Michael Stonebraker has a history of database breakthroughs. He was a founder of Ingres, Illustra, Cohera and StreamBase Systems. He also served as CTO of Informix and helped develop Postgres. His most recent creation is Vertica, which just launched version 2.o of its database.
The claims around Vertica Analytic Database 2.0 are significant for querying data warehouses and data marts. According to the company, Vertica can access dozens of terabytes at lower cost than other solutions and respond to ad hoc queries based on real-time data in seconds.
The speeds comes from a shared-nothing, column architecture; data compression of up to 90 percent less table space; and concurrent loading and querying. The company claims that queries against terabyte databases are answered 50 to 200 times faster than competing row-oriented databases and specialized hardware.
According to Vertica, organizing data in columns of values from the same attribute, rather than as rows of tabular records results in faster performance for analytic databases. The result is that when a query needs to access just a few columns of a table, it only has to read those columns from the disk. In row-oriented databases, all the values in a table are usually read, consuming bandwidth.
Vertica has a few marquee customers so far, including Comcast, Level 3 and JP Morgan.
Stonebraker isn't done with creating new kinds of databases. He is working with a group of MIT/Brown/Yale scientists to completely rethink the OLTP database concept. Curt Monash has coverage of it in his DBMS2 blog.