The Time Overhead Inside the Data Sync Procedure of Personal Cloud Storage Service

摘要

In recent years, Personal Cloud Storage (PCS) Service like Dropbox has achieved a great business success and has attracted a large interest from academia. However, considering the security of users' data, the PCS applications communicate with their servers by HTTPS protocol, which makes it hard for researchers to understanding how data sync works at the level of HTTP connection. As a result, the time overhead inside the procedure is unclear from the client view. Different from the previous work, by dynamic injection to the client and active measurement, we verify what's the role of each HTTP connection in the network communication procedure when the PCS applications synchronize the data to the server. As a result, we reveal the time overhead inside the procedure, and characterize the time overhead generated from both the processing and the network. Interestingly, the time overhead is different between the data upload and download in a real network with proxy servers. We believe that the findings could be an enlightenment for developing a more cost-effective PCS application.