A Measurement Study of Internet File Transfer Traffic
David J. Ewing, Richard S. Hall, Michael F. Schwartz · 2001
The lack of a topology-based information distribution mechanism in wide area networks causes users to waste a large amount of bandwidth on repeat transfers of widely accessed files. As a first step towards designing such a mechanism, in this paper we present measurements of the Internet File Transfer Protocol, gathered over two weeks at the main gateway to the University of Colorado network. We found that nearly 40% of all transfers were duplicate transmissions, accounting for over 54% of the file transmission traffic. A small proportion of files were significantly larger and more frequently transferred, with 6.30% of the files accounting for 19.30% of the transfers. 8.21% of duplicate file transmissions were caused by user errors when transferring binary data, underscoring the need to insulate users from such details. We also found that file transfers occurred in only 44% of the FTP connections. Other connections were probably directory-only requests, underscoring the need for better resource discovery support in the Internet. We present measurements of a number of other statistics as well, including distributions of file sizes, file types, peak transfer times, and sources and destinations.