The VaR is a statistical measure and is calculated with a simulation. This can either be a Monte Carlo simulation, which generates an artificial distribution, or a historical simulation, which uses historical data to generate a distribution. The calculation is lengthy because there is a lot of pricing of financial derivatives involved, which are expensive calculations. Another problem is the data-access; this can become the bottleneck for certain situations. Because this is such a lengthy calculation, it is done in parallel on a multi-processor environment.
Essentially, there are two different VaR calculations: a complete VaR calculation for the entire portfolio of the bank, and an incremental calculation, which tracks the development of the VaR during the day. These two different types of calculation require a different load balance.
The task in this internship is researching the structure and execution of the two types of calculation. Based on that research, an attempt will be made to improve the calculation structure and the load-balance, in order to make the calculation more efficient and faster.