The new KRM version 6.4 incorporates significant advances in speed due to the enhanced multithreading capability, which is an ability to run multiple processes concurrently on either Windows or UNIX operating systems. KRM now incorporates multi-threading for option and credit adjusted valuation, deterministic and stochastic interest rate risk simulation, multi-period stochastic credit portfolio management, single and multi-period value at risk, capital allocation, and stress testing. For most KRM processes, the speed improvement is almost proportional to the number of CPU processors used by the client machine, so that KRM now runs nearly four times faster on a four chip machine than the same process on a single chip machine. As part of the speed enhancement, Kamakura has also employed an advanced memory allocator used by organizations like AOL, British Telecom, Novell, and Reuters.
"KRM has now been significantly enhanced both in terms of speed and ease of use via an extensive link with Kamakura’s KRIS default probability service," said Warren Sherman, Kamakura President and Chief Operating Officer. "By combining a high speed Monte Carlo simulation with the full term structure of default probabilities available in KRIS, KRM users have a very powerful ability to simulate cash flows, losses, financial accruals, capital, and mark to market values on a multiperiod default-adjusted basis. The KRM users group of clients in 23 countries has provided excellent guidance on the way forward for KRM 6.4."
Version 6.4 of KRM also includes a number of other new features. Version 6.4 includes additional detail for the Basel II capital requirements of the Basel Committee on Banking Supervision. KRM version 6.4 also senses when various ‘stop loss’ triggers have been hit in a simulation, leading to a sale of the relevant security or portfolio. Version 6.4 includes application wide exception handling to detect errors in input data. Each “thread” in the multi-threaded application will evaluate the severity of the errors and take appropriate action and generate the appropriate error message or warning. KRM version 6.4 also includes the use of standard template libraries for efficient storage, retrieval and manipulation of in-memory data objects. Version 6.4 of KRM adds three new Monte Carlo simulation methodologies that can be selected by the user: randomized quasi-Monte Carlo sampling, Latin hypercube sampling, and antithetic sampling. Five new random number generators have been added to KRM version 6.4. The result is that clients can achieve significant reductions in sampling error in Monte Carlo simulations versus a traditional approach to Monte Carlo simulation. KRM version 6.4 also allows clients to audit for all KRM input and configuration data tables.