Processing a series of complex ALM models, including full balance sheet simulations, the tests demonstrated scalability from 5,000, 10,000 and 20,000 interest rate simulations in a near linear fashion, thus showing Ambit BancWare ALM5 to be five times more scalable than its previous release.
In addition to scalability, a new usage model has been created in which Ambit BancWare ALM users can successfully achieve a targeted computation time, executing over SunGard's core distributed compute functionality or the DataSynapse compatible managed grid. The test results also showed that graphical interface load times of large models have dropped by more than 1,800%, memory used by the compute engine has been reduced by more than 70% and Ambit BancWare ALM5 has been proven compatible with Microsoft SQL Server 2008.
The testing was performed on a 512-core cluster at Intel's state-of-the-art data center lab in Dupont, Washington. The 64 computer servers were configured with Intel Quad-Core Xeon 5400 series processors. The three data base servers contained Intel Quad-Core Xeon 7300 series processors running Microsoft SQL Server 2008 Enterprise 64-bit (CTP6). Intel 9501 10 GB and 82563 1 GB Ethernet adaptors with IOAT acceleration were also used to optimize performance.
Andreas Hug, group vice president, global risk and compliance solutions for SunGardâs banks and corporations business, commented: âBecause financial institutions are adopting more sophisticated ways of managing balance sheets and increasing the frequency of reporting, they are demanding higher-performance in their ALM systems. Working with Microsoft, Intel and DataSynapse has helped SunGard to optimize Ambit BancWare ALM for multi-core processors, 64-bit technology and grid computing environments. The result is improved scalability, reduced load times and increased overall performance. We appreciate the assistance we received allowing us to determine that Ambit BancWare
ALM can help provide our customers with greater efficiency, lower costs and optimized system performance.â