Compute Grid for mortage analytics cashflow simulations

Business Requirement

Having newly entered the US market, a leading investment bank planned on building its own infrastructure for running cashflow simulations for its new business. Starting on a clean slate, they were looking for options that were economical and dynamically scalable. More importantly, they were looking for options that could help them ramp up quickly without going through the whole exercise of building their own datacenter or entering long-term leases.


We were brought in to analyze their situation, recommend the best option, architect the solution and build the platform. Given their key requirements, a public cloud seemed to be the ideal solution. Amazon Web Services (AWS) was the chosen cloud platform. But, being a public cloud, there were new challenges such as security and compliance. We worked with AWS to ensure they were SAS-70 and FISAP complaint.

But, the biggest challenge came in the form of implementing the application that involved their complex algorithm for cashflow simulations of the entire mortgage-backed securities portfolio along various interest rate paths. In this application, interest rates were predicted to move along different paths over the following few months (a year or more) depending upon simulation of various factors. There was one run every night and for every run, there were thousands of such predicted paths. After generating the interest rate paths, every position in the portfolio was simulated along all these paths generating the corresponding cashflows. There were thousands of such positions. While generating these cashflows, additional data such as the indicative information for these securities and the historical data for the entire market were considered. The latter got into terabytes. Maximum number of servers required for one nightly run was close to one-hundred.

Our team comprised of a Project Manager, a lead architect, one security expert and some developers and testers. They worked closely with multiple developers, system administrators, DBAs, architects and senior leaders from the client side.


Because of the high, as well as, dynamic computing requirements of such cashflow simulations, the pay-as-you-go pricing model of the cloud turned out to be very economical. But, the biggest benefit came in the form of faster time-to-market. The client saved close to 4-6 months that would have otherwise taken them to build the infrastructure on their own – something highly critical to them because they were just entering a new market.