Cloud-computing-based distributed simulation platform elastic scheduling method and system

Abstract

The application discloses a cloud-computing-based distributed simulation platform elastic scheduling method and system, and relates to the technical field of computer systems.The cloud-computing-based distributed simulation platform elastic scheduling method and system comprise a cross-field dynamic dependency processing module, a real-time task ultrafast scheduling module, a hybrid cloud resource adaptation module and an elastic scheduling core engine; the cross-field dynamic dependency processing module is used for realizing dynamic processing of dependency relationships in multi-field collaborative simulation, cross-domain collaborative allocation of resources and breakpoint self-recovery of task chains.The cloud-computing-based distributed simulation platform elastic scheduling method and system realize dynamic analysis of multi-field dependency relationships, cross-domain collaborative allocation of resources and breakpoint self-recovery of task chains through the cross-field dynamic dependency processing module, solve the problem of rigid cross-field dependency processing mechanism in the prior art, reduce cross-field data synchronization delay, and improve resource utilization and stability of task chains.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer systems, in particular to an elastic scheduling method and system for a distributed simulation platform based on cloud computing. BACKGROUND

[0002] With the rapid development of digital twin cities and other fields, a distributed simulation platform based on cloud computing plays an important role in city fine management and intelligent decision-making. The platform needs to realize multi-field collaborative simulation and construct a virtual mirror that is synchronized with physical entities to support city-level dynamic deduction.

[0003] However, the prior art has significant defects in cross-field dependency processing and real-time task scheduling. In cross-field dependency processing, it relies on artificial preset static rules, which cannot adapt to dynamic changes in the scene, resulting in data synchronization delay.

[0004] Resource allocation adopts a field resource independent allocation mode, which cannot be dynamically adjusted, causing resource waste. The task chain formed by multi-field collaboration lacks a breakpoint self-healing mechanism, and any link delay will trigger a chain reaction, affecting decision accuracy.

[0005] In real-time task scheduling response, the existing system mostly adopts a resource load threshold trigger mode. The full-link delay from load exceeding to resource readiness is too long, which cannot meet the timeliness requirements of real-time tasks.

[0006] The resource pool lacks a preheating mechanism and is difficult to cope with the sudden and short peak demand of real-time tasks for resources.

[0007] In a multi-task concurrent scenario, there is a lack of fine-grained high-priority task preemption mechanism, resulting in high-priority task resource acquisition delay, and the existing preemption strategy is easy to cause the loss of interrupted task state, further exacerbating resource waste and delay.

[0008] In addition, in a hybrid cloud environment, there is a contradiction between cross-domain resource scheduling and security isolation, and the encryption performance loss of sensitive data transmission is large, affecting the overall performance and security of the distributed simulation platform.

[0009] Therefore, there is an urgent need for an elastic scheduling method and system that can solve the above problems and realize cross-field dynamic dependency processing, real-time task rapid scheduling, and hybrid cloud resource security adaptation. SUMMARY

[0010] The present application aims to at least solve one of the technical problems in the prior art, and provides an elastic scheduling method and system for a distributed simulation platform based on cloud computing, which can solve the above problems.

[0011] In order to achieve the above object, the application provides the following technical scheme: the elastic scheduling system of the distributed simulation platform based on cloud computing, comprising a cross-domain dynamic dependency processing module, a real-time task ultra-fast scheduling module, a hybrid cloud resource adaptation module and an elastic scheduling core engine;

[0012] The cross-domain dynamic dependency processing module is used for realizing dynamic processing of dependency relationships, cross-domain collaborative allocation of resources and breakpoint self-recovery of task chains in multi-domain collaborative simulation.

[0013] The real-time task ultra-fast scheduling module is used for realizing predictive resource warming, microsecond-level scheduling response and lossless preemption of high-priority tasks of real-time simulation tasks.

[0014] The hybrid cloud resource adaptation module is used for realizing hierarchical scheduling of private cloud and public cloud resources and hardware acceleration and encryption transmission of sensitive data according to task data sensitivity.

[0015] The elastic scheduling core engine is connected with the cross-domain dynamic dependency processing module, the real-time task ultra-fast scheduling module and the hybrid cloud resource adaptation module respectively, and is used for overall planning of collaborative work of the modules and dynamic adjustment of scheduling strategies.

[0016] Preferably, the cross-domain dynamic dependency processing module comprises:

[0017] The semantic dependency parser is used for automatically analyzing input-output parameter correlation of multi-domain simulation tasks and generating a real-time updated dependency relationship graph.

[0018] The cross-domain resource pool collaborative scheduler is used for dynamically adjusting resource quotas according to dependency relationships between domains and realizing cross-domain reuse of idle resources.

[0019] The task chain self-recovery controller is used for monitoring states of each link of a task chain formed by multi-domain collaboration and triggering breakpoint rescheduling when a certain link is delayed.

[0020] Preferably, the semantic dependency parser comprises a domain interface registration table, a dependency relationship graph generator and a dynamic rule engine.

[0021] The domain interface registration table is used for storing data interface information registered when each domain simulation task is started.

[0022] The dependency relationship graph generator is used for constructing a dynamic dependency network based on interface parameter correlation, and weights of the dynamic dependency network are adjusted in real time according to task types.

[0023] The dynamic rule engine is used for automatically updating data synchronization frequency when data demand changes are triggered in a burst scenario.

[0024] Preferably, the cross-domain resource pool cooperative scheduler comprises a domain resource profile library, a dependency sensitivity calculator and a resource elasticity allocation algorithm.

[0025] The domain resource profile library is used to record resource load information of each domain in real time.

[0026] The dependency sensitivity calculator is used to quantify the resource dependency strength between domains.

[0027] The resource elasticity allocation algorithm is used to allocate computing power from the idle resource pool of other domains when a domain needs to be expanded, while retaining the redundant resources required by the core tasks of other domains.

[0028] Preferably, the task chain self-healing controller comprises a link state monitoring probe, a delay conduction simulator and a breakpoint rescheduling engine.

[0029] The link state monitoring probe is used to collect the completion progress of each link of the task chain.

[0030] The delay conduction simulator is used to predict the impact of a link delay on subsequent links.

[0031] The breakpoint rescheduling engine is used to start the backup resource pool for computing power expansion when a link is delayed, and to pause the subsequent links until the breakpoint is restored and the progress is synchronized.

[0032] Preferably, the real-time task ultra-fast scheduling module comprises:

[0033] A predictive resource warm-up pool is used to predict resource demand based on historical load data and prepare resources in advance.

[0034] A microsecond-level scheduling engine is used to achieve sub-second response of real-time tasks.

[0035] A priority preemption arbitrator is used to achieve lossless preemption of low-priority tasks by high-priority tasks.

[0036] Preferably, the predictive resource warm-up pool comprises a historical load prediction model, a resource snapshot manager and a warm-up node pool.

[0037] The historical load prediction model is used to predict resource demand in subsequent periods based on historical load data.

[0038] The resource snapshot manager is used to generate common simulation environment images.

[0039] The warm-up node pool is used to start standby nodes before predicted peak and keep them in sleep-wake state.

[0040] Preferably, the hybrid cloud resource adaptation module comprises:

[0041] A cloud resource hierarchical scheduler is used to dynamically select private cloud or public cloud resources according to task data sensitivity;

[0042] A hardware acceleration encryption gateway is used to realize hardware-level encryption of sensitive data in cross-domain transmission and computing process;

[0043] The cloud resource hierarchical scheduler includes a data sensitivity label library, a cloud resource security level matrix, and a cross-cloud link optimizer. The data sensitivity label library marks the sensitivity of task data, such as military deployment simulation as top secret and public people flow heat map as public.

[0044] The cloud resource security level matrix matches cloud resources according to data sensitivity. Top secret level tasks are fixedly scheduled to private cloud nodes with hardware encryption cards, cross-cloud transmission is prohibited, public level tasks are automatically allocated to public cloud elastic nodes, and can be temporarily expanded to edge nodes during peak period. The cross-cloud link optimizer optimizes cross-cloud data transmission link and reduces transmission delay.

[0045] The hardware acceleration encryption gateway includes an SGX encryption chip, a national secret SM algorithm accelerator, and a transmission layer encryption tunnel manager. The SGX encryption chip provides a hardware-level secure environment, the national secret SM algorithm accelerator improves encryption efficiency, and the transmission layer encryption tunnel manager establishes an encrypted transmission tunnel, so that the encryption time of sensitive data such as emergency plan parameters in cross-domain transmission and computing process is reduced to 1 / 5 of software encryption.

[0046] Preferably, the microsecond-level scheduling engine includes a load monitoring high-frequency collector, a resource state memory database, and an ultra-speed decision maker.

[0047] The load monitoring high-frequency collector is used to monitor node load at high frequency and write data into the resource state memory database in real time.

[0048] The ultra-speed decision maker is used to directly call resources from the preheating pool when detecting real-time task triggering, and to compress resource ready time through kernel-level scheduling.

[0049] The priority preemption arbitrator includes a task priority registration table, a state saving buffer, and a resource switching controller.

[0050] The task priority registration table is used to preset a task priority matrix.

[0051] The state saving buffer is used to store memory data of low-priority tasks in real time when high-priority tasks are triggered.

[0052] The resource switching controller is used to release the resources of pre-empted nodes and allocate them to high-priority tasks. After the high-priority tasks are completed, the low-priority tasks are restored through the buffer data.

[0053] The elastic scheduling method of the distributed simulation platform based on cloud computing comprises the following steps:

[0054] S1, cross-domain dynamic dependency processing step: dynamically processing the dependency relationship in multi-domain collaborative simulation, realizing cross-domain collaborative allocation of resources and self-recovery of task chain breakpoints;

[0055] The cross-domain dynamic dependency processing step comprises:

[0056] The semantic dependency analysis step: automatically analyzing the input-output parameter correlation of the multi-domain simulation task, and generating a real-time updated dependency graph;

[0057] The cross-domain resource pool collaborative scheduling step: dynamically adjusting the resource quota according to the dependency relationship between domains, realizing cross-domain reuse of idle resources;

[0058] The task chain self-recovery step: monitoring the state of each link of the task chain, and triggering breakpoint rescheduling when a link is delayed;

[0059] S2, real-time task ultra-fast scheduling step: pre-heat resources for real-time simulation tasks, microsecond-level scheduling response and lossless preemption of high-priority tasks;

[0060] The real-time task ultra-fast scheduling step comprises:

[0061] The pre-heat resource step: based on historical load data, the resource demand is predicted, and the resource is prepared in advance;

[0062] The microsecond-level scheduling step: realizing sub-second level response of real-time tasks;

[0063] S3, hybrid cloud resource adaptation step: according to the sensitivity of task data, the private cloud and public cloud resources are hierarchically scheduled, and the sensitive data is encrypted and transmitted by hardware acceleration;

[0064] The hybrid cloud resource adaptation step comprises:

[0065] The cloud resource hierarchical scheduling step: dynamically selecting private cloud or public cloud resources according to the sensitivity of task data;

[0066] The hardware acceleration and encryption transmission step: the sensitive data is encrypted at the hardware level during cross-domain transmission and calculation;

[0067] S4, elastic scheduling core control step: overall coordination of the above steps, dynamic adjustment of the scheduling strategy according to real-time monitoring data.

[0068] Compared with the prior art, the beneficial effects of the present application are:

[0069] 1. The cloud computing-based distributed simulation platform elastic scheduling method and system, the cross-domain dynamic dependency processing module, realizes the dynamic analysis of multi-domain dependency, the cross-domain collaborative allocation of resources and the breakpoint self-recovery of task chain, solves the rigid problem of cross-domain dependency processing mechanism in the prior art, reduces the cross-domain data synchronization delay, and improves the resource utilization and the stability of the task chain.

[0070] 2. The cloud computing-based distributed simulation platform elastic scheduling method and system, the real-time task ultra-fast scheduling module realizes the prediction resource preheating, the microsecond-level scheduling response and the lossless preemption of high-priority tasks, significantly improves the scheduling response speed of the real-time task, meets the timeliness requirement of the real-time simulation task, and avoids the problems of high-priority task resource acquisition delay and interrupted task state loss.

[0071] 3. The cloud computing-based distributed simulation platform elastic scheduling method and system, the hybrid cloud resource adaptation module realizes the hierarchical scheduling of private cloud and public cloud resources according to task data sensitivity, and realizes the encryption transmission of sensitive data through the hardware acceleration encryption gateway, which reduces the encryption performance loss while ensuring data security, and solves the contradiction between security and performance in hybrid cloud resource adaptation.

[0072] 4. The cloud computing-based distributed simulation platform elastic scheduling method and system, the elastic scheduling core engine coordinates the cooperative work of each module and dynamically adjusts the scheduling strategy, so that the whole system can be self-adaptively optimized according to real-time monitoring data, and the elastic scheduling capability of the distributed simulation platform is further improved. BRIEF DESCRIPTION OF DRAWINGS

[0073] The application will be further described below in combination with the drawings and embodiments:

[0074] Figure 1 The application is a cloud computing-based distributed simulation platform elastic scheduling system framework schematic diagram. DETAILED DESCRIPTION

[0075] This part will describe the specific embodiments of the application in detail, the preferred embodiments of the application are shown in the drawings, the drawings are used to supplement the description of the text part, so that people can intuitively and visually understand each technical feature and the overall technical scheme of the application, but it cannot be understood as a limitation on the protection scope of the application.

[0076] In the description of the application, it should be understood that the orientation description, such as up, down, front, back, left, right, etc., is based on the orientation or positional relationship shown in the drawings, which is only for the convenience of describing the application and simplifying the description, and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, therefore it cannot be understood as a limitation on the application.

[0077] In the description of the present application, greater than, less than, more than, etc. are understood as not including the number, above, below, within, etc. are understood as including the number. If the first, second is described only for the purpose of distinguishing technical features, it cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features or implicitly indicating the sequence of indicated technical features.

[0078] In the description of the present application, unless otherwise expressly limited, the words such as setting, installing, connecting, etc. should be broadly understood, and those skilled in the art can reasonably determine the specific meaning of the above words in the present application in combination with the specific content of the technical scheme.

[0079] Please refer to Figure 1 The present application provides a technical scheme: a distributed simulation platform elastic scheduling system based on cloud computing, comprising a cross-domain dynamic dependency processing module, a real-time task ultra-fast scheduling module, a hybrid cloud resource adaptation module and an elastic scheduling core engine, the elastic scheduling core engine is connected with the cross-domain dynamic dependency processing module, the real-time task ultra-fast scheduling module and the hybrid cloud resource adaptation module, for overall planning of the modules and dynamic adjustment of the scheduling strategy according to real-time monitoring data;

[0080] The cross-domain dynamic dependency processing module comprises a semantic dependency parser, a cross-domain resource pool collaborative scheduler and a task chain self-healing controller.

[0081] The semantic dependency parser comprises a domain interface registration table, a dependency relationship graph generator and a dynamic rule engine. In actual application, when each domain simulation task is started, the data interface information is registered to the domain interface registration table, such as weather simulation-> rainfall intensity-> output frequency. The dependency relationship graph generator constructs a dynamic dependency network based on the correlation of these interface parameters, and the weight of the network is adjusted in real time according to the task type. When a sudden scene such as a traffic accident triggers a change in data demand, the dynamic rule engine automatically updates the synchronization frequency, for example, from 5 minutes / time to 100 ms / time.

[0082] The cross-domain resource pool collaborative scheduler comprises a domain resource portrait library, a dependency sensitivity calculator and a resource elastic allocation algorithm. The domain resource portrait library records the resource load information of each domain in real time, such as energy field GPU utilization and traffic field CPU occupancy.

[0083] The dependency sensitivity calculator quantifies the resource dependency strength between domains, such as the traffic dependency on the weather field increasing to 0.8 during heavy rain.

[0084] When the traffic field needs to expand 30% CPU due to the increase of rainfall intensity, the resource elastic allocation algorithm automatically allocates computing power from the idle resource pool of the energy field, while reserving 20% redundant resources required by the core task of the energy field.

[0085] The task chain self-healing controller includes a link state monitoring probe, a delay conduction simulator, and a breakpoint rescheduling engine. The link state monitoring probe collects the completion progress of each link of the task chain every second, such as whether the pollution diffusion simulation outputs the pollution concentration distribution on time.

[0086] The delay conduction simulator predicts the impact of a certain link delay on subsequent links, such as a 200ms delay in pollution simulation leading to a 1.2 second lag in traffic control decision-making.

[0087] The breakpoint rescheduling engine immediately starts the standby resource pool when a certain link is delayed, expands the computing power of the delayed link, and simultaneously suspends the subsequent links until the breakpoint is restored and the progress is synchronized.

[0088] The real-time task ultra-fast scheduling module includes a predictive resource warm-up pool, a microsecond-level scheduling engine, and a priority preemption arbitrator.

[0089] The predictive resource warm-up pool includes a historical load prediction model, a resource snapshot manager, and a warm-up node pool. The historical load prediction model uses the LSTM time series algorithm to predict resource demand for each time period the next day based on past day load data such as traffic simulation CPU utilization and GPU demand frequency at dawn.

[0090] The resource snapshot manager generates commonly used simulation environment images, such as GPU node snapshots with smoke diffusion algorithms.

[0091] The warm-up node pool starts standby nodes 30 minutes before the predicted peak and maintains a sleep-wake state, reducing the startup time from 90 seconds to 500ms.

[0092] The microsecond-level scheduling engine includes a load monitoring high-frequency collector, a resource state memory database, and an ultra-fast decision maker. The load monitoring high-frequency collector monitors node load such as CPU, memory, and network IO at a frequency of 10ms / second and writes data to the resource state memory database in real time.

[0093] When a real-time task such as traffic signal timing simulation is detected, the ultra-fast decision maker skips the traditional application-approval process and directly calls resources from the warm-up pool. For containerized tasks, it uses kernel-level scheduling methods such as K8s static Pod binding to reduce resource readiness time to within 2 seconds.

[0094] The priority preemption arbitrator includes a task priority registry, a state saving buffer, and a resource switching controller. The task priority registry presets a task priority matrix, such as P0 for earthquake emergency and P3 for daily simulation.

[0095] When a P0 task is triggered, the state saving buffer stores the memory data of P3 tasks in real time within ≤100ms.

[0096] The resource switching controller releases the preempted node resources and allocates to the P0 level task, and restores the P3 level task through the buffer data after the high priority task is completed, without re-computation.

[0097] The hybrid cloud resource adaptation module comprises a cloud resource hierarchical scheduler and a hardware acceleration encryption gateway.

[0098] The cloud resource hierarchical scheduler comprises a data sensitivity label library, a cloud resource security level matrix and a cross-cloud link optimizer, the data sensitivity label library marks the sensitivity of task data, for example, military deployment simulation is classified as top secret, and public pedestrian heat map is classified as public;

[0099] The cloud resource security level matrix matches the cloud resources according to the data sensitivity, the top secret level task is fixedly scheduled to the node of the private cloud with a hardware encryption card, cross-cloud transmission is prohibited, the public level task is automatically allocated to the elastic node of the public cloud, and the edge node can be temporarily expanded during peak period; the cross-cloud link optimizer optimizes the cross-cloud data transmission link and reduces transmission delay;

[0100] The hardware acceleration encryption gateway comprises an SGX encryption chip, a national secret SM algorithm accelerator and a transmission layer encryption tunnel manager, the SGX encryption chip provides a hardware level security environment, the national secret SM algorithm accelerator improves encryption efficiency, and the transmission layer encryption tunnel manager establishes an encryption transmission tunnel, so that the encryption time consumption of sensitive data such as emergency plan parameters in cross-domain transmission and calculation process is reduced to 1 / 5 of software encryption.

[0101] The embodiment of the application further provides a distributed simulation platform elastic scheduling method based on cloud computing, comprising the following steps:

[0102] S1, cross-domain dynamic dependency processing step:

[0103] S11, semantic dependency analysis step: register data interfaces when starting each domain task, construct a dynamic dependency network based on interface parameter correlation, and automatically update data synchronization frequency in a sudden scenario;

[0104] S12, cross-domain resource pool cooperative scheduling step: real-time record each domain resource load, quantize inter-domain resource dependency strength, and call computing power from other idle resource pools when a certain domain needs to be expanded;

[0105] S13, task chain self-healing step: collect the completion progress of each link of the task chain, predict the delay influence, start the standby resource pool expansion and suspend the subsequent link when a certain link is delayed, and wait;

[0106] S2, real-time task rapid scheduling step:

[0107] S21, a predicted resource preheating step: predicting resource demand based on historical load data, generating a simulation environment mirror, starting standby nodes before the predicted peak and keeping them in a "sleep-wake" state;

[0108] S22, a microsecond-level scheduling step: high-frequency monitoring of node load and real-time storage of data, direct calling of resources from the preheating pool when a real-time task is triggered, and compression of resource readiness time through kernel-level scheduling;

[0109] S23, a priority preemption step: a preset task priority matrix, storing low-priority task memory data when a high-priority task is triggered, releasing resources to high-priority tasks, and restoring low-priority tasks after completion;

[0110] S3, a hybrid cloud resource adaptation step:

[0111] S31, a cloud resource hierarchical scheduling step: marking task data according to sensitivity, and matching private cloud or public cloud resources according to sensitivity;

[0112] S32, a hardware acceleration encryption transmission step: using hardware encryption chips and algorithm accelerators to perform hardware-level encryption transmission of sensitive data;

[0113] S4, an elastic scheduling core control step: coordinating the above steps to work together, when detecting that the cross-domain synchronization delay is > 500ms, automatically increasing the atlas update frequency of the semantic dependency resolver; when the high-priority task preemption success rate is < 100%, triggering the rule recalibration of the priority registry.

[0114] Taking the rainstorm weather→traffic-emergency coordination simulation scene as an example, the working process of the application is described:

[0115] Dynamic dependency activation: the weather field triggers a "heavy rain" warning, and the semantic dependency resolver adjusts the synchronization frequency of the traffic field to weather data from 5 minutes / time to 100ms / time, and the dependency graph atlas weight is updated in real time;

[0116] Resource elasticity adaptation: the cross-domain resource pool coordination scheduler detects that the CPU demand of the traffic field increases by 30%, and 20% of the computing power is called from the idle resource pool of the energy field, while the energy core task redundancy is reserved, and the GPU node standby in the preheating pool is started;

[0117] Real-time task response: a sudden traffic accident triggers an emergency simulation P0 level, and the high-speed scheduling engine wakes up 2 GPU nodes from the preheating pool within 1.2 seconds; the priority arbiter saves the low-priority energy consumption simulation state for 80ms; the resource is switched to the emergency simulation task, and the hardware encryption gateway is started to protect the accident scene data;

[0118] Task chain self-recovery: if the pollution diffusion simulation is delayed by 150 ms due to resource fluctuation, the task chain self-recovery controller immediately suspends the subsequent traffic control deduction, starts the backup node expansion, and recovers the pollution simulation progress within 200 ms, ensuring that there is no delay in public warning release.

[0119] The embodiments of the present application are described in detail above with reference to the drawings, but the present application is not limited to the above-described embodiments, and various changes can be made within the knowledge of those skilled in the art in the technical field without departing from the gist of the present application.

Claims

1. A cloud computing-based distributed simulation platform elastic scheduling system, comprising a cross-domain dynamic dependency processing module, a real-time task ultra-fast scheduling module, a hybrid cloud resource adaptation module, and an elastic scheduling core engine; characterized in that The cross-domain dynamic dependency processing module is used to realize dynamic processing of dependency relationships, cross-domain collaborative allocation of resources, and breakpoint self-healing of task chains in multi-domain collaborative simulation; The real-time task ultra-fast scheduling module is used to realize predictive resource warming, microsecond-level scheduling response, and lossless preemption of high-priority tasks for real-time simulation tasks; The hybrid cloud resource adaptation module is used to realize hierarchical scheduling of private cloud and public cloud resources and hardware acceleration and encryption transmission of sensitive data according to task data sensitivity; The elastic scheduling core engine is connected with the cross-domain dynamic dependency processing module, the real-time task ultra-fast scheduling module, and the hybrid cloud resource adaptation module, respectively, and is used to coordinate the collaborative work of the modules and dynamically adjust the scheduling strategy; The cross-domain dynamic dependency processing module comprises: A semantic dependency parser is used to automatically analyze the input-output parameter correlation of multi-domain simulation tasks and generate a real-time updated dependency graph; A cross-domain resource pool collaborative scheduler is used to dynamically adjust resource quotas according to the dependency relationships between domains to realize cross-domain reuse of idle resources; A task chain self-healing controller is used to monitor the state of each link of the task chain formed by multi-domain collaboration and trigger breakpoint rescheduling when a link is delayed; The semantic dependency parser comprises a domain interface registry, a dependency graph generator, and a dynamic rule engine; The domain interface registry is used to store the data interface information registered when each domain simulation task is started; The dependency graph generator is used to construct a dynamic dependency network based on interface parameter correlation, and the weight of the dynamic dependency network is adjusted in real time according to the type of task; The dynamic rule engine is used to automatically update the data synchronization frequency when the data demand changes in a sudden scenario; The cross-domain resource pool collaborative scheduler comprises a domain resource profile library, a dependency sensitivity calculator, and a resource elastic allocation algorithm; The domain resource profile library is used to record the resource load information of each domain in real time; The dependency sensitivity calculator is used to quantify the resource dependency strength between domains; The resource elastic allocation algorithm is used to allocate computing power from the idle resource pool of other domains when a domain needs to be expanded, and reserve redundant resources required by core tasks in other domains; The task chain self-healing controller comprises a link state monitoring probe, a delay conduction simulator, and a breakpoint rescheduling engine; The link state monitoring probe is used to collect the completion progress of each link of the task chain; The delay conduction simulator is used to predict the impact of a link delay on subsequent links; The breakpoint rescheduling engine is used to start the standby resource pool for computing power expansion when a link is delayed, and pause the subsequent links until the breakpoint is restored. 2.The cloud computing-based distributed simulation platform elastic scheduling system of claim 1, wherein: The real-time task ultra-fast scheduling module comprises: A predictive resource warming pool is used to predict resource demand based on historical load data and prepare resources in advance; A microsecond-level scheduling engine is used to realize sub-second response of real-time tasks; A priority preemption arbitrator is used to realize lossless preemption of low-priority tasks by high-priority tasks. 3.The cloud computing-based distributed simulation platform elastic scheduling system of claim 2, wherein: The predicted resource pre-warming pool comprises a historical load prediction model, a resource snapshot manager and a pre-warming node pool; The historical load prediction model is used for predicting resource demand of a subsequent period based on historical load data; The resource snapshot manager is used for generating a commonly used simulation environment image; The pre-warming node pool is used for starting standby nodes before a predicted peak and keeping a sleep-wake state.

4. The cloud-computing-based distributed simulation platform elastic scheduling system of claim 1, wherein: The hybrid cloud resource adaptation module comprises: a cloud resource hierarchical scheduler, which is used for dynamically selecting private cloud or public cloud resources according to task data sensitivity; a hardware acceleration encryption gateway, which is used for realizing hardware-level encryption of sensitive data in cross-domain transmission and calculation processes; The cloud resource hierarchical scheduler comprises a data sensitivity label library, a cloud resource security level matrix and a cross-cloud link optimizer, the data sensitivity label library is used for marking sensitivity of task data, military deployment simulation is classified as top secret, and public people flow heat map is classified as public; The cloud resource security level matrix matches cloud resources according to data sensitivity, top secret level tasks are fixedly scheduled to nodes of the private cloud with hardware encryption cards, cross-cloud transmission is prohibited, and public level tasks are automatically allocated to public cloud elastic nodes, which can be temporarily expanded to edge nodes in a peak period; the cross-cloud link optimizer optimizes a cross-cloud data transmission link to reduce transmission delay; The hardware acceleration encryption gateway comprises an SGX encryption chip, a national secret SM algorithm accelerator and a transmission layer encryption tunnel manager, the SGX encryption chip provides a hardware-level secure environment, the national secret SM algorithm accelerator improves encryption efficiency, and the transmission layer encryption tunnel manager establishes an encrypted transmission tunnel, so that encryption time consumption of sensitive data in cross-domain transmission and calculation processes is reduced to 1 / 5 of software encryption.

5. The cloud-computing-based distributed simulation platform elastic scheduling system of claim 1, wherein: The microsecond-level scheduling engine comprises a load monitoring high-frequency collector, a resource state memory database and an ultrafast decision maker; The load monitoring high-frequency collector is used for high-frequency monitoring of node load and real-time writing of data into the resource state memory database; The ultrafast decision maker is used for directly calling resources from the pre-warming pool when detecting a real-time task trigger, and compressing resource ready time through kernel-level scheduling; The priority preemption arbitrator comprises a task priority registration table, a state saving buffer and a resource switching controller; The task priority registration table is used for presetting a task priority matrix; The state saving buffer is used for real-time storage of memory data of low-priority tasks when a high-priority task triggers; The resource switching controller is used for releasing pre-empted node resources and allocating them to high-priority tasks, and restoring low-priority tasks through the buffer data after completion of the high-priority tasks.

6. A cloud-computing-based distributed simulation platform elastic scheduling method, characterized in that: The method comprises the following steps: S1, a cross-domain dynamic dependency processing step: dynamically processing dependency relationships in multi-domain collaborative simulation to realize cross-domain collaborative allocation of resources and self-recovery of a task chain; The cross-domain dynamic dependency processing step comprises: a semantic dependency analysis step: automatically analyzing input-output parameter correlation of multi-domain simulation tasks to generate a real-time updated dependency graph; a cross-domain resource pool collaborative scheduling step: dynamically adjusting resource quotas according to dependency relationships between domains to realize cross-domain reuse of idle resources; a task chain self-recovery step: monitoring states of each link of a task chain and triggering breakpoint rescheduling when a certain link is delayed. S2, real-time task ultra-fast scheduling step: real-time simulation task is predicted resource preheating, microsecond-level scheduling response and high-priority task lossless preemption; The real-time task ultra-fast scheduling step comprises: Predictive resource preheating step: predict resource demand based on historical load data, and prepare resources in advance; Microsecond-level scheduling step: sub-second response of real-time task is realized; S3, hybrid cloud resource adaptation step: according to the task data sensitivity, the private cloud and the public cloud resource are scheduled in stages, and the sensitive data is transmitted by hardware acceleration encryption; The hybrid cloud resource adaptation step comprises: Cloud resource hierarchical scheduling step: dynamically selecting private cloud or public cloud resources according to task data sensitivity; Hardware acceleration encryption transmission step: sensitive data is encrypted at the hardware level during cross-domain transmission and calculation; S4, elastic scheduling core control step: the above steps work together, and dynamically adjust the scheduling strategy according to real-time monitoring data.

Images