These are plans for a generic open source server that will support full static and stochastic balance sheet simulation for liquidity management and capital structure optimization. This generic simulation code will be a reference for a Consortium of banks and financial service firms to promote technical standards development, competitive floating-point performance, and computational stability in balance sheet simulation/optimization. LCR, NSFR, GSIB, TLAC, and CCAR preparations and rollouts through the next several years will be quite challenging to existing bank systems. There are GSIBs working on significant balance sheet system speed-ups just to control hardware costs. Some of the system speed-ups are mandatory simply to meet static simulation regulatory reporting requirements. Kinetix will maintain AWS NIMo Reference Servers for Firms to check their internal balance sheet simulation results against standard product models. Firm’s are welcome to use the reference server code internally, optimized to their production environment. Kinetix will also provide code and expertise to GSIBs, banks, financial service firms, and the FRB to assist in optimized simulation. Up-to-date competitive balance sheet simulation tools on commodity microprocessors should promote safety in the US Banking system. The Consortium will regularly solicit improvements to the basic design, technology standards, and quantitative analytic implementation of the NIMo Reference Server from Consortium members, academics, and practitioners by maintaining an Open Source policy. The Balance Sheet data, the operations models, the security models, and the detailed code optimizations required to make the balance sheet simulation’s performance competitive and relevant to each Firm’s capital planning and regulatory reporting, obviously remain wholly controlled by the respective Firm. The long-term goal is to automate the management of revenue margin (NIM) by adjusting business mix, product pricing, and funding through security/contract level Monte Carlo simulation and contemporary optimization algorithms assisting Firm Management.
The reason Firms work with the Consortium and Kinetix is raw simulation performance applied to their balance sheet. In 2016 the NIMo Reference Server runs a static worst-case 5-year full CCAR regression-fit simulation of a multi-trillion dollar balance sheet on a single commodity x86 processor in several seconds. If the Reference code is optimized it can run 10 to 20 times faster on a single microprocessor core, a few tenths of a second. That is important if there is need to run this balance sheet simulation in a Monte Carlo expected case framework. It is also important for:
- Internal Audit and Model Review
- Development Testing
- Market and Credit Risk Review
- External Regulatory Reporting
- Firm Management & Asset/Liability Committee Reporting
If a bank needs that kind of world-class competitive performance they run it with their internal High Performance Computing groups or work with the Consortium. The Consortium uses commodity x86 microprocessors, compilers, and development tools so the Firm’s simulations can be maintained and debugged by conventional means. Given a modest size multiprocessor running the optimized NIMo simulation code, the entire annual capital allocation plan for a GSIB can be automated and optimized with a dynamic feedback system.