Jun 30, 2010 (03:06 AM EDT)
Image Gallery: 11 Leading Data Warehousing Appliances
Read the Original Article at InformationWeek
Teradata won't call its flagship enterprise data warehouse an appliance, but it was a pioneering product in pre-integrating all the technology needed for a data warehouse. Over the last three years, Teradata has released a range of niche appliances, including this solid-state-disk Extreme Performance Appliance 4555.
Netezza led the mid-decade appliance boom by offering a lower-cost, higher-performance alternative to conventional data warehouses. Netezza last year moved off proprietary hardware to the TwinFin platform illustrated here. The company now claims more than 300 customers and pricing under $20,000 per terabyte.
Greenplum also fueled the mid-decade boom, though now as then it provides a database that's ready to run on suggested configurations of third-party hardware. The key to performance is massively parallel processing (MPP) on low-cost commodity hardware, an approach also embraced by Kognitio and others.
Oracle, the database and data warehouse marketshare leader, finally responded to appliance competition with the Exadata in 2008. Oracle soon ditched hardware partner HP and introduced Sun Oracle Exadata V2 in 2009. The update delivers faster query speeds by way of Sun's solid-state-memory F5100 Flash Array.
The IBM Smart Analytics System 7600 is one of three pre-integrated appliances introduced in 2009. In addition to the usual combo of database, servers and storage (all from IBM), the system also offers optional information management, BI, analytics and vertical-industry application modules that can be preinstalled.
HP went after the top end of the data warehousing market by introducing Neoview in 2007. Aimed at enterprises with tens of thousands of users, tens of thousands of simultaneous queries, and hundreds of terabytes to petabytes of data, Neoview targets would-be Teradata users. Customer wins have been few.
Sybase IQ is the top column-store DBMS for data warehousing with more than 2,500 customers. Column-store databases have compression and speed advantages over row-store DBMS (like Oracle and MS SQL Server) when querying against a few, columnar attributes. IQ runs on symmetric multiprocessor hardware.
Vertica is Sybase IQ's primary competitor. As shown here, the column-oriented analytic database runs on suggested configurations of third-party hardware and exploits an MPP (rather than Sybase IQ's SMP) approach. ParAccel also offers a column-store database supporting MPP on third-party hardware.
There's still room for innovation. Latecomer Aster Data Systems (founded in '05) offers a MapReduce appliance, partner software modules and deep support for in-database analytic processing. MapReduce is a framework for processing huge data sets across many compute nodes.
SAP has expanded on the in-memory and column-store database technology behind its SAP BW Accelerator to deliver SAP BusinessObjects Explorer, a tool that can quickly explore large data sets across multiple sources. The same technology will drive a High-Performance Analytic Appliance set to debut by year end.
Microsoft plans to debut its SQL Server 2008 R2 Parallel Data Warehouse (PDW) by year-end. The edition will support massively parallel processing on HP or IBM hardware. It will also support hub-and-spoke integration of marts and satellite data warehouses. Don't expect exotic flash memory options or petabyte scalability.