VoltDB and the Jepsen test: What we learned about data accuracy and consistency

VoltDB promises full ACID with strong serializability in a fault-tolerant, distributed SQL platform, as well as higher throughput than other systems that promise much less. But why should users believe this? John Hugg discusses VoltDB's internal testing and support processes, its work with Kyle Kingsbury on the VoltDB Jepsen testing project, and where VoltDB will continue to improve.

VoltDB promises full ACID with strong serializability in a fault-tolerant, distributed SQL platform, as well as higher throughput than other systems that promise much less. But why should users believe this? VoltDB has built a culture where data safety, correctness, and consistency are the highest priorities. John Hugg covers some specific scenarios where trade-offs have been made in support of these goals. Anyone can say their product makes no compromises, but it’s precisely the compromises that expose engineering values. John discusses VoltDB’s internal development, support, and testing processes and explores how VoltDB has improved over the years, how it tries to identify blindspots, and what kind it has planned for the future. John concludes by diving into third-party validation, with a focus on Kyle Kingsbury’s recent Jepsen test, which analyzed VoltDB and published the results in a nonbiased report. John offers an overview of this process and its ramifications. This session is sponsored by VoltDB.