Abstract

Distributed key-value stores are now a standard component of high-performance web services and cloud computing applications. While key-value stores offer significant performance and scalability advantages, the first wave of NoSQL stores typically compromise on consistency, fault-tolerance, performance, or functionality, and sometimes on all four. This talk will present HyperDex, a novel, open-source, distributed key-value store developed by my group that provides (1) strong consistency guarantees, (2) fault-tolerance for failures and partitions affecting up to f nodes, and (3) a rich API which includes ACID transactions and a unique search primitive that enables queries on secondary attributes. HyperDex achieves these properties through the combination of three recent technical advances called hyperspace hashing, value-dependent chaining and linear transactions. Despite offering stronger guarantees than first-gen NoSQL data stores, HyperDex is also a factor of 2-13 faster than Cassandra and MongoDB. This talk will outline these techniques, identify new research directions, and discuss how these breakthroughs relate to the oft-quoted, but mostly misunderstood, CAP credo.