CN-120407100-B - Engine scheduling method of digital model oriented to mixed granularity

CN120407100BCN 120407100 BCN120407100 BCN 120407100BCN-120407100-B

Abstract

The invention relates to an engine scheduling method of a digital model oriented to mixed granularity, which comprises the following steps of S1, model sequencing, constructing a model interaction relation matrix DSM, carrying out hierarchical sequencing on a simulation model, adjusting the matrix sequence to determine the execution priority of the model, S2, designing a task scheduling strategy, adopting an improved centralized scheduling strategy, globally controlling task allocation by a scheduling host, actively reporting a load by a node and executing partial task scheduling, S3, executing a mixed task scheduling algorithm based on the priority based on a frame period and a multi-rate distributed simulation node synchronization mechanism of system simulation, wherein the priority sequence is periodic task > sporadic task > background task. The invention can realize the engine scheduling of the digital model under the mixed granularity.

Inventors

QU WEI
YIN JIANFENG
SU HUIFANG
CANG JIE
WANG TIANZHE
WANG QIAN
WANG JINWANG
GUO JING

Assignees

中国空间技术研究院

Dates

Publication Date: 20260512
Application Date: 20250324

Claims (3)

1. The engine scheduling method of the digital model facing the mixed granularity is characterized by comprising the following steps of: s1, model ordering, namely constructing a model interaction relation matrix DSM, carrying out hierarchical ordering on simulation models, and adjusting the matrix order to determine the execution priority of the models; The method for classifying the simulation models comprises the following steps of: S11, constructing a model interaction relation matrix, taking a model without input as a first stage in a grading manner, deducing a subsequent model step by step, wherein the grading method comprises the following steps: taking the model without other model input as a first stage, taking a data receiving model of the first stage model as a second stage model, and so on, wherein the classified models do not participate in subsequent classification; Step S12, adjusting the matrix sequence, and completing model execution sequence planning through source point and sink point positioning and closed loop processing, wherein the closed loop processing comprises the following steps: performing fixed coupling operation on the model with the directed loop, regarding the coupled model as a sub-module by using normalization operation, and selecting source points from the sub-module to continue sequencing; Step S2, designing a task scheduling strategy, adopting an improved centralized scheduling strategy, globally controlling task allocation by a scheduling host, and actively reporting load by a node and executing partial task scheduling; In step S2, the improved centralized scheduling policy includes: Step S21, periodically or threshold triggering type actively reporting load information to a dispatching host by a node; step S22, the dispatching host computer lowers part of dispatching decision weights to the nodes, and the nodes execute task allocation according to the local state; the threshold triggering type active load information reporting method comprises the following steps: when the load variable of the node exceeds a preset threshold value, the load variation exceeds a difference threshold value or a request of a dispatching host is received, the node actively reports load information; And S3, executing a mixed task scheduling algorithm based on priority by a frame period and multi-rate distributed simulation node synchronization mechanism based on system simulation, wherein the priority order is periodic tasks, sporadic tasks and background tasks.
2. The method according to claim 1, wherein in step S3, the multi-rate distributed emulated node synchronization mechanism comprises: step S31, taking a frame period of system simulation as a logic time unit, wherein the step length of a node is an integer multiple of the frame period; step S32, the node starts or pauses operation according to the time stamp of the received data and the logic time relation.
3. The method of claim 2, wherein the frame period of the system simulation is the greatest common divisor of all node step sizes, and the scheduling host only performs data interaction with the nodes meeting the step size requirement.

Description

Engine scheduling method of digital model oriented to mixed granularity Technical Field The invention relates to the technical field of simulation and modeling, in particular to an engine scheduling method of a digital model oriented to mixed granularity. Background Along with the continuous increase of simulation demands of large-scale systems in the fields of aviation, aerospace, ships, weapons and the like, the complexity and the scale of a digital model are exponentially increased. The simulation system needs to integrate different sources, different granularities (such as a high-precision physical model and a low-precision behavioral model) and different-scale physical models to support multidisciplinary collaborative simulation. In this context, the task scheduler of the simulation engine becomes a core challenge for efficient system operation. In the prior art, the simulation modeling and scheduling engine is an important component in large-scale system simulation in the fields of aviation, aerospace, ships, weapons and the like. The method is characterized in that the method is used for determining the execution sequence among the models, formulating an engine scheduling mechanism and realizing the synchronous scheduling of the models with different priorities and different granularity, which is a problem to be solved urgently. Disclosure of Invention In view of the above technical problems, the present invention provides an engine scheduling method for a digital model with mixed granularity, which implements engine scheduling of the digital model with mixed granularity. The technical solution for solving the technical problem of the invention is that an engine scheduling method of a digital model facing to mixed granularity comprises the following steps: s1, model ordering, namely constructing a model interaction relation matrix DSM, carrying out hierarchical ordering on simulation models, and adjusting the matrix order to determine the execution priority of the models; Step S2, designing a task scheduling strategy, adopting an improved centralized scheduling strategy, globally controlling task allocation by a scheduling host, and actively reporting load by a node and executing partial task scheduling; And S3, executing a mixed task scheduling algorithm based on priority by a frame period and multi-rate distributed simulation node synchronization mechanism based on system simulation, wherein the priority order is periodic tasks, sporadic tasks and background tasks. According to one technical scheme of the invention, in step S1, the simulation models are ranked in a grading manner, and the method specifically comprises the following steps: s11, constructing a model interaction relation matrix, taking a model without input as a first stage in a hierarchical mode, and deducing a subsequent model step by step; and step S12, adjusting the matrix sequence, and completing model execution sequence planning through source point and sink point positioning and closed loop processing. According to an embodiment of the present invention, in step S11, the classification method includes: And taking the model without other model inputs as a first stage, taking a data receiving model of the first stage model as a second stage model, and so on, wherein the classified models do not participate in subsequent classification. According to an aspect of the present invention, in step S12, the closed loop processing includes: and performing fixed coupling operation on the model with the directed loop, regarding the coupled model as a sub-module by using normalization operation, and selecting source points from the sub-module to continue sequencing. According to an aspect of the present invention, in step S2, the improved centralized scheduling policy includes: Step S21, periodically or threshold triggering type actively reporting load information to a dispatching host by a node; and S22, the dispatching host computer lowers part of dispatching decision weights to the nodes, and the nodes execute task allocation according to the local state. According to one technical scheme of the invention, the threshold triggering type active load information reporting method comprises the following steps: when the load variable of the node exceeds a preset threshold value, the load variation exceeds a difference threshold value or a request of a dispatching host is received, the node actively reports the load information. According to one embodiment of the present invention, in step S3, the multi-rate distributed simulation node synchronization mechanism includes: step S31, taking a frame period of system simulation as a logic time unit, wherein the step length of a node is an integer multiple of the frame period; step S32, the node starts or pauses operation according to the time stamp of the received data and the logic time relation. According to one technical scheme of the invention, the frame period of the system simulation is the g