“In-Database” Analytical Applications
MPP systems are particularly adapted to analytical processing of large volumes of data with complex datamodels. This sets MPP apart from the two more recent approaches: vector databases and big-table/hadoop.
MPP systems are perfectly suited to data warehousing requirements: complex patterns, massive loads, large volumes, scalability and high computing power (relative to the number of simultaneous queries). Some vendors now integrate native SQL calculation engines with advanced analytic functions for high performance data analytics. The distribution of CPU and storage also supports other computing paradigms such as map-reduce.
Greenplum and the democratization of MPP
In the world of MPP, our partner Greenplum holds a special position. Greenplum was the first company to offer an MPP architecture that was agnostic to hardware, whereas other players have chosen to design dedicated hardware systems, with specific adaptations for example at the network layer or in storage management.
Greenplum therefore can address a broad range of applications, from Tera to Peta levels, while lowering the cost of access to technology. It is now possible using standard multi-core CPU hardware and basic storage racks (JBOD DAS) to build a 1TB to 4TB MPP database system.
By doing that, Greenplum is probably the first publisher to democratize the MPP technology and make it accessible to many companies. Such an entry-level system allows for web companies to innovate through analytics…
(to be continued)