Research And Implement Of Data Collection System For Parallel Discrete Event Simulation

PDES (Parallel Discrete Event Simulation) only concerns about the imfluential events on the simulation results, so it has high execution efficiency, which is attracting more and more attention in the field of analysis simulation. In order to carries on the effective analysis and evaluation of the result of simulation, it is necessary to collect the relevant data in the process of simulation. As PDES using the event-based modeling method and optimistic time management, its operation mechanism is different from the traditional real-time simulation and DIS (Distributed Interactive Simulation), which make the existing strategies for simulation data collection are not effective for PDES.This paper studies deeply the key technologies of data collection in PDES based on the characteristics of PDES and the analysis of the relevant technologies of the existing simulation data collection system.In PDES, the changes of simulation objects' state are triggered by the discrete events and that the states updating time is uncertain, so the paper proposes a method of gaining the simulation objects' state data through the "object proxy" mechanism, in which the data are sent to the data collection system only when the simulation objects' state are updated. This doesn't only ensure the integrity of data collection, but also effectively reduces the redundant data collected, which improves the efficiency of the system.The speed of simulation object state updating sometimes may exceed the real-time processing capability of the data collection system. To ensure that the data collection system is not losing data, the paper proposes a "Multi-threaded Event Buffer Pool" strategy. In this strategy, a thread receives the messages of updating the objects' state data and sends them to the pending event buffer pool, while another thread takes responsible for processing the events fetched from the pool. The strategy can be used to effectively enhancing the data processing capability of the data collection system and assure the integrity of data collection.The data types in PDES not only including the simple ones but also the complex ones such as position, class and so on, the paper proposes a "Parallel Simulation Metadata" sextuple to express the collected data. This method is not only conducive to the unity of documents and database storage, but it also helps identifying and reconstructing the preserved data, and then enhancing the generality and readability of the collected data.To effectively analyze and evaluate the results of the simulation, the analysts often need to repeatedly study the simulation process or some link after simulation. Aiming at PDES being event-driven, the paper proposes the design for playbacking based on data recorded file and message file respectively. The first method only need to parse the data file, and read the data records sequentially, which is easy to control the process of playbacking. The second method will change the messages into events, and then process them through the event schedule mechanism, which is advantageous for simulation application system. Based on the above, a PDES data collection system is designed and implemented. This system not only supports users visibly choosing the needed data on the interface, but also supports configuring the collected contents through the file menu. The system can transmit real-time data; store the data to file or database, and playback in two ways. The results of test show that the system performances greatly.