Turbocharging the Hardware
No data warehousing-related release would be complete without also addressing hardware. Teradata has oblidged, announced upgrades across its entire product line by upgrading to latest-generation dual Intel Xeon Westmere Processors with six computing cores.
Taking advantage of Intel hyper-threading, the new chips let Teradata deliver 24 virtual cores per compute server, matching processing power to physical drives and, according to Teradata, greatly enhancing performance.
Teradata's flagship Active Enterprise Data Warehouse 5650, for instance, is said to pack a 43% performance gain over the previous-generation product on a compute-server to compute-server bases (meaning, taking virtualization into account). The EDW platform now scales from 7 terabytes (Tbytes) to 86 Petabytes of uncompressed data, truly putting the "Big" in Big-Data analysis.
Other highlights of the hardware refresh include a compact new footprint for the Teradata Data Warehouse Appliance 2650 thanks to a move to 2.5-inch drives. The switch is said to deliver a 3.3-times performance improvement, with systems available from 2 to 275 Tbytes of uncompressed data.
The appliance upgrade also offers more granular workload management and a new software bundle including a free appliance backup utility. The utility sounds like a first response to the EMC Greenplum Data Computing Appliance, a recent launch featuring multiple backup and data recovery options.
The Teradata product line is rounded out by the Teradata Extreme Performance Appliance 4600, the Extreme Data Appliance 1650 and the Data Mart Appliance 560. The 4600 is all solid-state disk and is said to deliver 18 times the speed of conventional hard-drive based appliances. It's aimed at online retailers, financial institutions, network providers and others that demand the ultimate in real-time performance.
The 1650, which will be available in December, is designed for extremely large data sets and moderate performance. It might be used for historical analyses such as retrieving year-old call records to meet legal or regulatory requirements.
On both performance and scalability, Teradata's new hardware line will no doubt give competitors new benchmarks to target.
A Question of Depth
Teradata has been a follower rather than a leader in supporting MapReduce techniques and Hadoop, areas where Aster Data and EMC's Greenplum unit were early movers. Teradata is catching up through two partnerships aimed at sharing data and analyses across Hadoop and Teradata environments. MapReduce and Hadoop are often selected for analyzing large sets of loosely structured data, such as e-mail or social media data, or sparse data, with inconsistencies that aren't handled well by conventional relational databases.
On Monday Teradata unveiled a Plug-in for Eclipse that will enable developers to use an application development environment from Karmasphere to write MapReduce applications in drag-and-drop environment. These apps can tap into data in Teradata stores while handling MapReduce processing in Hadoop environments. Result sets can then be returned to Teradata for conventional relational analysis.
In a deal announced last month, Teradata partnered with software and services provider Cloudera to build an integration to that vendor's distribution of Hadoop. This, too, will enable passing of data and results between Teradata and Hadoop, but the free pipeline has yet to be released.
In the same way that Teradata stopped clinging to its EDW strategy and quickly introduced a broad and capable hardware family, the company now appears to be articulating a clear strategy for what had been a sprawling array of partnerships and initiatives. Netezza, which will set to be acquired by IBM, took similar steps to articulate its iClass analytics strategy.
The depth and breadth of the Teradata Accelerated Analytics portfolio is impressive and unmatched. Some of the individual pieces, such as Hadoop support, have yet to be fleshed out. But the strategy is there and execution is not an area where Teradata tends to stumble.