H-Store
H-Store is an experimental database management system. It was designed for online [transaction processing] applications. H-Store was developed by a team at Brown University, Carnegie Mellon University, the Massachusetts Institute of Technology, and Yale University in 2007 by researchers Michael Stonebraker, Sam Madden, Andy Pavlo and Daniel Abadi.
Architecture
H-Store was promoted as a new class of parallel database management systems, called NewSQL, that provide the high-throughput and high-availability of NoSQL systems, but without giving up the transactional consistency of a traditional DBMS known as ACID. Such systems operate across multiple machines, as opposed to a single, more powerful, more expensive machine.H-Store is able to execute transaction processing with high throughput by forgoing many features of traditional relational database management systems.
H-Store was designed as a parallel system to run on a cluster of shared-nothing, main memory executor nodes. The database is partitioned into disjoint subsets each assigned to a single-threaded execution engine assigned to one core on one node. Each engine has exclusive access to all of the data in its partition. Because it is single-threaded, only one transaction at a time can access the data stored on that partition. No physical locks or latches are included in the system, and once a transaction is started, it cannot stall waiting for another transaction to complete. Throughput is increased by increasing the number of nodes in the system and reducing partition sizes.