The Research And Implementation Of Large-scale Distributed Network Simulation System And Semi-physical Simulation

Network simulation technology is widely used in the study of network problems,but most of the traditional network simulation technologies are limited to single-host simulation,the simulation scale is limited,and the single-host simulation cannot be easily scaled to multiple simulation hosts.Moreover,traditional network simulation techniques use their own independently implemented network protocol stacks,which are far from the physical node operating environment and cannot simulate the actual program operating environment.In this thesis,based on the project and actual requirements,we study the distributed discrete simulation algorithm,design a distributed discrete simulation algorithm that can dynamically adapt to the network latency changes between physical simulation hosts distributed across the network.And based on this algorithm,we combine with container virtualization technology,which simulates the running environment of physical nodes on physical hosts,and design and implement a set of simulation environment to meet the seamless integration with the running programs of physical nodes.Then we design and implement a large-scale distributed network simulation framework to meet the seamless integration of simulation environment and physical node running procedures and support semi-physical simulation verification.First,this thesis designs a distributed conservative synchronous discrete event simulation algorithm DT-DDNS(Dynamic Tree-Distributed Discrete Network Synchronization)based on the central node election mechanism and dynamic tree topology mechanism to meet the requirements of distributed network simulation,combining the traditional tree distributed discrete event simulation algorithm,through which the distributed network simulation system can support multiple physical simulation hosts to perform network simulation in a distributed manner and can adapt to the network delay variation between physical simulation hosts across networks.The distributed network emulation system can support distributed network emulation by multiple physical emulation hosts,and can adapt to the network delay variation among physical emulation hosts across networks.Considering the strong dependence of distributed discrete-event simulation synchronization on the central node,this thesis introduces a distributed election algorithm based on the traditional distributed tree synchronization algorithm,so that the synchronization algorithm can elect the central node in the cluster of simulation hosts,and the central node is responsible for the discrete-event synchronization progress decision in distributed simulation.In addition,the election algorithm dynamically changes the central node after the central node becomes overloaded or offline,enhancing the flexibility of distributed network simulation.For the scenario that the network communication delay between physical simulation hosts changes dynamically,this thesis introduces a dynamic topology mechanism in DT-DDNS,which dynamically changes the event synchronization path in response to the change of network delay of physical simulation hosts to ensure the synchronization efficiency.In this thesis,in order to compare the synchronization efficiency and performance overhead of DT-DDNS and traditional distributed conservative synchronous discrete simulation algorithms under different network conditions,a numerical simulation environment is built to simulate different node sizes and different network delay conditions to verify that the DT-DDNS algorithm can effectively guarantee the simulation synchronization efficiency under distributed network simulation conditions.Based on the DT-DDNS algorithm,this thesis designs and implements a large-scale distributed virtualized network simulation framework by combining container virtualization technology.The framework is able to use containers to create emulation nodes to simulate the actual operating environment and network protocol stack of physical nodes and run real network programs seamlessly.The framework also combines the virtual container with the physical channel model to provide controllable physical channels for the emulation nodes and achieve flexible emulation network topology control.To address the network protocol verification of physical nodes,the framework implements a semi-physical gateway architecture,creates mapping nodes of physical nodes in the simulation network,collects simulation message sequences from mapping nodes interacting with other simulation nodes during the simulation process,and replay the collected message sequences to physical nodes in the semi-physical simulation environment to verify the consistency of protocol implementation of physical nodes and the performance of physical nodes and improve the simulation results This thesis combines the project scenario with a large-scale distributed simulation environment.In this thesis,we combine the project scenario to functionally validate the large-scale distributed virtualized network framework and semi-physical gateway architecture.The validation results show that the system can well support the discrete simulation of largescale networks and can verify the physical node functions and performance.