HPC from A-Z (part 9) - I


Like all financial services companies today, insurance companies (whether national or global in scale) are under constant pressure. The rise of consumer empowerment through the growth of online comparison sites means insurance companies’ websites need to manage increased volume of traffic, requiring additional processing power. Meanwhile, the reporting requirements from the regulatory environment are unrelenting. And this all comes against a backdrop of consumer uncertainty about the economic future and an incredibly competitive industry.

Recently I was impressed with the bold action a major international insurance customer, specializing in retail services, took to plan ahead for these types of challenges. The company wanted to support strategic decision-making around the complex issue of the ‘inherited estate’ by scaling out from a small, overloaded cluster to hundreds of machines. It also needed to manage the workload brought on by the strict UK regulatory environment – PriceWaterhouseCoopers had dauntingly explained it would need a ‘weapon of mass computation’ to achieve this.

HPC gave the company the ability to run up to 10 times more scenario models at once, providing it with the information needed to make a strategic decision not to reallocate inherited estate money. However, the overall strategic value gained is something which the wider insurance industry would benefit from considering.

While this customer’s approach is not unique, there are still many companies in the industry that have not implemented tools that allow them to improve the performance of their compute applications and hardware. And beyond this, there’s even more value to be derived from existing HPC infrastructure: for example, extending modelling to Individual Capital Assessments. These tools help firms identify major sources of risk within their business (credit, market, liquidity, operational and insurance risks) and perform appropriate stress scenario testing for each of the identified risks. Surely this is hugely desirable, if not essential, for the industry at large?


Post a Comment