized in that, A server deployment evaluation channel is introduced to evaluate and optimize the server deployment adjustment space, thereby obtaining the deployment adjustment optimization space, including: The g-th deployment adjustment strategy is extracted based on the server deployment adjustment space, where g is a positive integer; The content delivery network is simulated and optimized according to the g-th deployment adjustment strategy to obtain the g-th network optimization simulation data. Activate the server deployment evaluation channel, wherein the server deployment evaluation multi-dimensional indicators include service reliability, service scalability, and service security; The g-th network optimization simulation data is input into the server deployment evaluation channel to obtain the g-th deployment evaluation sequence; If the g-th deployment evaluation sequence satisfies the deployment evaluation tolerance condition, the g-th deployment adjustment strategy is added to the deployment adjustment optimization space.

6. The OTT platform content delivery network optimization method as described in claim 1, characterized in that, The deployment adjustment optimization space is iteratively optimized using the server deployment evaluation channel and the server deployment adjustment space to obtain the deployment adjustment optimization result, including: We obtain a deployment quality evaluation coordinate system by configuring weights based on multiple deployment evaluation indicators. Based on the deployment quality evaluation coordinate system, the deployment quality is analyzed in the deployment adjustment and optimization space to obtain the current deployment quality distribution; Based on the server deployment evaluation channel and the server deployment adjustment space, the deployment adjustment optimization space is analyzed according to the current deployment quality distribution to obtain the first deployment optimization neighborhood. Based on the deployment quality evaluation coordinate system, the deployment quality of the first deployment optimization neighborhood is analyzed to obtain the deployment quality distribution of the first neighborhood. Based on the server deployment evaluation channel and the server deployment adjustment space, the first deployment optimization neighborhood is further parsed according to the first neighborhood deployment quality distribution until the f-th deployment optimization neighborhood is obtained, where f is the predetermined number of neighborhoods. Based on the deployment quality evaluation coordinate system, a global iterative optimization of deployment quality is performed on the deployment adjustment optimization space, the first deployment optimization neighborhood, and up to the f-th deployment optimization neighborhood to generate the deployment adjustment optimization result.

7. The OTT platform content delivery network optimization method as described in claim 6, characterized in that, Based on the server deployment evaluation channel and the server deployment adjustment space, a neighborhood analysis is performed on the deployment adjustment optimization space according to the current deployment quality distribution to obtain a first deployment optimization neighborhood, including: Based on the current deployment quality distribution, the deployment adjustment optimization space is used to optimize the deployment quality and obtain the current deployment optimization population. Based on the current deployment optimization population, the server deployment adjustment space is analyzed for differences to obtain the current deployment difference distribution; Based on the current deployment difference distribution, the server deployment adjustment space is cross-mutated to obtain the current deployment adjustment mutation space; The server deployment evaluation channel is introduced to perform deployment evaluation and optimization on the current deployment adjustment variation space, and the first deployment optimization neighborhood is generated.

8. The OTT platform content delivery network optimization method as described in claim 1, characterized in that, The full set of service operation characteristics includes user geographic distribution characteristics, user operator distribution characteristics, user terminal access characteristics, content popularity characteristics, and network operation characteristics.

9. The OTT platform content delivery network optimization method as described in claim 5, characterized in that, The deployment evaluation tolerance conditions include service reliability evaluation constraints, service scalability evaluation constraints, and service security evaluation constraints.

10. An OTT platform content delivery network optimization system, characterized in that, The system is used to implement the OTT platform content delivery network optimization method according to any one of claims 1 to 9, the system comprising: The runtime model building module is used to build OTT service runtime models based on the full set of service runtime characteristics of the OTT platform. The hierarchical model acquisition module is used to dynamically hierarchize the content distribution network of the OTT platform to obtain a distribution network hierarchical model, which includes a core node layer, a regional node layer, and an edge node layer. The bottleneck prediction module is used to predict server deployment bottlenecks in the distribution network hierarchical model based on the OTT service operation model, and to establish a multi-level deployment bottleneck map. The server adjustment module is used to adjust the multi-level server deployment of the content delivery network according to the multi-level deployment bottleneck map, and to establish a server deployment adjustment space. The deployment evaluation and optimization module is used to introduce the server deployment evaluation channel to perform deployment evaluation and optimization on the server deployment adjustment space, and obtain the deployment adjustment optimization space. The hierarchical deployment optimization module is used to perform iterative neighborhood optimization on the deployment adjustment optimization space through the server deployment evaluation channel and the server deployment adjustment space to obtain the deployment adjustment optimization result, and to perform hierarchical deployment optimization of the content delivery network based on the deployment adjustment optimization result.

Images