The application of new insurance and reinsurance regulation introduced by the European Directive 2009/138 (Solvency II) [4] leads to a complex valuation process to assess risks and determine the overall solvency needs. The development of an “internal model” – “a risk management system developed by an insurer to analyse its overall risk position, to quantify risks and to determine the economic capital required to meet those risks” [5] – generates hard computational problems. The perfect timing of measurements and consequent management actions must be further safeguard. It stands to reason that the computational performance of the valuation process plays a relevant role; this motivates the need to develop both accurate and efficient numerical algorithms and to use High Performance Computing (HPC) methodologies and resources. The literature on the application of HPC in the development of “internal model” is very poor; a relevant contribution is given in [1] where is introduced DISAR (Dynamic Investment Strategy with Accounting Rules), a Solvency II compliant system designed to work on a grid of conventional computers. In [2] numerical experiments carried out applying to DISAR a parallelisation strategy based on the distribution of Monte Carlo simulations among processors are reported. The developed parallel software is tested on an IBM Bladecenter using pure MPI implementation and treating every core as a separate entity with its own address space. Now, we show some of experiences in adding a layer of shared memory threading trying to optimize the application built using MPI and OpenMP. At this aim, we use some tools and techniques for tuning the hybrid MPI/OpenMP DISAR implementation.

Performance Analysis of an Hybrid MPI/OpenMP ALM Software for Life Insurance Policies on Multi-core Architectures

PERLA, Francesca;ZANETTI, Paolo
2012-01-01

Abstract

The application of new insurance and reinsurance regulation introduced by the European Directive 2009/138 (Solvency II) [4] leads to a complex valuation process to assess risks and determine the overall solvency needs. The development of an “internal model” – “a risk management system developed by an insurer to analyse its overall risk position, to quantify risks and to determine the economic capital required to meet those risks” [5] – generates hard computational problems. The perfect timing of measurements and consequent management actions must be further safeguard. It stands to reason that the computational performance of the valuation process plays a relevant role; this motivates the need to develop both accurate and efficient numerical algorithms and to use High Performance Computing (HPC) methodologies and resources. The literature on the application of HPC in the development of “internal model” is very poor; a relevant contribution is given in [1] where is introduced DISAR (Dynamic Investment Strategy with Accounting Rules), a Solvency II compliant system designed to work on a grid of conventional computers. In [2] numerical experiments carried out applying to DISAR a parallelisation strategy based on the distribution of Monte Carlo simulations among processors are reported. The developed parallel software is tested on an IBM Bladecenter using pure MPI implementation and treating every core as a separate entity with its own address space. Now, we show some of experiences in adding a layer of shared memory threading trying to optimize the application built using MPI and OpenMP. At this aim, we use some tools and techniques for tuning the hybrid MPI/OpenMP DISAR implementation.
2012
978-3-642-30960-1
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11367/30855
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 1
  • ???jsp.display-item.citation.isi??? ND
social impact