Project staff and support

Shilan Mistry (Intern, University of Oxford)
Jurgen Gaiser-Porter (Company Supervisor, Willis)
Patrick McSharry (Academic Mentor, University of Oxford)
Tristram Armour (Technology Translator, Industrial Mathematics KTN)

This Internship project was carried out at Willis, in conjunction with the University of Oxford. It is part of the KTN's Industrial Mathematics Internships Programme and funded by NERC. The project is also part of the Environmental Risk Management Special Interest Group funded by NERC. Start date: October 2009; duration: 3 months.

Project overview

Reinsurance provides protection for insurance companies against the risk of losses including catastrophic events such as earthquakes, floods and wind storms. Quantification of the associated risk relies on intensive simulations and the optimisation of complex contracts. The challenge of obtaining these calculations could benefit greatly from recent developments in parallel computing.

Willis is currently investing significant resources in the development of the next generation of catastrophe models for support of the transactional business. Willis have a new experimental prototype tool for optimising the placement strategy of client reinsurance programmes. This is particularly computationally demanding task demanding running tens of thousands of different strategies. However, timetables for their clients to make certain decisions about reinsurance placements can be measured in days so having fast tools is key.

Project outcomes

The results indicate a significant increase in speed of the optimisation calculation. A test contract with 5 layers and 10000 simulations can be optimised in 2 minutes using 36 standard desktops, where a naive, nonparallel approach could take up to 50 minutes. The behaviour of the system under multiple users was explored and it was noticed that with an appropriate number of client computers, a robust system with reasonable waiting times for job completion was made.

By using Nereus, Willis is harnessing available, idle computer power and so the cost is negligible. A scalable platform to develop parallel algorithms was created. As a large amount of scientific computing at Willis is done in Java, this makes the transition to parallel computing simple. For example, problems that are embarrassingly parallel such as Monte-Carlo simulations, can be trivally mapped to the Nereus system. Should more computing power be required, the Nereus network can be expanded by simply connecting more client computers.

"At Willis Re, we are were looking at a financial modelling problem, for which we had a prototype solution. It would take several hours to run and we were trying to find ways for a faster throughput so that the calculation could be run regularly. The algorithm we were trying to implement contains several mathematical algorithms hence we were very pleased to take on Shilan Mistry, a mathematics PhD student for this project. With his mathematical background and also using the expertise of his Oxford research group he has provided us with a possible solution which speeds up the runtime by several factors. We are now analysing Shilan's findings and hope to make a decision soon how to implement his solution," said industrial supervisor Jürgen Gaiser-Porter, Willis Re.

"While the progress of science and mathematics is an academic pursuit, the real test of science is the ability to implement and apply these new ideas and methodologies to actual problems. Through the collaboration with Willis we have be able to establish connections between techniques and methodologies developed within the University of Oxford and the needs of the insurance sector to address problems of both a scientific and industrial nature. Through the KTN internship I have had the opportunity to engage with practitioners in the field and focus the scope of my research to deal with problems relevant to industry," said intern Shilan Mistry, University of Oxford.