AI cloud desktop operation strategy intelligent issuing method and system

Abstract

The invention particularly relates to an AI cloud desktop operation strategy intelligent issuing method and system. The AI cloud desktop operation strategy intelligent issuing method comprises the following steps: dividing a client cluster into a plurality of resource pools, and assigning an initial main client for each resource pool; each client acquires request platform state data, request response time and response content in real time, encrypts the acquired data and uploads the encrypted data to the algorithm analysis module; the algorithm analysis module analyzes the collected data and generates an adaptive request strategy; and the strategy broadcasting module broadcasts the generated request strategy to all the slave clients in the same resource pool from the master client, and each client executes request operation according to the request strategy and feeds back an execution result to the algorithm analysis module for dynamic adjustment. According to the AI cloud desktop operation strategy intelligent issuing method and system, the peak pressure of the server is relieved, the strategy issuing efficiency is improved, the operation cost is reduced, and the use satisfaction degree of the user is improved.

Description

Technical Field

[0001] This invention relates to the fields of cloud computing and desktop virtualization technology, and in particular to an AI cloud desktop operation strategy intelligent distribution method and system. Background Technology

[0002] With the popularization of cloud computing technology, AI cloud desktops, with their advantages of resource sharing, flexible deployment, and centralized management, have been widely used in enterprise offices, education and training, finance and government affairs. As the core control hub, the AI ​​cloud desktop management platform needs to issue operational policies (such as desktop configuration parameters, software update instructions, resource allocation rules, security policies, etc.) to a massive number of clients to ensure the stable operation and functional iteration of the desktop cluster.

[0003] Currently, the distribution of AI cloud desktop operation strategies mainly adopts the traditional "client-initiated request - server-passive response" model: when a client (such as an employee's office terminal or teaching terminal) needs to obtain the latest operation strategy, it directly sends a request to the management platform server, and the server processes each client's request one by one and returns the strategy content. However, in practical application scenarios, this distribution method has the following problems:

[0004] 1) Excessive peak pressure on the server: When the number of clients reaches thousands or even tens of thousands (such as when enterprises turn on their machines in the morning or when schools synchronize their terminals before class), high-density, high-concurrency requests will be generated, causing the server's CPU utilization to surge, memory usage to exceed the limit, and network bandwidth to become congested, forming a "request flood".

[0005] 2) Slow distribution of operational strategies: When the server handles a large number of concurrent requests, there will be a response delay. Some clients need to wait for several minutes or even longer to obtain the strategy, which seriously affects the user experience of AI cloud desktop (such as slow desktop startup, inability to load functions normally), and may even cause the strategy distribution to fail.

[0006] 3) Limitations of existing solutions: Currently, the industry mostly alleviates pressure by adding server nodes (horizontal scaling) or improving the hardware performance of a single node (vertical scaling), but this approach is costly, has low resource utilization, and cannot fundamentally solve the "request flood" problem; some other solutions use "rate limiting strategies" to avoid server crashes by limiting the number of concurrent requests, but this will cause some client requests to be rejected, further reducing the efficiency of strategy implementation.

[0007] To address the issues of concentrated server-side pressure and low distribution efficiency in existing AI cloud desktop operation strategy distribution methods, this invention proposes an intelligent distribution method and system for AI cloud desktop operation strategies. Summary of the Invention

[0008] To overcome the shortcomings of existing technologies, this invention provides a simple and efficient method and system for intelligently distributing AI cloud desktop operation strategies.

[0009] This invention is achieved through the following technical solution:

[0010] A method for intelligently distributing AI cloud desktop operation strategies includes the following steps:

[0011] Step S1: Resource pool initialization

[0012] Based on the client's physical location, network topology, and associated hardware resources, the client cluster is divided into several resource pools, and an initial master client is assigned to each resource pool.

[0013] Step S2: Real-time data acquisition

[0014] Each client collects real-time data on the platform's status, request and response times, and response content, and then encrypts and uploads the collected data to the algorithm analysis module.

[0015] In step S2, the client and server synchronously obtain the request platform status data through heartbeat packets, including the number of server connections, CPU utilization, memory usage and network bandwidth usage; the response content includes the operation strategy content and data format returned by the server.

[0016] In step S2, each client collects data once every 10 seconds, and the upload frequency is dynamically adjusted according to the server load; the higher the load, the higher the upload frequency.

[0017] Step S3: Algorithm Model Analysis

[0018] The algorithm analysis module analyzes the collected data using K-means clustering and decision tree algorithms to generate appropriate request strategies;

[0019] Among them, the K-means clustering algorithm is used to classify the request characteristics of clients within the same resource pool and identify high-frequency similar requests within the same resource pool whose request frequency exceeds a custom threshold.

[0020] The decision tree algorithm is responsible for generating request strategies based on the server's load status and request response time;

[0021] In step S3, after receiving the data, the algorithm analysis module performs the following operations:

[0022] Step S3.1: Data Preprocessing

[0023] Filter out abnormal data, including invalid data with a response time exceeding 30 seconds, and encode the request type;

[0024] Step S3.2, Cluster Analysis

[0025] The K-means clustering algorithm is used to classify requests within the same resource pool into 3-5 categories, and high-frequency similar requests whose frequency exceeds a custom threshold are identified.

[0026] Step S3.3, Strategy Generation

[0027] Determining server status based on a decision tree model:

[0028] If the server's CPU utilization does not exceed 60% and the response time does not exceed 3 seconds, a regular request strategy will be generated, which means that the client can directly initiate a request.

[0029] If the server's CPU utilization exceeds 60% but does not exceed 80%, and the response time exceeds 3 seconds but does not exceed 5 seconds, a strategy to merge similar requests will be generated, meaning that the main client will initiate a unified request.

[0030] If the server's CPU utilization exceeds 80% and the response time exceeds 5 seconds, a cache request strategy or a priority request strategy will be generated, which means either calling the cache from the client or prioritizing requests from the main client.

[0031] Step S4: Policy Broadcasting and Execution

[0032] The strategy broadcasting module uses the WebSocket protocol to broadcast the generated request strategy from the master client to all slave clients in the same resource pool. Each client executes the request operation according to the request strategy and feeds back the execution result to the algorithm analysis module.

[0033] In step S4, the received request strategy includes at least one of the following: regular request strategy, merging similar requests strategy, caching request strategy, and priority request strategy.

[0034] If it is to merge similar requests, the main client will send a single request to the server to obtain the policy and then distribute it to the slave clients.

[0035] If the response content is cached, the local cached response content is directly retrieved from the client without sending a request to the server.

[0036] If a critical policy is to be prioritized, the client will request the core policy first, and request non-core policies later.

[0037] Step S5: Dynamic Strategy Optimization

[0038] The algorithm analysis module dynamically optimizes the request strategy based on the client's feedback.

[0039] In step S5, the algorithm analysis module updates the algorithm model parameters every 5 minutes based on the execution results fed back by the client, to ensure that the request strategy matches the server load and client demand in real time.

[0040] An AI cloud desktop operation strategy intelligent distribution system is used to implement the above method, including an AI cloud desktop management platform server, a client cluster, a resource pool partitioning module, a data acquisition module, an algorithm analysis module, and a strategy broadcasting module;

[0041] Resource pool partitioning module: Responsible for dividing the client cluster into several resource pools based on the client's physical location, network topology, and associated hardware resources, and assigning an initial master client to each resource pool;

[0042] Data acquisition module: Deployed on each client, responsible for collecting real-time platform status data, request response time and response content, and encrypting and uploading the collected data to the algorithm analysis module;

[0043] Algorithm Analysis Module: Deployed on the management platform server or edge node, it is responsible for receiving multi-dimensional data uploaded by the data collection module, analyzing the collected data through K-means clustering algorithm and decision tree algorithm, and generating appropriate request strategies;

[0044] Policy broadcasting module: The main client deployed within the resource pool is responsible for broadcasting the generated request policy to other slave clients within the same resource pool using the WebSocket protocol.

[0045] Policy execution module: Responsible for executing relevant operations according to the requested policy after the client receives the broadcast request policy.

[0046] A computing device for intelligently distributing AI cloud desktop operation strategies includes:

[0047] One or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for performing any of the methods described above.

[0048] A computer-readable storage medium storing one or more programs, the one or more programs including instructions that, when executed by an AI cloud desktop operation strategy intelligent distribution computing device, cause the AI ​​cloud desktop operation strategy intelligent distribution computing device to perform any of the methods described above.

[0049] The beneficial effects of this invention are: the intelligent distribution method and system for AI cloud desktop operation strategies have strong dynamic adaptability, reduce the processing of repeated requests on the server side, alleviate peak pressure on the server side, improve the efficiency of strategy distribution, reduce operating costs, avoid strategy distribution failure or delay caused by server congestion, ensure the normal loading and stable operation of AI cloud desktop functions, and improve user satisfaction. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Appendix Figure 1 This is a schematic diagram of the AI ​​cloud desktop operation strategy intelligent distribution method of the present invention. Detailed Implementation

[0052] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions in the embodiments of this invention will be clearly and completely described below in conjunction with the embodiments of this invention. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0053] The intelligent distribution method for this AI cloud desktop operation strategy includes the following steps:

[0054] Step S1: Resource pool initialization

[0055] Based on the client's physical location, network topology, and associated hardware resources (such as server cluster affiliation), the client cluster is divided into several resource pools (e.g., 1 resource pool for every 50 clients), and an initial master client is assigned to each resource pool.

[0056] Clients within the same resource pool have similar operational strategy requirements (such as terminals in the same department or on the same floor).

[0057] Step S2: Real-time data acquisition

[0058] Each client collects real-time data on the platform's status, request response time (the time interval between the client sending the policy request and receiving the server's response, accurate to the millisecond level), and response content, and then encrypts and uploads the collected data to the algorithm analysis module.

[0059] In step S2, the client and server synchronously obtain the request platform status data through heartbeat packets, including the number of server connections, CPU utilization, memory usage and network bandwidth usage; the response content includes the operation strategy content (such as configuration parameters and instruction codes) and data format returned by the server.

[0060] In step S2, each client collects data once every 10 seconds, and the upload frequency is dynamically adjusted according to the server load; the higher the load, the higher the upload frequency.

[0061] Step S3: Algorithm Model Analysis

[0062] The algorithm analysis module analyzes the collected data using K-means clustering and decision tree algorithms to generate appropriate request strategies;

[0063] Among them, the K-means clustering algorithm is used to classify the request characteristics (such as request time, policy type, and response time) of clients in the same resource pool, and identify high-frequency similar requests (such as multiple clients requesting the same software update policy within the same time period) that exceed a custom threshold.

[0064] The decision tree algorithm is responsible for generating request strategies based on the server's load status and request response time;

[0065] In step S3, after receiving the data, the algorithm analysis module performs the following operations:

[0066] Step S3.1: Data Preprocessing

[0067] Filter out abnormal data, including invalid data with a response time exceeding 30 seconds, and encode the request type (e.g., software update policy is encoded as 01, security policy is encoded as 02).

[0068] Step S3.2, Cluster Analysis

[0069] The K-means clustering algorithm is used to classify requests within the same resource pool into 3-5 categories, and high-frequency similar requests with a frequency exceeding a custom threshold are identified (e.g., 60% of clients in a certain resource pool request "Office software update policy").

[0070] Step S3.3, Strategy Generation

[0071] Determining server status based on a decision tree model:

[0072] If the server's CPU utilization does not exceed 60% and the response time does not exceed 3 seconds, a regular request strategy will be generated, which means that the client can directly initiate a request.

[0073] If the server's CPU utilization exceeds 60% but does not exceed 80%, and the response time exceeds 3 seconds but does not exceed 5 seconds, a strategy to merge similar requests will be generated, meaning that the main client will initiate a unified request.

[0074] If the server's CPU utilization exceeds 80% and the response time exceeds 5 seconds, a cache request strategy or a priority request strategy will be generated, which means either calling the cache from the client or prioritizing requests from the main client.

[0075] Step S4: Policy Broadcasting and Execution

[0076] The strategy broadcasting module uses the WebSocket protocol to broadcast the generated request strategy from the master client to all slave clients in the same resource pool. Each client executes the request operation according to the request strategy and feeds back the execution result (such as whether the strategy was successfully obtained) to the algorithm analysis module.

[0077] In step S4, the received request strategy includes at least one of the following: regular request strategy, merging similar requests strategy, caching request strategy, and priority request strategy.

[0078] If it is to merge similar requests, the main client will send a single request to the server to obtain the policy and then distribute it to the slave clients.

[0079] If the response content is cached, the local cached response content is directly called from the client (synchronously sent by the main client), without needing to send a request to the server;

[0080] If critical policies are to be prioritized, the client will request core policies (such as security authentication policies) first, and non-core policies (such as desktop theme configuration) will be requested later.

[0081] Step S5: Dynamic Strategy Optimization

[0082] The algorithm analysis module dynamically optimizes the request strategy based on the client's feedback.

[0083] In step S5, the algorithm analysis module updates the algorithm model parameters (such as adjusting the clustering threshold and strategy triggering conditions) every 5 minutes based on the execution results fed back by the client, to ensure that the request strategy matches the server pressure and client demand in real time.

[0084] The AI ​​cloud desktop operation strategy intelligent distribution system is used to implement the above methods, including an AI cloud desktop management platform server, a client cluster, a resource pool partitioning module, a data acquisition module, an algorithm analysis module, and a strategy broadcasting module;

[0085] Resource pool partitioning module: Responsible for dividing the client cluster into several resource pools based on the client's physical location, network topology, and associated hardware resources, and assigning an initial master client to each resource pool;

[0086] Data acquisition module: Deployed on each client, responsible for collecting real-time platform status data, request response time and response content, and encrypting and uploading the collected data to the algorithm analysis module;

[0087] Algorithm Analysis Module: Deployed on the management platform server or edge node, it is responsible for receiving multi-dimensional data uploaded by the data collection module, analyzing the collected data through K-means clustering algorithm and decision tree algorithm, and generating appropriate request strategies;

[0088] Policy Broadcast Module: The main client deployed within the resource pool (selected by the algorithm analysis module from the same resource pool as the client with the best hardware performance and lowest network latency) is responsible for broadcasting the generated request policy to other slave clients within the same resource pool using the WebSocket protocol to ensure low latency and high reliability.

[0089] Policy execution module: Responsible for executing relevant operations according to the requested policy after the client receives the broadcast request policy.

[0090] This AI cloud desktop operation strategy is intelligently distributed to computing devices, including:

[0091] One or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for performing any of the methods described above.

[0092] The computer-readable storage medium stores one or more programs, the one or more programs including instructions that, when executed by an AI cloud desktop operation strategy intelligent distribution computing device, cause the AI ​​cloud desktop operation strategy intelligent distribution computing device to perform any of the methods described above.

[0093] This AI-powered cloud desktop operation strategy intelligent distribution method is based on the logic of "client collaboration - data-driven - strategy sharing". By collecting request data from clients within the same resource pool, it uses an algorithm model to generate and broadcast appropriate request strategies, thereby reducing the processing of duplicate requests on the server side and achieving peak reduction and efficiency improvement in strategy distribution.

[0094] Compared with existing technologies, it has the following characteristics:

[0095] (1) Alleviated peak pressure on the server: By sharing strategies within the same resource pool, the number of repeated requests processed by the server was reduced. According to tests, in a scenario with 1,000 concurrent requests from clients, the number of requests received by the server can be reduced by 60%-70%, and the peak CPU utilization rate is reduced by 40%-50%.

[0096] (2) Improved policy delivery efficiency: Merging similar requests and cached response content can shorten the client waiting time. In the "request flood" scenario, the average time for delivery of operational policies is reduced from 15 seconds to less than 3 seconds, and the delivery success rate is increased from 85% to 99.5%.

[0097] (3) Reduced operating costs: No additional server nodes are needed. Peak reduction is achieved through client collaboration and algorithm optimization, which can reduce hardware expansion costs by 30%-40%.

[0098] (4) Improved user experience: It can avoid policy delivery failure or delay due to server congestion, ensure that the functions of AI cloud desktop are loaded normally and run stably, and improve user satisfaction.

[0099] (5) Strong dynamic adaptability: The algorithm model can adjust the request strategy in real time according to the server pressure and client demand, adapting to AI cloud desktop clusters of different sizes (from hundreds to tens of thousands of units).

[0100] The embodiments described above are merely one specific implementation of the present invention. Ordinary changes and substitutions made by those skilled in the art within the scope of the technical solution of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligently distributing AI cloud desktop operation strategies, characterized in that: Includes the following steps: Step S1: Resource pool initialization Based on the client's physical location, network topology, and associated hardware resources, the client cluster is divided into several resource pools, and an initial master client is assigned to each resource pool. Step S2: Real-time data acquisition Each client collects real-time data on the platform's status, request and response times, and response content, and then encrypts and uploads the collected data to the algorithm analysis module. Step S3: Algorithm Model Analysis The algorithm analysis module analyzes the collected data using K-means clustering and decision tree algorithms to generate appropriate request strategies; Among them, the K-means clustering algorithm is used to classify the request characteristics of clients within the same resource pool and identify high-frequency similar requests within the same resource pool whose request frequency exceeds a custom threshold. The decision tree algorithm is responsible for generating request strategies based on the server's load status and request response time; Step S4: Policy Broadcasting and Execution The strategy broadcasting module uses the WebSocket protocol to broadcast the generated request strategy from the master client to all slave clients in the same resource pool. Each client executes the request operation according to the request strategy and feeds back the execution result to the algorithm analysis module. Step S5: Dynamic Strategy Optimization The algorithm analysis module dynamically optimizes the request strategy based on the client's feedback.

2. The AI ​​cloud desktop operation strategy intelligent distribution method according to claim 1, characterized in that: In step S2, the client and server synchronously obtain the request platform status data through heartbeat packets, including the number of server connections, CPU utilization, memory usage and network bandwidth usage; the response content includes the operation strategy content and data format returned by the server.

3. The AI ​​cloud desktop operation strategy intelligent distribution method according to claim 2, characterized in that: In step S2, each client collects data once every 10 seconds, and the upload frequency is dynamically adjusted according to the server load; the higher the load, the higher the upload frequency.

4. The method for intelligently distributing AI cloud desktop operation strategies according to claim 1, characterized in that: In step S3, after receiving the data, the algorithm analysis module performs the following operations: Step S3.1: Data Preprocessing Filter out abnormal data, including invalid data with a response time exceeding 30 seconds, and encode the request type; Step S3.2, Cluster Analysis The K-means clustering algorithm is used to classify requests within the same resource pool into 3-5 categories, and high-frequency similar requests whose frequency exceeds a custom threshold are identified. Step S3.3, Strategy Generation Determining server status based on a decision tree model: If the server's CPU utilization does not exceed 60% and the response time does not exceed 3 seconds, a regular request strategy will be generated, which means that the client can directly initiate a request. If the server's CPU utilization exceeds 60% but does not exceed 80%, and the response time exceeds 3 seconds but does not exceed 5 seconds, a strategy to merge similar requests will be generated, meaning that the main client will initiate a unified request. If the server's CPU utilization exceeds 80% and the response time exceeds 5 seconds, a cache request strategy or a priority request strategy will be generated, which means either calling the cache from the client or prioritizing requests from the main client.

5. The AI ​​cloud desktop operation strategy intelligent distribution method according to claim 4, characterized in that: In step S4, the received request strategy includes at least one of the following: regular request strategy, merging similar requests strategy, caching request strategy, and priority request strategy. If it is to merge similar requests, the main client will send a single request to the server to obtain the policy and then distribute it to the slave clients. If the response content is cached, the local cached response content is directly retrieved from the client without sending a request to the server. If a critical policy is to be prioritized, the client will request the core policy first, and request non-core policies later.

6. The AI ​​cloud desktop operation strategy intelligent distribution method according to claim 1, characterized in that: In step S5, the algorithm analysis module updates the algorithm model parameters every 5 minutes based on the execution results fed back by the client, to ensure that the request strategy matches the server load and client demand in real time.

7. An AI-powered cloud desktop operation strategy intelligent distribution system, characterized in that: The method for implementing any one of claims 1 to 6 includes an AI cloud desktop management platform server, a client cluster, a resource pool partitioning module, a data acquisition module, an algorithm analysis module, and a policy broadcasting module; Resource pool partitioning module: Responsible for dividing the client cluster into several resource pools based on the client's physical location, network topology, and associated hardware resources, and assigning an initial master client to each resource pool; Data acquisition module: Deployed on each client, responsible for collecting real-time platform status data, request response time and response content, and encrypting and uploading the collected data to the algorithm analysis module; Algorithm Analysis Module: Deployed on the management platform server or edge node, it is responsible for receiving multi-dimensional data uploaded by the data collection module, analyzing the collected data through K-means clustering algorithm and decision tree algorithm, and generating appropriate request strategies; Policy broadcasting module: The main client deployed within the resource pool is responsible for broadcasting the generated request policy to other slave clients within the same resource pool using the WebSocket protocol. Policy execution module: Responsible for executing relevant operations according to the requested policy after the client receives the broadcast request policy.

8. An AI cloud desktop operation strategy intelligent distribution computing device, characterized in that: include: One or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, the one or more programs including instructions for performing the method according to any one of claims 1 to 6.

9. A computer-readable storage medium for storing one or more programs, characterized in that: The one or more programs include instructions that, when executed by the AI ​​cloud desktop operation strategy intelligent distribution computing device, cause the AI ​​cloud desktop operation strategy intelligent distribution computing device to perform the method according to any one of claims 1 to 6.

Images