Load Optimization And Application Of Flight Data Processing System Based On LSTM

In recent years,my country’s demand for aviation flight has grown rapidly.The rapid increase in the types and numbers of aircraft has led to a sharp increase in the amount of flight data.In addition,air traffic affairs have higher requirements for safety and real-time,which has become a new challenge.Improving the concurrency of server clusters is one of the most effective means to solve real-time and reliability.This thesis is based on the design and implementation of flight data processing system,and based on the current problems of flight data processing system clusters in terms of load balancing,an optimization plan is proposed.The research in this thesis mainly includes the following contents:(1)Aiming at the problem of uneven load distribution caused by the built-in load balancing algorithm of Nginx,this thesis proposes a dynamic load balancing scheme based on Long ShortTerm Memory(LSTM)network,which uses the LSTM algorithm to request the number of connections to the server Make predictions,and calculate load weights based on the server’s real-time performance indicators and load information.The program comprehensively considers the server’s performance indicators,real-time load information,and the suddenness of server traffic to ensure real-time and stability for the system.(2)In this thesis,the optimal combination weighting method is used to determine the weight coefficients between the load indicators,and then the real-time load indicators of the cluster are collected periodically.After the calculated weight coefficients,the dynamic routing table of the load balancer is updated by consul+upsync.(3)The load balancing algorithm that combines LSTM and the optimal combination weighting method is applied to the actual air traffic control automation system project,and the key functions in the system are summarized,and the actual effect diagram is supplemented for display.(4)In order to verify the feasibility and effectiveness of the load optimization algorithm proposed in this thesis,response time and throughput are selected as evaluation indicators,and the Apache JMeter performance testing tool is used to test the Nginx built-in round-robin scheduling algorithm,weighted round-robin algorithm,and source address.Hash scheduling algorithm and the dynamic load optimization algorithm proposed in this thesis are compared for performance testing.The simulation results show that compared with load balancing algorithms such as weighted round-robin and source address hashing,the improved algorithm proposed in this thesis not only responds to concurrent requests of the cluster The time was reduced by about 20%,and the throughput rate of the cluster was increased by 17%.And applying this algorithm to the flight data processing system,the improved system has improved response time and throughput.