Cloud-based distributed discrete event simulation platform

Abstract

The invention provides a distributed discrete event simulation platform based on cloud. The distributed discrete event simulation platform effectively solves the problems of computing resource shortage and low computing efficiency in a traditional single-machine discrete event simulation technology. By combining cloud computing and distributed computing technologies, efficient deployment and operation of simulation tasks in a cloud environment are realized. The platform adopts a Master-Slave structure, and a Master node serves as a central dispatching center and is responsible for distribution, real-time monitoring, state feedback and result summarization of simulation tasks; and the Slave node is responsible for executing a specific simulation task. A dynamic allocation strategy based on node resource real-time evaluation is adopted, simulation resources can be dynamically adjusted according to actual requirements, simulation efficiency and result reliability are ensured, and meanwhile simulation cost is reduced. The method has high practical value and wide application prospect.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of cloud services, and more particularly relates to a cloud-based distributed discrete event simulation platform. BACKGROUND

[0002] In recent years, in the fields of large-scale system modeling and optimization, complex process analysis and multi-objective decision evaluation, discrete event simulation (DES) has become an important tool to support scientific research, industrial and engineering decision-making. Traditional discrete event simulation platforms mostly adopt single-machine deployment mode, and use multi-core and multi-thread technology to improve the parallelism and efficiency of simulation calculation. For small and medium-sized simulation tasks, single-machine multi-core architecture can meet the needs to a certain extent, relying on the expansion ability of local computing resources, through increasing the number of processor cores, improving CPU frequency and increasing memory capacity, to realize the linear growth of computing power. However, with the continuous expansion of application fields, the simulation scale is rapidly expanding, and the number and complexity of model parameters are greatly improved. Relying only on the computing power and resource expansion of a single machine has been difficult to cope with the high concurrency and high throughput pressure brought by multi-parameter groups, multi-random seeds and batch experiments. The power ceiling of a single machine limits the horizontal scalability of the simulation platform, and when executing large batch simulation tasks or super-large scale model calculations, resource bottlenecks are prone to occur, task scheduling and management are significantly increased in difficulty, and it is also difficult to guarantee the high availability and high reliability of the service.

[0003] At the same time, with the rapid development of cloud computing and distributed system technology, more and more computing-intensive and data-intensive applications are gradually migrating to the cloud, making full use of the elastic resource pool and distributed parallelism of the cloud platform to realize efficient operation and automatic elastic scaling of large-scale tasks. At present, the mainstream discrete event simulation platforms are mostly limited to local environment or cluster-type static resource scheduling, and lack the deep integration of cloud computing and distributed dynamic scheduling capabilities. Users often need to manually configure the hardware environment, and the flexibility of simulation experiment scheduling and resource management is insufficient, making it difficult to adapt to complex business scenarios and changing resource needs. In addition, traditional solutions often rely on manual intervention in experiment monitoring, exception handling, automatic statistics and reporting, etc. In large-scale multi-round batch experiments, it is easy to cause uncontrollable progress, difficult fault location, experimental data fragmentation and other problems, which greatly restricts the construction of efficient and reliable simulation service system.

[0004] In view of the above technical bottleneck and the increasing actual demand, it is urgent to have a distributed simulation platform capable of fully utilizing cloud computing resources, having elastic scaling and high concurrency scheduling capability, breaking through the "ceiling" of single-machine multi-core architecture, and realizing linear or even super-linear expansion of simulation task execution capability. Therefore, the application proposes a cloud-based distributed discrete event simulation platform with a new architecture, introduces Master-Slave distributed architecture, realizes high interconnection and intercommunication of nodes through the Internet, combines the elastic resource scheduling capability of the cloud platform, and can automatically perceive and dynamically allocate multi-dimensional resources such as computing, storage and network bandwidth. The platform collects the resource state and historical performance of each node in real time, combines the stability and task execution efficiency of the node, intelligently scores and dynamically assigns simulation tasks, and realizes efficient parallel simulation scheduling. During task execution, the system can monitor the progress and health status of each node in real time, automatically migrate and reschedule when an exception is found, and ensure the continuity and high availability of the entire simulation process. In the aspect of experiment management, the platform supports users to flexibly configure complex experiment parameters through a graphical interface, automatically generates parameter groups and random seed combinations, and submits large-scale simulation tasks in batches, realizes elastic scaling in the cloud, and seamlessly handles high-concurrency experiment scenarios. SUMMARY

[0005] The main technical problem to be solved by the application is how to combine cloud computing and distributed computing technology, realize efficient deployment and running of discrete event simulation tasks in a cloud computing environment, provide reasonable task scheduling and management strategies, fully utilize the elastic resources of cloud computing, realize dynamic allocation and adjustment of simulation tasks, thereby improve simulation efficiency, expand simulation scale, reduce simulation cost, and ensure the reliability of simulation results.

[0006] To achieve the above purpose, the application adopts the following technical solutions:

[0007] The platform comprises a Master node and a plurality of Slave nodes, the Master node is connected with the plurality of Slave nodes through the Internet, the Master node serves as a central scheduling center, is responsible for allocation, real-time monitoring, exception handling and result summarization of simulation tasks, each Slave node is responsible for execution of specific simulation tasks, and the platform comprises the following steps:

[0008] (1) The experiment manager generates a plurality of simulation tasks containing different input parameters and observation numbers according to the experiment scene built by the user and the simulation time set, and submits the tasks to the Master node for scheduling;

[0009] (2) After the Master node receives the tasks, the rationality and integrity of the simulation model and parameters are checked;

[0010] (3) The Master node obtains resource usage of all Slave nodes in real time, performs comprehensive quantitative scoring based on CPU, memory, bandwidth, historical execution efficiency, node stability and elastic expansion capacity, and dynamically allocates different simulation tasks to each Slave node according to the scoring results;

[0011] (4) The Master node records scheduling logs and generates a task ID, the Slave node starts to execute the allocated task after initializing the simulation environment, and reports progress and resource usage to the Master node in real time during running;

[0012] (5) The Master node monitors the task execution state of each Slave node in real time, and if it detects node exception or task timeout, it migrates the abnormal task to other suitable Slave node for continuous execution through the rescheduling mechanism, to ensure the continuity of the simulation task;

[0013] (6) After all Slave nodes complete the task, the results are returned to the Master node, the Master node combines and counts all task results, generates a comprehensive report including task completion time, throughput, delay, resource utilization and other indicators, and sends the report to the user in the form of system notification or email.

[0014] In one scheme, the resource status evaluation of the Master node to the Slave node includes periodically collecting CPU utilization, memory utilization and network bandwidth occupation of each Slave node, and combining historical task execution efficiency, node stability and elastic expansion capacity of the node to perform comprehensive scoring on each node, which is used to guide optimal allocation of simulation tasks.

[0015] In one scheme, the parameter configuration of the simulation task supports automatic enumeration of all experimental parameter value combinations, and generates independent experimental observations according to multiple different random number seeds for each parameter group configuration, to realize repeatable and statistically sufficient simulation experiments.

[0016] In one scheme, during the execution of the simulation task, the Slave node can report simulation progress and self-resource consumption data to the Master node every second, and the Master node dynamically adjusts the scheduling strategy according to the real-time collected information, to improve the stability and efficiency of system running.

[0017] In one scheme, when the Master node detects that a Slave node has resource overload, failure or task timeout, etc., it can record abnormal logs, and based on the current resource evaluation results, automatically trigger the task migration mechanism to reassign the abnormal task to the Slave node with the lightest load or the most abundant resources for re-execution, to ensure the integrity and continuity of the task.

[0018] In one scheme, the test manager supports user graphical configuration of experimental scenes, including input parameters, output parameter settings and simulation observation number settings, and can automatically count all experimental groups and observation numbers, facilitating generation and management of large-scale automatic simulation tasks.

[0019] In one scheme, the Master node can automatically count and analyze simulation data reported by all Slave nodes after the task is completed, including calculation of mean value, maximum value, minimum value, standard deviation and confidence interval of each experiment, and automatic generation of a structured detailed report.

[0020] In one scheme, the final simulation report generated by the Master node can be notified to the user through the system interface or pushed to the user in the form of an email, and the report content includes task execution details, resource utilization, main performance indicators and abnormal processing records, so that the user can comprehensively evaluate and trace analyze the simulation results.

[0021] Advantages of the present application:

[0022] The platform of the present application can dynamically adjust and allocate simulation resources according to the actual needs of simulation tasks, fully utilize the advantages of elastic resource supply of cloud computing, and maximize the use of cloud resources.

[0023] Reasonable task scheduling strategies are introduced, which effectively balance the resource utilization rate and task execution efficiency, especially in the case of node abnormality or task timeout, task rescheduling can be quickly performed to ensure the continuity of simulation.

[0024] Compared with the traditional single-machine multi-core mode, the cloud-based distributed simulation platform of the present application not only greatly expands the computing scale and concurrent processing capacity of simulation, but also realizes a qualitative leap in experiment automation, resource utilization rate, reliability and traceability. The built-in whole-process automatic monitoring, statistics and report analysis mechanism can generate comprehensive experimental reports including key indicators such as throughput, delay and resource utilization for the user, and actively push them through system notification or email, which significantly improves the experimental management efficiency of the user and the overall intelligent level of simulation services. In summary, the cloud-based distributed discrete event simulation platform not only breaks through the limitations of traditional single-machine simulation, but also provides a new technical path and solid platform support for simulation experiments under large-scale, high-complexity and multi-parameter scenarios, and has broad popularization prospects and important innovation value in the fields of scientific research and engineering application. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 The platform architecture diagram of the present application is shown in the figure;

[0026] Figure 2Task scheduling flowchart for the present application;

[0027] Figure 3 Experimental scene built for users of the present application;

[0028] Figure 4 Configuration interface one for platform test manager of the present application;

[0029] Figure 5 Configuration interface two for platform test manager of the present application;

[0030] Figure 6 Configuration interface three for platform test manager of the present application. DETAILED DESCRIPTION

[0031] In order to facilitate the understanding of the present application, the present application will be described more fully below with reference to the accompanying drawings. The drawings show typical embodiments of the present application. However, the present application can be implemented in many different forms and is not limited to the embodiments described in the present application. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive.

[0032] Unless otherwise defined, all technical and scientific terms used in the present application have the same meaning as understood by one skilled in the art to which the present application belongs. The terms used in the present application in the specification of the present application are only for the purpose of describing the specific embodiments and are not intended to limit the present application. In order to facilitate the understanding of the present application, the present application will be described more fully below with reference to the accompanying drawings. The drawings show typical embodiments of the present application. However, the present application can be implemented in many different forms and is not limited to the embodiments described in the present application. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive.

[0033] (1) Overall architecture of the platform

[0034] As shown in Figure 1 , the present application adopts Master-Slave structure, Master node as the central scheduling center, responsible for simulation task allocation, real-time monitoring, state feedback and result summary; multiple Slave nodes are connected with Master node through the Internet, responsible for the execution of specific simulation tasks.

[0035] The cloud-based distributed discrete event simulation platform proposed in the application is designed and implemented around a Master-Slave structure. In this architecture, the Master node plays the core role of the central scheduling center, controlling the running rhythm of the entire distributed simulation system. When the system is initialized, the Master node first starts the registration service and listens to the access requests of all Slave nodes. Slave nodes are distributed in the cloud or intranet environment and establish stable remote communication connection with the Master node through the Internet. These Slave nodes will actively report their computing power, available resources, network status and other basic information to the Master node when they first join the platform, so as to facilitate the Master to manage and schedule the global resources.

[0036] In the simulation task starting stage, the user submits the simulation experiment and its parameter configuration through the platform interface or API. These task data are first transmitted to the Master node. The Master node checks the received task content, confirms the integrity of the model file and the validity of the parameters. After the task is verified correctly, the Master node uses the built-in resource monitoring and task scheduling module to analyze and quantify the load and availability of all Slave nodes in real time, intelligently allocates each simulation sub-task to the most suitable Slave node, and realizes the balanced distribution of the task.

[0037] In the specific task execution process, the Slave node initializes its simulation environment according to the instructions of the Master, loads the corresponding model and parameters, and starts to complete the simulation process independently. The Slave node will periodically report its task progress, resource consumption and running state to the Master node during the simulation running period. The Master node continuously collects and monitors the feedback of all Slave nodes and can timely find out abnormal conditions such as failure, overload or delay of a node. Once an abnormality is found, the Master node will flexibly migrate the related simulation task to other healthy Slave nodes according to the actual situation, ensuring the stable operation and efficient execution of the whole system.

[0038] After the simulation task is completed, the slave node returns information, including simulation logs, statistics, and output results, to the master node via a secure network path. The master node organizes and summarizes the result data from each slave node, ultimately generating a detailed simulation report. This report not only includes the completion status and performance indicators of each subtask, but also integrates the resource utilization efficiency and scheduling strategy of the entire distributed system to provide users with comprehensive evaluation and decision support. By implementing this master-slave overall architecture, the platform can fully leverage the elasticity and distributed computing capabilities of cloud resources, significantly improving the efficiency, reliability, and scalability of large-scale discrete event simulation.

[0039] (2) Task scheduling and management methods

[0040] like Figure 2 As shown in the figure, the Master node receives simulation experiment tasks submitted by users and dynamically assigns them to different slave nodes based on the number of tasks and the load of the slave nodes. The Master node monitors the operating status of the slave nodes in real time and quickly reschedules tasks in the event of an anomaly to ensure overall simulation efficiency.

[0041] Step 1. Initialize the test manager and generate tasks

[0042] The experiment manager is based on the experimental scenario built by the user and the simulation time set as follows Figure 3 As shown:

[0043] like Figure 4 As shown, configure the input parameters and values ​​(input parameter Buffer1.Capacity, value range 1-8), and output parameters (Sink1.MetricsOutputPerHour) in the "Experiment Manager" object. Each set of parameters constitutes an experiment.

[0044] like Figure 5 As shown, configure the number of observations for each experiment (default 5). The simulation will run multiple observation models for each experiment using different random number seeds. The simulation will calculate the mean, maximum, minimum, confidence interval, and standard deviation based on the results of multiple experimental observations. In this case, the simulation will run a total of 8 (number of experiments) * 5 (number of observations) = 40 experiments.

[0045] Select "Cloud Distributed Simulation", click "Reset" and "Start", and the experiment manager will pass these tasks to the Master node for scheduling.

[0046] Step 2. Master receives and verifies task parameters

[0047] After receiving the task, the Master verifies whether the simulation model has any defects and whether it can run normally; and verifies the rationality and completeness of the parameter values ​​set by the experiment manager.

[0048] After receiving a simulation task submitted by a user, the Master node first performs rigorous validation on all task parameters and the simulation model itself to ensure smooth, efficient, and accurate execution of subsequent simulations. Specifically, the Master node performs an integrity check on the task description file and associated simulation model, including ensuring the completeness of the model files, compatibility of the file format with supported simulation engines, and compliance with agreed-upon specifications. The Master node then automatically parses the model's internal structure, checking for syntax errors, logical conflicts, or missing dependencies. For example, the Master node verifies that the modules and functions called in the model are available and that the relevant data input and output interfaces are complete and free of type mismatches. For each parameter set by the Experiment Manager, the Master node performs bounds checking and validity analysis based on the task configuration template to ensure that all parameters are within the permitted range and that parameter value combinations do not cause execution exceptions due to conflicts, omissions, or duplications. For enumerated or Boolean parameters, the Master node also verifies that their values ​​are strictly within the defined set.

[0049] In addition, the Master node will also check the consistency and integrity of metadata such as the grouping and number of observations of tasks in the experiment manager to ensure that the number of tasks and the hierarchical relationship are consistent with the user's intentions. In order to prevent simulation failures due to user configuration omissions, the Master node adopts a multi-layer security mechanism in the verification process. Once the model file is damaged, the parameters are abnormal, or there are potential inconsistencies, the error information will be immediately fed back to the upper-level user interface or API call end, prompting the user to make corrections and prohibiting non-compliant tasks from entering the next scheduling link. Through the above series of rigorous inspection and verification processes, the Master node can maximize the availability and accuracy of simulation tasks, laying a solid foundation for distributed scheduling and execution. Step 3. Query the list of available SLAVE nodes

[0050] The master node first evaluates the resource status (CPU, memory, bandwidth) of all computing nodes in the current environment. For example, there are 4 computing nodes in the cluster (Slave 1, Slave 2, Slave 3, Slave 4).

[0051] Step 4. Evaluate the resource situation of each node

[0052] Their resource usage is as follows:

[0053] Slave 1: 40% CPU, 60% memory

[0054] Slave 2: 50% CPU, 70% memory

[0055] Slave 3: 30% CPU, 50% memory

[0056] Slave 4: 90% CPU, 95% memory

[0057] Step 5. Generate allocation strategy for each simulation task and allocate to corresponding node

[0058] In the process of simulation task scheduling, in order to maximize the utilization of cluster resources and improve the efficiency of task execution, this scheme proposes a dynamic allocation strategy based on real-time evaluation of node resources:

[0059] After receiving the newly generated simulation task, the Master node will conduct real-time resource monitoring and evaluation of all Slave nodes. Since each simulation task is essentially similar in type, the main difference lies in the difference in resource demand such as parameter scale, computational amount, etc. Therefore, the core of this strategy is to real-time fine evaluation of the CPU usage rate, memory usage rate, and network bandwidth occupancy rate of each node, and to comprehensively consider the historical execution efficiency, stability, and the potential of resource dynamic adjustment of the node, to quantitatively obtain the comprehensive score of each node, so as to determine the most suitable execution node for the current task.

[0060] The specific quantitative calculation formula is as follows:

[0061] S = w1 x (1 - L cpu ) + w2 x (1 - L mem )

[0062] + w3 x (1 - L net ) + w4 x E hist + w5 x R stab + w6 x E ftex

[0063] Wherein:

[0064] Lcpu: current CPU occupancy rate of node (value 0-1, the lower the better) ;

[0065] Lmem: current memory occupancy rate of node (value 0-1, the lower the better) ;

[0066] Lnet: current network bandwidth occupancy rate of node (value 0-1, the lower the better) ;

[0067] Ehist: historical task execution efficiency score of node (value 0-1, the higher the better) ;

[0068] Rstab: recent stability score of node (value 0-1, 1 represents recent no failure) ;

[0069] Eflex: Node elastic resource adjustment capability score (value 0-1, 1 indicates high elastic adjustment capability).

[0070] The weights are initially set as follows (which can be dynamically adjusted according to actual conditions): CPU resource weight w1=0.25; memory resource weight w2=0.25; network resource weight w3=0.15; historical execution efficiency weight w4=0.2; stability weight w5=0.1; elastic adjustment capability weight w6=0.05.

[0071] Each weight coefficient satisfies:

[0072] w1+w2+w3+w4+w5+w6=1

[0073] Take the actual task allocation as an example:

[0074] Task1 is allocated to Slave 3 because the CPU and memory resource comprehensive score of Slave 3 is the highest at this time;

[0075] Task2 is allocated to Slave 1 because Slave 1 has the lowest current load and can quickly handle new tasks;

[0076] Task3 is allocated to Slave 2 because Slave 2 has a relatively wide network bandwidth and can better support frequent data interaction;

[0077] Task4 is allocated to Slave 4, although the current load is heavy, Slave 4 has an elastic expansion mechanism and can meet the task execution demand through resource dynamic adjustment.

[0078] This strategy effectively balances resource utilization and task execution efficiency through detailed real-time resource quantification and comprehensive scoring, ensuring stable and efficient operation of the system under different load conditions.

[0079] Step 6. Master records the scheduling log and returns each task ID

[0080] After the allocation of simulation tasks is completed, the Master node will immediately start the process of recording the scheduling log and returning the task identification. In the implementation, the Master node first generates a globally unique task ID for each allocated simulation subtask. This task ID is usually encrypted by a timestamp, task source information, node sequence number, and other factors to ensure that all tasks in the system can be uniquely tracked. Subsequently, the Master node persistently saves the complete task scheduling information, including the ID of each task, the simulation experiment and parameters to which the task belongs, the identification of the target Slave node to which the task is allocated, the allocation time, the expected completion time, the node resource snapshot at the time of scheduling, priority information, and the basis for scheduling strategy, and other detailed contents. At the same time, the log system automatically generates a version number for the task scheduling record, which facilitates the tracing of historical states during subsequent fault recovery or task rescheduling. This scheduling log is not only written to the local database in real time, but can also be synchronized to the distributed log service to improve data security and system scalability. After the task scheduling information record is completed, the Master node will return the generated task ID and related information to the experiment manager or front-end user system in a timely manner through a designated message channel or API interface, supporting the user's subsequent real-time query of the task running state, execution progress, and results. Through this series of operations, the platform realizes full-process traceable management of task allocation, providing a solid data foundation for fault troubleshooting, performance analysis, and system auditing. Step 7. Slave task execution and monitoring

[0081] As shown in Figure 6 , each Slave node receives the task, initializes the simulation environment, and starts execution. During the execution process, the Slave node reports the task progress and resource usage to the Master node every second. The Master node monitors the execution status of each task in real time and detects whether there are any exceptions or timeout conditions.

[0082] After the Master node assigns simulation tasks to each Slave node, the Slave node first parses the received task instructions, loads the corresponding simulation model and parameter configuration, and then automatically initializes its simulation environment, including scheduling the required computing resources, memory space, and network ports, and loading the required dependent libraries and local middleware to ensure high compatibility between the running environment and the simulation task. After the simulation environment is initialized, the Slave node immediately starts the simulation process and enters the formal task execution phase. During the entire simulation execution process, the Slave node maintains a high-precision task monitoring sub-thread that collects the current task progress information, resource consumption indicators such as CPU and memory usage, and the health status of the process or container. To facilitate real-time control by the Master, the Slave node actively reports its task progress, resource usage, and running log summary to the Master node every second through a pre-agreed message protocol and secure channel, ensuring that the Master node can real-time perceive the execution dynamics of each sub-task. After receiving the progress report from the Slave, the Master node compares the monitoring data with the scheduling log and continuously judges whether the execution status of each task is normal, relying on the built-in state analysis and alarm mechanism. If it finds that a task has been stuck for a long time, has abnormal resource consumption, or has timed out, the Master node will immediately issue an alarm and take emergency measures such as retry, migration, or task termination according to the exception category, thereby avoiding single-point failure affecting the overall simulation progress and result accuracy of the system. Through this high-frequency two-way communication and fine-grained monitoring mechanism, the platform not only realizes the visual management of distributed simulation task execution process, but also significantly improves the reliability and execution efficiency of the task.

[0083] During task execution, Slave 4 (the node executing Task4) experiences an exception due to resource overload (CPU and memory are close to 100%), and the simulation task cannot continue. After detecting the exception, the Master node immediately triggers the rescheduling mechanism and migrates Task4 from Slave 4 to Slave 3 (the node with the lightest current load). After the task migration is complete, Slave 3 starts to re-execute Task4, ensuring the continuity of the simulation task.

[0084] Step 9. Task completion and result aggregation

[0085] All tasks complete execution on their respective Slave nodes and send the calculation results (including simulation logs, statistical data, performance indicators, etc.) to the Master node. The Master node collects and merges the execution results of all tasks, generates the final report, and sends a notification to the user through the system.

[0086] Step 10. Report generation and user notification

[0087] The Master node generates a comprehensive report based on the execution data of all tasks, including task completion time, throughput, latency, resource utilization, and other indicators. The final report system notifies or sends an email to the user for viewing.

[0088] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, can include the processes of the above-mentioned embodiments of each method. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), or the like.

[0089] It should be understood that the above detailed description of the technical solutions of the present application by means of preferred embodiments is illustrative rather than limiting. A person of ordinary skill in the art can modify the technical solutions recorded in each embodiment or make equivalent substitutions for part of the technical features on the basis of the present application; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. A cloud-based distributed discrete event simulation platform, characterized by: The platform includes a Master node and multiple Slave nodes. The Master node and the multiple Slave nodes are connected via the Internet. The Master node serves as a central dispatching center and is responsible for the allocation, real-time monitoring, exception handling, and result aggregation of simulation tasks. Each Slave node is responsible for the execution of a specific simulation task. The platform includes the following steps: (1) The experiment manager generates multiple simulation tasks with different input parameters and observation numbers based on the user-built experimental scenario and the set simulation time, and submits these tasks to the Master node for scheduling; (2) After receiving the task, the Master node verifies the rationality and integrity of the simulation model and parameters; (3) The Master node obtains the resource usage of all Slave nodes in real time, and makes a comprehensive quantitative score based on factors such as CPU, memory, bandwidth, historical execution efficiency, node stability, and elastic expansion capability. Based on the score results, different simulation tasks are dynamically allocated to each Slave node; (4) The Master node records the scheduling log and generates a task ID. The Slave node initializes the simulation environment and starts executing the assigned task. During the execution, it reports the progress and resource usage to the Master node in real time. (5) The Master node monitors the task execution status of each Slave node in real time. If a node abnormality or task timeout is detected, the abnormal task is migrated to another suitable Slave node through the rescheduling mechanism to continue execution, ensuring the continuity of the simulation task; (6) After completing the task, all slave nodes return the results to the master node. The master node merges and counts all task results, generates a comprehensive report including indicators such as task completion time, throughput, latency, resource utilization, etc., and sends the report to the user in the form of system notification or email.

2. The cloud-based distributed discrete event simulation platform according to claim 1, wherein: The Master node's resource status assessment of the Slave node includes regularly collecting the CPU utilization, memory utilization, and network bandwidth occupancy of each Slave node, and combining the node's historical task execution efficiency, node stability, and elastic expansion capability to give each node a comprehensive score to guide the optimal allocation of simulation tasks.

3. The cloud-based distributed discrete event simulation platform according to claim 1, wherein: The parameter configuration of the simulation task supports automatic enumeration of value combinations of all experimental parameters, and configures multiple different random number seeds according to each parameter group to generate independent experimental observations, thereby achieving repeatable and statistically significant simulation experiments.

4. The cloud-based distributed discrete event simulation platform according to claim 1, wherein: During the execution of the simulation task, the Slave node can report the simulation progress and its own resource consumption data to the Master node every second. The Master node dynamically adjusts the scheduling strategy based on the real-time collected information to improve the stability and efficiency of the system operation.

5. The cloud-based distributed discrete event simulation platform according to claim 1, wherein: When the Master node detects an abnormality such as resource overload, failure or task timeout on a Slave node, it can record the abnormality log and automatically trigger the task migration mechanism based on the current resource evaluation results to reallocate the abnormal task to the Slave node with the lightest load or the most abundant resources for re-execution, ensuring the integrity and continuity of the task.

6. The cloud-based distributed discrete event simulation platform according to claim 1, wherein: The experiment manager supports users to graphically configure experimental scenarios, including input parameter and output parameter settings and simulation observation number settings, and can automatically count all experimental groups and observation numbers, facilitating the generation and management of large-scale automated simulation tasks.

7. The cloud-based distributed discrete event simulation platform according to claim 1, wherein: After all tasks are completed, the Master node can automatically perform statistics and analysis on the simulation data reported by all Slave nodes, including calculating statistical indicators such as the mean, maximum, minimum, standard deviation and confidence interval of each experiment, and automatically summarize and generate a structured detailed report.

8. The cloud-based distributed discrete event simulation platform according to claim 1, wherein: The final simulation report generated by the Master node can be notified to the user through the system interface or pushed to the user in the form of an email. The report content includes task execution details, resource utilization, main performance indicators and exception handling records, so that the user can conduct a comprehensive evaluation and retrospective analysis of the simulation results.

Images