VaR is the percentile of loss (random variable) probability function. It determines the pobability of some extreme cases of loss. It's a function of time and probability required. Say Prob.{Loss > -VaR_alpha} = alpha.
There are many ways to estimate this measure. Historical simulation, Model Building, Monte Carlo Simulation.
In the industry, people use monte carlo simulation quite often. But one thing we have to pay attention to when doing MC simulation is the variance. There are bunch of papers discussing how to reduce the variance and save the computional time. Some methods people always mentioned are important sampling and stratified sampling etc.
If you are interested in this field, you can read check out some papers written by P.P. Boyle(he is the first guy who introduces monte carlo simulation method to asset pricing, Univ. of Waterloo), P. Glasserman(he has written many papers on this topic, Columbia Univ.) , M. Broadie(Columbia Univ.)