VoltDB, an open-source “next generation” database, graduated from beta last week. The company, under the leadership of Postgres and Ingres co-founder Mike Stonebraker, claims its product is extremely scalable and 45-times faster than traditional database products. VoltDB is targeting enterprises such as software-as-a-service providers, financial traders, online businesses or any organization that requires “systems with large, fast-growing transaction volumes.” VoltDB is one of many projects trying to re-invent the database.
After over 30 years of dominance, the traditional relational DB model is being challenged by several companies. Charles Babcock, writing for Information Week, compares VoltDB to Cassandra, CouchDB and MongoDB, three non-relational database projects. Babcock notes that unlike those projects, which aren’t designed as transactional systems, VoltDB will “always return the same answer to a query,” and preserves developers’ ability to use SQL.
Other VoltDB features include:
- Data is held in memory (instead of on disk) for maximum throughput, which eliminates buffer management.
- VoltDB distributes data – and an SQL engine to process it – to each CPU core in the server cluster.
- Each single-threaded partition operates autonomously, eliminating the need for locking and latching.
- Data is automatically replicated for intra-cluster high availability, which eliminates logging.
- VoltDB performance scales out near linearly by simply adding additional servers to the cluster.
Other database projects that store data in memory include Oracle‘s TimesTen and MySQL when using Memcached.
The recently funded startup FluidInfo‘s FluidDB is also experimenting with new approaches to the database.
Last year we asked whether relational database is doomed due to its problems scaling. VoltDB’s may save the relational database paradigm by solving the scalability issue and implementing innovative new features while retaining the advantages of the traditional model. Non-relational databases will probably continue to play a role in enterprise, but relational databases clearly aren’t dead yet.
