Measuring Default Risk in a Parallel ALM Software for Life Insurance Portfolios

In this paper we investigate the computational issues in the use of a stochastic model - the doubly stochastic intensity default model - to measure default risk in the development of "internal models", according to the new rules of the Solvency II project. We refer to the valuation framework used in DISAR, an asset-liability management system for the monitoring of portfolios of "Italian style" profit sharing life insurance policies with minimum guarantees. The computational complexity of the overall valuation process requires both efficient numerical algorithms and high performance computing methodologies and resources. Then, to improve the performance, we apply to DISAR a parallelisation strategy based on the distribution of Monte Carlo simulations among the processors of a last generation blade server.