Research And Implementation Of Multi-threaded Segmental Download Algorithm

As the office gradually presents the trend of intellectualization,the transmission of information at work is gradually transformed into digital form,and people’s demand for downloading files is gradually increasing.How to download files at high speed has also become a technology hotspot in recent years.Based on the LAN download system of a company,this paper studies a multi-thread segmented download algorithm based on download speed prediction to improve the file download speed,increase the disaster resistance of the download process,and save download time.The main work completed in this paper includes:First of all,taking time series prediction as the research object,a time series prediction method named PA-LSTM is constructed to predict the download speed.In this method,CEEMDAN is used to preprocess the data,and PSO algorithm is used to adjust the weight of temporal attention mechanism.The experimental data shows that the PA-LSTM method is superior to the current superior ConvLSTM method in terms of the time series prediction RMSLE and MAE evaluation indicators,and can more accurately predict the change of download speed in the future.The RMSLE index is 0.009 lower than ConvLSTM,and the MAE index is 0.003 lower than ConvLSTM.At the same time,with the actual download file as the background,a large number of real download data were collected and filtered manually,and the download speed data set under different bandwidth and network environment was constructed,which included 8 categories and 7534 download speed data for PA-LSTM training and testing.Secondly,a new multi-threaded segmented download method is constructed.In this method,different threads run different download links.For each download link,PA-LSTM is used to predict the average download speed in the future,and gray model is used to solve the problem of data prediction cold start.According to the prediction results,different numbers of data blocks are allocated to the downloaded files in proportion to reduce the time of thread switching and data block allocation during the download process.This download method can make full use of the actual download speed of each download link,thus greatly improving the overall file download speed,saving download time,and meeting the future file download trend.Finally,on the basis of the PA-LSTM download speed prediction method and the multithreaded segmented download method,this paper uses the relevant technologies such as Electron,JavaScript,and Vue to design and develop a multithreaded segmented download system based on the download speed prediction,and implements the download chain download,BT download,download monitoring,download queue storage and other related functions.The system is simple to operate,practical,and has a wide range of use,and has certain use value.