Elastic cloud resources provisioning for life insurance undertaking applications

Andrea La Rizza

Abstract:
The Solvency II Directive (Directive 2009/138/EC) [14] is a European law adopted in November 2009, and amended by Directive 2014/51/EU of the European Parliament and of the Council of 16 April 2014 (the so-called "Omnibus II Directive"). It was enacted by the European Union to regulate the insurance sector through risk management. Solvency II requires that European insurance companies conduct consistent evaluations and constant monitoring of risk. The main regulation requires the insurance undertaking to compute the SCR (Solvency Capital Requirement) and the pdf (Probability Distribution Forecast). The SCR is the capital amount that should hedge the losses up to a given level. To calculate the SCR the directive allows the company to choose between two different computational methodology: the standard formula and the internal model. The pdf is a mathematical function that assigns to an exhaustive set of mutually exclusive future events a probability of realisation. Its computation relies on non closed form calculus based on several Monte Carlo simulations and different hypothesis.
To fulfill the directives, companies should equip themselves with particular computational systems which, due to their complexity, require for their execution significant underlying IT infrastructures. This IT equipments represent a significant outlay for the company, not only for what concerns their direct costs due to the purchase, but also those for power consumption, cooling, maintenance and upgrades. Moreover, the technological progress that characterizes the IT field makes these computing resources become obsolete quickly. On the other hand, we face the risk that such resources become underused since these procedures could be used only periodically (at most monthly). If for the insurance company the utilization of "on premises" resources may not be advantageous, for a cloud services provider, increasing steadily its resources, leads to reduce the costs and consequently reduce the rates charged to the user.
The idea of this work is to put together the needs arising from Solvency II accomplishment with the always more advantageous scenario of cloud services. Within the project, as a case study, the focus is on a particular "module" of a more complex system named DISAR® 1, that deals with the calculation of the pdf and other significant items for an insurance company. This procedure relies on a Monte Carlo simulation algorithm developed on parallel MPI environment. The goal of the project is to develop a framework that, by exploiting a learning algorithm is able to: calculate the computational power and the size of the IT infrastructure necessary for the execution of the procedure for a given input, deploy and make it available in a few minutes on the cloud, transfer to the grid the data necessary to the simulation, execute it and retrieve the output in a fully automated way transparent to the final user. The latter has the only task of choosing the segregated funds the to work on. Moreover after each elaboration, the data relative to it are stored and subsequently utilized to improve the model for the next simulation. The cloud services provider adopted for this project is Amazon Web Services (AWS), however the core algorithm as the methodology can be applied independently of the provider.

BibTeX Entry: