Skip to main content

 
IBM Systems  > System p  > Solutions  > 

SAS Forecast Server Performance and Sizing for AIX on IBM System p5 Servers

  

Forecasting and predictive analytics have become increasingly important in today's highly competitive business environment increasing the demand for forecasting solutions. SAS Forecast Server, designed for organizations in any industry, answers the need for large-scale and highly analytic forecasting as well as automated forecasting tools. IBM's unmatched expertise in hardware and software technology along with services enables the SAS Forecast Server for AIX 5L™ on IBM POWER™ processor-based systems solution to be deployed on an infrastructure that is designed to improve reliability, performance and scalability across diverse forecasting projects commonly used in real-life production environments.

This white paper reviews how SAS Forecast Server performs on the IBM System p5™ 550Q Express server running the AIX® Version 5.3 operating system. The p5-550Q configured with 1.65 GHz POWER5+™ processors demonstrated exceptional performance on the SAS Forecast Server application based on testing performed by IBM and SAS in November 2006. 50,000 series were processed within 29 minutes, 500,000 series within 2 hours and 20 minutes, and 3,000,000 series within 15 hours in configurations ranging from two cores to eight cores. Test results include the two most prevalent forecasting operations - the "Select" operation determines the best analytical model to use based on the input data and the "Fit" function runs the model to produce the forecasts. In most organizations, the Select operation is performed less frequently, often only once a quarter, while the Fit function may be performed as often as daily or weekly. These results show how organizations can deploy the p5-550Q to deliver uncompromising performance and scaling capabilities across three diverse configurations to meet the demands of small, medium and large enterprise forecasting requirements. With these powerful, yet affordable entry servers, companies are now able to perform forecasting Fit functions more frequently and on more forecasting projects concurrently while enjoying fast turn-around times.


 
Full text white paper
View Acrobat
(323 KB) June 2007