Analytics vendor will bow retail and financial applications optimized for pre-integrated HP hardware.
SAS today announced plans to introduce, by year-end 2010, in-memory financial risk assessment and retail pricing and product-mix applications that will run on blade-based hardware grids from Hewlett Packard. The combination is aimed at the seemingly insatiable customer demand for faster analysis.
By combining its analytic application know-how with pre-tuned blade hardware from HP, SAS is promising high-performance computing aimed at complex, big-data problems. SAS High-Performance Risk, the financial application, handles portfolio stress testing and scenario analysis. SAS says a prototype slashed the time required to analyze a complex financial portfolio against 100,000 possible market states and two time horizons; what took 18 hours on a single server required only three minutes in the High-Performance Risk in-memory environment. The SAS High-Performance Markdown Optimization and SAS High-Performance Assortment Planning applications promise similar speed advantages for common retail analyses.
"Speed is everything in scoring models and in accelerating all the operations needed to develop, tune and deploy predictive models," observed James Kobielus, an analyst with Forrester Research. "We're going to see most advanced analytics being executed on in-memory architectures, and this is another sign that we're headed in that direction."
SAS bills the planned applications as its first in-memory offerings, but it can hardly be accused of taking a me-too approach. Unlike vendors including QlikTech, IBM Cognos (TM1), Spotfire, Microstrategy and Microsoft (with PowerPivot for Excel), which offer all-purpose in-memory analysis tools, the SAS offerings are application specific. On the surface they have have more in common with high-end, data-warehouse offerings now delivered or in the works among the likes of Teradata, Netezza, SAP, and IBM. But rather than speeding specific queries within a data warehouse, the SAS apps bring an entire application onto a high-performance computing platform.
Just last month, SAP announced plans for a High-Performance Analytic Appliance to be built on HP and IBM hardware. IBM's Smart Analytic System, introduced last year, is available with generic data-integration and business-intelligence modules. IBM has also announced, and has extensive plans to add, industry-specific application modules tuned and optimized to run on the hardware platform. This week Netezza announced plans for advanced i-Class analytic capabilities, and rivals including Aster Data Systems are also embedding more and more analytic processing capabilities into their databases.
SAS has partnered with Teradata, Netezza and Aster Data on in-database processing of SAS algorithms and analyses. But the applications announced today rely on in-memory rather than in-database processing.
"The in-database approach provides an edge in analytic data preparation whenever lots of data movement is required," explained Tapan Patel, a product marketing manager at SAS. "In-memory delivers speed and fast response in the model development stage. If computations and models can be stored in memory, different scenarios can then be tested and explored against those analytic result sets."
In the High-Performance Risk application, for example, thousands of different possible models of a portfolio can be quickly examined in the context of variables such as time, exchange rates, interest rates and other factors.
In other news from SAS today, the company announced it is acquiring Memex, a Glasgow, Scotland-based company focused on public safety, homeland security and intelligence-management solutions. SAS says the deal will enhance its portfolio of solutions and capabilities aimed at national security, intelligence, and law enforcement agencies. SAS also introduced an integrated Governance, Risk and Compliance platform. SAS Enterprise GRC is aimed at improved audit and contingency planning, and better strategy formulation and capital planning. SAS rivals SAP, Oracle and IBM all offer extensive GRC applications and add-on capabilities.