In this talk, we focus on the delay performance of cloud storage, particularly that of Amazon S3. We first present a summary of our observations based on extensive measurements with various object sizes. In line with early studies, our measurements indicate that although the mean delay performance of Amazon S3 is quite robust, there is a high delay spread in both retrieving from and placing objects into the cloud. Furthermore, we establish that at the expense of additional storage and bandwidth, we can significantly improve the delay performance by applying forward error correction (FEC) techniques when multiple threads are utilized in parallel. We show that FEC introduces a trade off between service delays and queueing delays, requiring load adaptive strategies. For further analysis of this trade off, we resort to Queueing Theory and tackle down an extremely non-trivial queue model that incorporates multiple servers and FEC. We define different types of scheduling policies referred to as repeat-allowed and repeat-forbidden policies. We develop schemes that are throughput optimality and delay optimal when the time for retrieving/placing objects is exponentially distributed.
Shengbo Chen received his B. E. and M. E. degrees in the department of Electronic Engineering from Tsinghua University, Beijing, China in 2006 and 2008, respectively. He is currently a Ph.D. candidate in the Department of Electrical and Computer Engineering at The Ohio State University. His research interests include resource allocation in rechargeable sensor networks, scheduling policy design in the smart grid and cloud computing. He is a student member of the IEEE.