Intelligent resource dynamic sharing and task scheduling method based on cloud desktop system

Abstract

The invention provides an intelligent resource dynamic sharing and task scheduling method based on a cloud desktop system. According to the method, a resource monitoring module is deployed on a local cloud desktop to monitor the utilization rate of key resources in real time, resources are requested from a cloud desktop system when the resources are insufficient, a cloud desktop resource management module selects a proper cloud desktop to provide the resources, and tasks are reasonably allocated by using a resource allocation algorithm. Meanwhile, efficient operation of the compute-intensive application in the cloud desktop system is realized through task scheduling and monitoring, data synchronization and a feedback adjustment mechanism. According to the method, the resource utilization rate is improved, the task execution time is shortened, the system stability is enhanced, the cost is reduced, and the problems that an existing cloud desktop system is unreasonable in resource allocation and low in task execution efficiency are effectively solved.

Description

Technical Field

[0001] This invention relates to the field of cloud desktop technology, and specifically to a method for intelligent dynamic resource sharing and task scheduling based on a cloud desktop system. Background Technology

[0002] With the rapid development of information technology, computationally intensive applications, such as big data analysis, high-definition image processing, and complex model simulation, are becoming increasingly common in daily work and study scenarios. These applications have extremely high demands on hardware resources, especially CPU and memory, and performance bottlenecks are easily encountered when running on a single machine. Traditional solutions, such as upgrading hardware, are not only costly but also have low resource utilization, resulting in resource waste. Although cloud desktop systems can provide elastic computing resources and alleviate resource shortages to some extent, they currently lack efficient and intelligent mechanisms for dynamic resource sharing and task scheduling between local and cloud environments. In actual use, unreasonable resource allocation and low task execution efficiency often occur, failing to fully leverage the advantages of cloud desktop systems and making it difficult to meet users' demands for efficient operation of computationally intensive applications. Therefore, a new technical solution is urgently needed to address these issues. Summary of the Invention

[0003] Purpose of the invention

[0004] This invention aims to provide an intelligent resource dynamic sharing and task scheduling method based on a cloud desktop system, in order to solve the problems of unreasonable resource allocation and low task execution efficiency of computationally intensive applications in cloud desktop systems in the prior art, and to achieve efficient resource utilization and rapid task processing.

[0005] Technical solution

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] 1. Local Resource Monitoring: A local resource monitoring module is deployed on the user's local cloud desktop. This module continuously monitors the utilization of key resources such as CPU and memory in real time. Once the utilization of any resource exceeds a pre-set threshold, a resource request process is immediately triggered to prepare for acquiring additional resources later.

[0008] 2. Resource Request: Upon detecting resource overruns, the resource request module automatically sends a resource request message to the cloud desktop system. This message includes detailed information such as the type of resources required (e.g., number of CPU cores, memory capacity), quantity, and the urgency of the task, ensuring an accurate response from the cloud desktop system.

[0009] 3. Resource Provider Selection: Upon receiving a resource request, the cloud desktop resource management module comprehensively considers various factors such as the load, geographical location, and network latency of each cloud desktop in the current cloud desktop system. Through intelligent algorithms, it selects one or more suitable cloud desktops from among many cloud desktops as resource providers, ensuring that the provided resources not only meet the needs but also guarantee efficient transmission and use.

[0010] 4. Task Allocation: Employ a carefully designed resource allocation algorithm that comprehensively considers task characteristics (e.g., computational intensity, data dependencies) and the capabilities of resource providers (e.g., remaining computing power).

[0011] (Power, storage space, etc.) The computing task is rationally divided into multiple subtasks and accurately allocated to selected cloud desktops to achieve optimal matching of tasks and resources.

[0012] 5. Task Scheduling and Monitoring: Assigned computing tasks are launched on selected cloud desktops, and a heartbeat mechanism is used to monitor task progress and resource consumption in real time. If task execution anomalies or unreasonable resource usage are detected, timely adjustments can be made to ensure smooth task progress.

[0013] 6. Data Synchronization: A dedicated, high-efficiency data synchronization protocol is designed to ensure real-time data and computation results synchronization among all cloud desktops participating in the computation. This maintains the integrity and consistency of the final results, avoiding computational errors caused by data inconsistencies.

[0014] 7. Feedback and Adjustment: Based on real-time data collected during task execution, such as task progress and resource consumption rate, dynamically adjust resource allocation and task scheduling strategies. In case of unforeseen circumstances...

[0015] (e.g., when a cloud desktop malfunctions) or to further optimize overall performance, timely and reasonable adjustments can be made to ensure the stable and efficient operation of the system. Attached Figure Description

[0016] Appendix Figure 1 The flowchart illustrates the same dynamic resource sharing and task scheduling process, with a brief description of the process:

[0017] Resource shortage feedback: Cloud Desktop 1 detects that it is experiencing resource shortage and sends a "Resource shortage" feedback message to the Cloud Desktop server.

[0018] Task scheduling decision: After receiving the information, the cloud desktop server performs resource assessment and schedules the computing tasks of cloud desktop 1 to cloud desktops 2...N with sufficient resources, thereby balancing the resource load and ensuring the overall efficient operation of the cloud desktop system.

[0019] If cloud desktop 1 lacks sufficient resources, it will seek coordination with the server, which will then distribute the task to other cloud desktops with available resources.

[0020] Appendix Figure 2 Here is a diagram illustrating the intelligent resource dynamic sharing and task scheduling method for cloud desktop systems, with a brief flowchart:

[0021] I. Local Cloud Desktop Side Process

[0022] 1. Resource monitoring triggered

[0023] The local resource monitoring module continuously collects hardware resource data of the local cloud desktop (such as CPU utilization, memory usage, storage I / O, etc.). When a certain type of resource (such as CPU utilization exceeding 80% for 5 consecutive minutes) reaches the preset "over-limit threshold", a resource over-limit signal is generated, marking the type of over-limit resource and its current usage status, triggering the resource expansion requirement.

[0024] 2. Resource request initiation

[0025] After receiving the over-limit signal, the resource request module sorts out the resource requirements of the currently running tasks on the local cloud desktop (such as the resource details required by the running office software, background services, etc.), encapsulates them into a standardized resource request package (including the type, quantity, urgency priority, etc. of the required resources), and sends the resource request to the cloud desktop system over the network.

[0026] II. Cloud Desktop System Side Process

[0027] (1) Resource Management and Scheduling Preparation

[0028] 3. Resource Provider Screening

[0029] After receiving a request, the cloud desktop resource management module iterates through the resource pool within the cloud desktop system (including other available cloud desktops and server-level resource nodes in the cluster), and determines the resource based on the type and quantity of the request, combined with the real-time remaining resources of each candidate node (e.g., Node1 has 16GB of remaining memory).

[0030] Based on factors such as the number of idle CPU cores (e.g., 4), and network topology (prioritizing low-latency nodes on the same network segment), suitable resource providers (which may be a single or multiple cloud desktop nodes) are selected.

[0031] 4. Allocation Strategy Calculation

[0032] The resource allocation algorithm, based on the selected providers, invokes preset strategies (such as load balancing strategies that prioritize distributing load evenly, and performance-first strategies that select high-configuration nodes) to calculate the resource allocation scheme: determining which nodes and which resources to allocate to the local cloud desktop (e.g., allocating 2 cores from Node2).

[0033] (CPU, 8GB memory) Generate and allocate task instructions, specifying the exact parameters and execution order for resource transfer / sharing.

[0034] (2) Task execution and closed-loop monitoring

[0035] 5. Scheduling Execution and Data Acquisition

[0036] After receiving the assigned task, the task scheduling and monitoring module sends execution instructions to the resource provider and the local cloud desktop:

[0037] • Resource provider side: Release / share specified resources according to instructions (such as enabling memory sharing channels, marking CPU cores as available);

[0038] • Local cloud desktop side: Integrate new resources and adjust task execution strategies (such as migrating high-load processes to newly added CPU cores).

[0039] Meanwhile, this module collects task execution data (such as resource migration time of 2 seconds) and task progress (such as the completion of office software rendering task) in real time through data synchronization and integration mechanisms.

[0040] The data on resource consumption (30% → 70%) and the amount of newly added memory used (5GB) are all aggregated in the system data platform.

[0041] Feedback and dynamic optimization

[0042] The feedback and adjustment mechanism periodically pulls the above data from the data platform and compares it with the expected target of resource requests (such as expecting CPU utilization to drop below 60%):

[0043] • If the actual results meet the target (e.g., CPU utilization is stable at 55%), then generate a "task completed" confirmation and end the current scheduling.

[0044] If the target is not met (e.g., memory allocation still exceeds 90%), analyze the reasons for the discrepancy (e.g., insufficient actual available resources from the resource provider, memory leaks in the task itself), adjust the strategy (e.g., change the resource provider, allocate additional resources), and send the adjustment strategy back to the resource allocation algorithm to trigger a secondary adjustment.

[0045] Continue this process until resource needs are met.

[0046] The entire process forms a closed loop of "monitoring → requesting → allocating → executing → feedback → optimizing", ensuring that local cloud desktop resources dynamically adapt to business loads and demonstrating the elastic scaling and intelligent scheduling capabilities of the cloud desktop system.

[0047] Beneficial effects

[0048] 1. Improve resource utilization: By monitoring and dynamically allocating resources in real time, idle and wasted resources are avoided, and resources in the cloud desktop system are utilized more fully. Compared with traditional methods, resource utilization can be increased by more than 30%.

[0049] 2. Improved Task Execution Efficiency: The intelligent task scheduling and allocation mechanism optimizes matching based on the actual situation of tasks and resources, effectively reducing task execution time. Tests show that the runtime of compute-intensive applications is reduced by an average of approximately 40%.

[0050] 3. Enhanced system stability: The data synchronization mechanism ensures data consistency, and the feedback and adjustment mechanism can respond promptly to emergencies, enabling the entire cloud desktop system to operate stably even in complex environments.

[0051] Yes, it can reduce the system failure rate by more than 50%.

[0052] 4. Reduced costs: Rational resource utilization reduces the need for hardware upgrades, thereby lowering hardware procurement and maintenance costs. In the long run, this can save approximately 40% of costs. Detailed Implementation

[0053] Example 1: Basic Implementation

[0054] 1. Local Resource Monitoring Settings: Install the local resource monitoring module on your local cloud desktop and configure CPU settings.

[0055] Resource usage exceeding 80% and memory usage exceeding 70% are considered resource overrun thresholds. This module collects resource usage data every 10 seconds.

[0056] 2. Resource Request Configuration: Upon receiving a resource overrun signal, the resource request module sends a request to the cloud desktop system according to a preset format. For example, if CPU resources are insufficient, the request information will be "Request..."

[0057] 2 CPU cores, urgency level: high.

[0058] 3. Resource Provider Selection Strategy: The cloud desktop resource management module uses a weighted algorithm to select resource providers. Load conditions account for 40% of the weight, geographical location accounts for 30%, and network latency accounts for [missing percentage].

[0059] 30%. Based on these weights, each cloud desktop is evaluated comprehensively, and the cloud desktop with the highest score is selected to provide resources.

[0060] 4. Task Allocation Strategy: For a big data analysis task, the resource allocation algorithm divides it into 5 subtasks based on the data volume and computational complexity. The subtasks are then allocated to suitable cloud desktops based on the remaining computing power and storage space of each resource provider. For example, cloud desktops with strong computing power and large storage space are assigned to subtasks with higher computational demands.

[0061] 5. Task Scheduling and Monitoring: Subtasks are launched on the selected cloud desktop. The task scheduling and monitoring module uses a heartbeat mechanism to obtain task progress and resource consumption data every 30 seconds. If a subtask is found to be progressing slowly, resource allocation is adjusted promptly.

[0062] 6. Data Synchronization Implementation: A message queue-based data synchronization protocol is adopted. When data changes, each cloud desktop encapsulates the changed data into a message and sends it to the message queue. Other cloud desktops retrieve the message from the message queue and synchronize it to ensure data consistency.

[0063] 7. Feedback and Adjustment Implementation: Based on data feedback during task execution, if the CPU utilization of a certain cloud desktop continues to exceed 90%, resources will be allocated from other idle cloud desktops to reallocate tasks and ensure the smooth progress of the overall task.

[0064] Example 2: Optimization in Complex Task Scenarios

[0065] In computationally intensive tasks with complex data dependencies, such as complex model simulation:

[0066] 1. Task Partitioning and Allocation Optimization: When partitioning tasks, the resource allocation algorithm fully considers data dependencies, assigning interdependent data and computational tasks to the same cloud desktop or a cloud desktop cluster with low network latency to reduce data transmission overhead. For example, the model's initialization data and related computational tasks are assigned to the same cloud desktop, and subsequent iterative computational tasks are assigned to appropriate cloud desktops based on data flow and computational needs.

[0067] 2. Dynamic Resource Adjustment: During task execution, resources are dynamically adjusted according to the phased requirements of the simulation task. For example, during the model building phase, memory requirements are high, so memory resource allocation is increased; during the model calculation phase, CPU requirements increase, so CPU resource investment is increased to further improve task execution efficiency.

Claims

1. A method for intelligent dynamic resource sharing and task scheduling based on a cloud desktop system, characterized in that, Includes the following steps: o Local resource monitoring: Utilizes a monitoring module deployed on the local cloud desktop to monitor the utilization of key resources such as CPU and memory in real time. When the utilization of any resource exceeds a preset threshold, a resource request process is triggered. o Resource Request: After detecting that the resource limit has been exceeded, the resource request module automatically sends a resource request to the cloud desktop system. The request information includes the type, quantity, and urgency of the required resources. o Resource provider selection: The cloud desktop resource management module receives resource requests and selects one or more suitable cloud desktops as resource providers based on factors such as the load, geographical location, and network latency of each cloud desktop in the current cloud desktop system through intelligent algorithms. o Task allocation: Using resource allocation algorithms, the computing tasks are reasonably divided and allocated to selected cloud desktops by comprehensively considering factors such as task characteristics (e.g., computational intensity, data dependencies) and resource provider capabilities (e.g., remaining computing power, storage space). o Task scheduling and monitoring: Start the assigned computing tasks on each selected cloud desktop, and use the heartbeat mechanism to monitor the task progress and resource consumption in real time to ensure the smooth execution of the tasks. o Data synchronization: Through a designed high-efficiency data synchronization protocol, all cloud desktops participating in the computation can synchronize data and computation results in real time, maintaining the integrity and consistency of the final result. o Feedback and Adjustment: Based on real-time data during task execution, dynamically adjust resource allocation and task scheduling strategies to cope with emergencies or optimize overall performance.

2. The method according to claim 1, characterized in that, When selecting resource providers, the intelligent algorithm prioritizes cloud desktops that are geographically close and have low network latency. Under the premise of load balancing, it prioritizes allocating resources to cloud desktops with strong remaining computing power.

3. The method according to claim 1, characterized in that, The data synchronization protocol uses an incremental synchronization method, synchronizing only the changed data to reduce network transmission pressure and synchronization time.

4. The method according to claim 1, characterized in that, In the feedback and adjustment mechanism, when the task execution speed is too slow, the resources allocated to the task are increased; when the resource utilization rate is too low, idle resources are reallocated.

Images