Towards Reliable User Collaboration over Cloud-based File Synchronization System:Dropbox as a Case Study

Abstract

Recent advances in cloud technology have turned the idea of Cloud Storage into a reality. Over the past few years, cloud-based file storage/synchronization systems such as Dropbox, Google Drive and One Drive, have gained tremendous success among Internet users. The cloud storage systems, most notably Dropbox, not only provide reliable file storage but also enable effective file synchronization with diverse user collaborations. It is thus interesting to explore if this cloud-based file storage system can provide reliable collaboration service for the users. In this thesis, we, for the first time, collect samples of content shared on Dropbox and analyze it. We then take our first step to understand the reliable user collaboration over cloud-based file storage systems using the popular Dropbox as a case study. Our observations indicate that users' updates would be discarded without any warning in a classic collaborative file editing session on Dropbox like file storage systems. In particular, to ensure a reliable service, the users need to wait for over 40 seconds between two updates when their uploading capacity is lower than 100 KBytes/sec. This already exceeds the auto-saving interval of many real world applications and severely hinders the system applicability. We further investigate the root causes of this phenomenon and find that such a problem is due to the three-layer design(i.e. user client, cloud virtual machines and cloud storage) of cloud storage systems. To address this problem of losing updates in multi-user collaborative file editing session, we propose an enhancement to the system that keeps track of and saves all the user updates locally and provides the opportunity for the recovery of any historical updates. Our proposed enhancement ensures that none of the user updates will be lost during a file editing session irrespective of the time interval between the updates and thereby making Dropbox user collaboration more resilient and less prone to potential data loss.