Rozier wins IEEE best paper for deduplication research


CSL graduate student Eric Rozier and CSL Director William Sanders won the Best Paper Award at the 30th IEEE Symposium on Reliable Distributed Systems, which took place Oct. 5-7 in Madrid.

Eric Rozier
Eric Rozier

Rozier, a graduate student in computer science, was the primary author of the paper, titled, “Modeling the Fault Tolerance Consequences of Deduplication.” Rozier and Sanders co-authored the paper along with Pin Zhou, Nagapramod Mandagere, Sandeep Uttamchandani and Mark L. Yakushev.

For the symposium, researchers submitted original research papers about distributed system design, development and evaluation. The accepted papers were presented during the three-day event.

Rozier and Sanders’ paper focuses on deduplication, which is the process of trying to fit large amounts of information in a distributed system into a smaller space by storing data only once, essentially eliminating any copies. Deduplication can be a problem for large scale distributive storage systems because it reduces reliability.

The team developed techniques to analyze this problem, and then did an analysis of a real system. They found that few errors occurred in the system, but when errors did occur, the magnitude of the impact was very large.

Rozier explained this impact by describing a scenario in which many people have a similar file. Deduplication would combine the information into one copy, but if this copy is lost, the information from everyone is lost.

To solve this problem, Rozier and the other authors of the paper suggested finding how to keep an additional copy in such a way that the copy is independent from the original. This way, the system is more reliable, but data can still be stored compactly.

Yahoo! sponsored the Best Paper Award at the symposium, and the winners were announced before the papers were presented at the event. Because of this, “we got a lot of attention at the conference,” Rozier said. The winners of Best Paper Awards also received 1,000 euros.

Rozier felt that he and the other co-authors received good recognition and that winning this award “was important in our field” because this “was a top notch conference,” he said.