USENIX supports diversity, equity, and inclusion and condemns hate and discrimination.
Redundancy Does Not Imply Fault Tolerance: Analysis of Distributed Storage Reactions to Single Errors and Corruptions
;login: Enters a New Phase of Its Evolution
For over 20 years, ;login: has been a print magazine with a digital version; in the two decades previous, it was USENIX’s newsletter, UNIX News. Since its inception 45 years ago, it has served as a medium through which the USENIX community learns about useful tools, research, and events from one another. Beginning in 2021, ;login: will no longer be the formally published print magazine as we’ve known it most recently, but rather reimagined as a digital publication with increased opportunities for interactivity among authors and readers.
Since USENIX became an open access publisher of papers in 2008, ;login: has remained our only content behind a membership paywall. In keeping with our commitment to open access, all ;login: content will be open to everyone when we make this change. However, only USENIX members at the sustainer level or higher, as well as student members, will have exclusive access to the interactivity options. Rik Farrow, the current editor of the magazine, will continue to provide leadership for the overall content offered in ;login:, which will be released via our website on a regular basis throughout the year.
As we plan to launch this new format, we are forming an editorial committee of volunteers from throughout the USENIX community to curate content, meaning that this will be a formally peer-reviewed publication. This new model will increase opportunities for the community to contribute to ;login: and engage with its content. In addition to written articles, we are open to other ideas of what you might want to experience.
We analyze how modern distributed storage systems behave in the presence of file-system faults such as data corruption and read and write errors. We characterize the behaviors of eight popular distributed storage systems, including Cassandra, Redis, and ZooKeeper. The major result of our study is that a single file-system fault introduced in one node of the cluster can induce catastrophic outcomes such as data loss, corruption, and unavailability. We find that most systems do not consistently use redundancy to recover from file-system faults. We also find that the above outcomes arise due to fundamental problems in file-system fault handling that are common across many systems. Our results have implications for the design of next generation fault-tolerant distributed storage systems.