The damage caused by natural disasters can be devastating, and the resulting distress is only heightened by the financial shock and unpredictability of such an event to the insurance industry. In order to provide greater clarity as to the frequency and severity of such catastrophes, insurance companies must use specialist software packages to predict and model the potential outcomes of major events such as flooding, hurricanes or even tsunamis. This allows them to provide more accurate pricing and assess the adequacy of their capital - thus protecting their businesses.
Oasis Loss Modelling Framework is one such software provider, and is the leading supplier of open architecture loss modelling for the global community. One challenge for Oasis was to demonstrate that their software would run quickly on a range of computing platforms. Their existing use of a Massively Parallel Processing (MPP) “appliance" supercomputer had shown that a proprietary and expensive version of HPC could run fast, but performance on industry-standard and open source servers was poor. Options for customers were therefore limited for the very large datasets that the latest catastrophe loss models require.
The Hartree Centre provided computational expertise to demonstrate that the Oasis platform would run on a distributed computing environment. The company used the Hartree Centre's HPC facilities, including the iDataplex Blue Wonder, alongside the expertise of STFC's computational scientists, to successfully demonstrate that the platform will scale up significantly on a distributed environment using MySQL. They also recommended that the Oasis team investigate memory-based calculations for the Oasis "kernel" processes.
Through this project, the Hartree Centre provided Oasis Loss Modelling Framework with a roadmap for future developments to enhance their offering. Specifically, Oasis has now developed a distributed version and will develop an in-memory version for its future releases. Faster, more efficient compute capability will benefit all users of their software.
“The independence and depth of knowledge enabled us to get world leading expertise to demonstrate the practicality of parallel processing the potential of an in-memory option."
Dr Peter Taylor
Oasis Loss Modelling Framework