Understanding the Distribution of File Transmission Duration in the Web

It is well known that the distribution of files transmission duration in the Web is heavy-tailed [9]. This paper attempts to understand the reasons for this phenomenon by isolating the three major factors influencing the transmission duration: file size, network conditions and server load. We present evidence that the transmissionduration distribution of the same file from the same server to the same client in the Web is Pareto and therefore heavy tailed. Furthermore, transmission delay for a specific client/server pair is not significantly affected by file size: all files transmitted from the same server to the same client have a very similar transmission duration distribution, regardless of their size. We use simulations to estimate the impact of network conditions and server load on the transmission-duration distribution. When the server and the client are on the same network, the transmission duration distribution of each file is usually Pareto as well (for server files and client requests that are distributed in a realistic way). By examining a wide-area network situation, we conclude that the network conditions do not have a major influence on the heavy-tailed behavior of transmission-duration distribution. In contrast, the server load is shown to have a significant impact on the high variability of the distribution. ∗This research was supported by the Israel Science Foundation (grant number 105/01) and by Technion V.P.R. fund – I. Goldberg fund for electronics Research. †Dept. of Computer Science, Technion, Haifa, 32000, Israel, E-mail: [email protected] ‡Dept. of Computer Science, Technion, Haifa, 32000, Israel, E-mail: [email protected]