FOCUS AREAS
Large Pharma Company Uses SAS Grid to Speed Optimization
Problem Description
Solution
Benefits
Problem Description
A large pharmaceutical company had several optimization jobs that they had developed using SAS. They were running these jobs on an NT server and they were taking entirely too long to complete. For example, one of the optimization jobs took over 24 hours to complete. One of the SAS experts at this site decided to try one of the SAS jobs on a large Unix server. Even though the execution time went down to approximately 15 hours, this customer was still unable to get the results he needed in a timely fashion to benefit his organization. So he decided to investigate what type of performance benefits he might achieve running these jobs in a SAS grid.
Solution
Initially, this customer picked one of the SAS optimization jobs that he thought would be a good candidate to be run on a SAS grid. The job could be easily decomposed into multiple independent units of work that could be distributed across the nodes in the grid and executed in parallel. He also decided to utilize SAS/CONNECT to do the distribution and management of the work across the grid. What makes this scenario especially interesting is that the hardware that was used to create the grid was a bunch of laptops otherwise considered obsolete. A dedicated grid of 60 laptops, ranging from 266-400 MHz, made up the initial grid.
Benefits
The grid ran an optimization of 600 sales territories in about 2 hours! This same job previously took 26 hours on an NT server and 15 hours on a powerful Unix server. In this instance, computer hardware considered obsolete and slow was able to be used to significantly benefit an operating arm of this customer's company so they could get their reporting done in a time-frame consistent with their needs. By utilizing existing hardware thought to be obsolete, this customer realized a very high return on investment and was able to do much more with less.
As a result of the large performance gains achieved with the implementation of the initial optimization job in their SAS grid, this pharma organization has since grown thier SAS grid to include 300 nodes and they have added 3 additional optimization jobs to run in this grid environment.