Teradata's new Extreme Data Appliance 1550 is designed for high-data-scale applications that are characterized by focused queries, departmental scale and not-so-time-sensitive querying. Examples include Web site clickstream analysis, multi-year regulatory compliance, manufacturing processing and testing, RFID-product movement and cell phone network usage. Many of these apps were heretofore viewed as impractical, or they were relegated to server farms and flat-file processing. The 1550 combines one-terabyte hard drives with Intel quad-core nodes and built-in data protection software to offer extreme storage density starting at a list price of $16,500 per terabyte.
"This gives many people an option to get [vast sets of data] into a relational format and start doing analytics on it for discovery and huge data sifting," says Scott Gnau, Teradata's Chief Development Officer. "It's for applications in which performance isn't extremely important, but ease of use and ease of integration will be important."
Teradata 13, the latest upgrade of the vendor's core database, is said to boost performance in several ways. New data extract, load and transform tools, for instance, are said to speed loading while workload optimization has accelerated OLAP query performance by as much as 30 percent. Perhaps the biggest news in the database upgrade is Teradata Virtual Storage, an innovation that enables customers to manage heterogeneous drive types as a virtual storage pool.
"In parallel processing environments you always need the system to be balanced, because the slowest thread will dictate how fast the overall warehouse performs," Gnau explains. "With Virtual Storage, you can mix drive sizes, say 73-GB and 300-GB drives, in the same storage pool, and the software will automatically move the data, based on service-level requirements and the frequency of access, among the hot and cold [meaning fast and slow] areas of available storage."
Thus, the new Virtual Storage capability lets Teradata customers mix generations of devices without crimping overall performance. Low-demand data will be moved to the inner tracks of the slower, fatter drives while the high-demand data will be migrated to the faster, outer tracks of the higher performance drives.
Teradata's prototype SSD-based appliance appearing in Las Vegas this week is strictly for demonstration purposes. SSDs are still far too expensive to go into a production data warehousing device, but Gnau says they may be more practical by 2010 or 2012. The concept device looks like Teradata's 2550 appliance, but inside are 128-GB SSDs rather than spinning disks. Currently making their way into laptops and ultraportable devices, SSDs promise 5 to 10 times the performance and an order-of-magnitude lower power consumption than conventional drives with motors and moving parts.